2020 |
Fernando, Heshan; Marshall, Joshua What lies beneath: Material classification for autonomous excavators using proprioceptive force sensing and machine learning Journal Article 119 , pp. 103374, 2020, ISSN: 0926-5805. Abstract | Links | BibTeX | Tags: Force sensing, Loader automation, Machine Learning, Material classification, Robotic excavation @article{FERNANDO2020103374, title = {What lies beneath: Material classification for autonomous excavators using proprioceptive force sensing and machine learning}, author = {Heshan Fernando and Joshua Marshall}, url = {http://www.sciencedirect.com/science/article/pii/S0926580520309547}, doi = {https://doi.org/10.1016/j.autcon.2020.103374}, issn = {0926-5805}, year = {2020}, date = {2020-09-04}, volume = {119}, pages = {103374}, abstract = {The ability of robotic excavators to acquire meaningful knowledge about materials during digging can augment their autonomous functionality, as well as optimize downstream operations in construction and mining. Some material properties, such as rock sizes, can be determined visually, but these methods cannot see what lies beneath. In this work, a classification methodology that utilizes only proprioceptive force data acquired from an autonomous digging system and machine learning algorithms is proposed for excavation material identification. The consistent performance synonymous with autonomous digging systems allows for the use of basic features extracted from the force data for classification. A proof of concept of this novel approach to excavation material classification is demonstrated through a binary classification of rock and gravel materials. Force data were obtained from full-scale autonomous loading trials with a 14-tonne capacity load-haul-dump machine at a mining and construction test facility. Preliminary results achieved a classification accuracy of 90%.}, keywords = {Force sensing, Loader automation, Machine Learning, Material classification, Robotic excavation}, pubstate = {published}, tppubtype = {article} } The ability of robotic excavators to acquire meaningful knowledge about materials during digging can augment their autonomous functionality, as well as optimize downstream operations in construction and mining. Some material properties, such as rock sizes, can be determined visually, but these methods cannot see what lies beneath. In this work, a classification methodology that utilizes only proprioceptive force data acquired from an autonomous digging system and machine learning algorithms is proposed for excavation material identification. The consistent performance synonymous with autonomous digging systems allows for the use of basic features extracted from the force data for classification. A proof of concept of this novel approach to excavation material classification is demonstrated through a binary classification of rock and gravel materials. Force data were obtained from full-scale autonomous loading trials with a 14-tonne capacity load-haul-dump machine at a mining and construction test facility. Preliminary results achieved a classification accuracy of 90%. |
Kudrinko, K; Flavin, E; Zhu, X; Li, Q Wearable Sensor-Based Sign Language Recognition: A Comprehensive Review Journal Article IEEE Reviews in Biomedical Engineering, pp. 1-1, 2020, ISSN: 1941-1189. Abstract | Links | BibTeX | Tags: Gesture recognition;Assistive technology;Cameras;Auditory system;Three-dimensional displays;Sociology;Statistics;wearable sensors;sign language;gesture recognition;machine learning;sensor systems @article{9178440, title = {Wearable Sensor-Based Sign Language Recognition: A Comprehensive Review}, author = {K Kudrinko and E Flavin and X Zhu and Q Li}, doi = {10.1109/RBME.2020.3019769}, issn = {1941-1189}, year = {2020}, date = {2020-08-26}, journal = {IEEE Reviews in Biomedical Engineering}, pages = {1-1}, abstract = {Sign language is used as a primary form of communication by many people who are Deaf, deafened, hard of hearing, and non-verbal. Communication barriers exist for members of these populations during daily interactions with those who are unable to understand or use sign language. Advancements in technology and machine learning techniques have led to the development of innovative approaches for gesture recognition. This literature review focuses on analyzing studies that use wearable sensor-based systems to classify sign language gestures. A review of 72 studies from 1991 to 2019 was performed to identify trends, best practices, and common challenges. Attributes including sign language variation, sensor configuration, classification method, study design, and performance metrics were analyzed and compared. Results from this literature review could aid in the development of user-centred and robust wearable sensor-based systems for sign language recognition.}, keywords = {Gesture recognition;Assistive technology;Cameras;Auditory system;Three-dimensional displays;Sociology;Statistics;wearable sensors;sign language;gesture recognition;machine learning;sensor systems}, pubstate = {published}, tppubtype = {article} } Sign language is used as a primary form of communication by many people who are Deaf, deafened, hard of hearing, and non-verbal. Communication barriers exist for members of these populations during daily interactions with those who are unable to understand or use sign language. Advancements in technology and machine learning techniques have led to the development of innovative approaches for gesture recognition. This literature review focuses on analyzing studies that use wearable sensor-based systems to classify sign language gestures. A review of 72 studies from 1991 to 2019 was performed to identify trends, best practices, and common challenges. Attributes including sign language variation, sensor configuration, classification method, study design, and performance metrics were analyzed and compared. Results from this literature review could aid in the development of user-centred and robust wearable sensor-based systems for sign language recognition. |
Davoodnia, Vandad; Slinowsky, Monet; Etemad, Ali Deep multitask learning for pervasive BMI estimation and identity recognition in smart beds Journal Article Journal of Ambient Intelligence and Humanized Computing, pp. 1–15, 2020. Abstract | Links | BibTeX | Tags: Artificial Intelligence, Healthcare, Internet of Things @article{davoodnia2020deep, title = {Deep multitask learning for pervasive BMI estimation and identity recognition in smart beds}, author = { Vandad Davoodnia and Monet Slinowsky and Ali Etemad}, doi = {10.1007/s12652-020-02210-9}, year = {2020}, date = {2020-06-26}, journal = {Journal of Ambient Intelligence and Humanized Computing}, pages = {1--15}, publisher = {Springer}, abstract = {Smart devices in the Internet of Things (IoT) paradigm provide a variety of unobtrusive and pervasive means for continuous monitoring of bio-metrics and health information. Furthermore, automated personalization and authentication through such smart systems can enable better user experience and security. In this paper, simultaneous estimation and monitoring of body mass index (BMI) and user identity recognition through a unified machine learning framework using smart beds is explored. To this end, we utilize pressure data collected from textile-based sensor arrays integrated onto a mattress to estimate the BMI values of subjects and classify their identities in different positions by using a deep multitask neural network. First, we filter and extract 14 features from the data and subsequently employ deep neural networks for BMI estimation and subject identification on two different public datasets. Finally, we demonstrate that our proposed solution outperforms prior works and several machine learning benchmarks by a considerable margin, while also estimating users’ BMI in a 10-fold cross-validation scheme.}, keywords = {Artificial Intelligence, Healthcare, Internet of Things}, pubstate = {published}, tppubtype = {article} } Smart devices in the Internet of Things (IoT) paradigm provide a variety of unobtrusive and pervasive means for continuous monitoring of bio-metrics and health information. Furthermore, automated personalization and authentication through such smart systems can enable better user experience and security. In this paper, simultaneous estimation and monitoring of body mass index (BMI) and user identity recognition through a unified machine learning framework using smart beds is explored. To this end, we utilize pressure data collected from textile-based sensor arrays integrated onto a mattress to estimate the BMI values of subjects and classify their identities in different positions by using a deep multitask neural network. First, we filter and extract 14 features from the data and subsequently employ deep neural networks for BMI estimation and subject identification on two different public datasets. Finally, we demonstrate that our proposed solution outperforms prior works and several machine learning benchmarks by a considerable margin, while also estimating users’ BMI in a 10-fold cross-validation scheme. |
Mitchell, Jordan; Marshall, Joshua Autorotating unmanned aerial vehicle surveying platform Patent 2020, (US Patent App. 16/046,436). Abstract | Links | BibTeX | Tags: Mapping, Signal Processing, Unmanned Aerial Vehicles @patent{mitchell2019autorotating, title = {Autorotating unmanned aerial vehicle surveying platform}, author = { Jordan Mitchell and Joshua Marshall}, url = {https://patents.google.com/patent/US20190031342A1/en}, year = {2020}, date = {2020-06-09}, abstract = {An autorotating unmanned aerial vehicle (UAV) has a data acquisition system and a rotor assembly including a hub that couples the rotor assembly to the UAV. Although not limited thereto, the UAV is suitable for collecting data about the inside of a cavity. The data acquisition system includes a processor and one or more sensors that obtain data about motion of the UAV and at least one parameter of the cavity as the UAV descends though the cavity. Features of the cavity may be mapped by generating a 3D point cloud from the data. The cavity may be natural or man-made, such as a mine.}, note = {US Patent App. 16/046,436}, keywords = {Mapping, Signal Processing, Unmanned Aerial Vehicles}, pubstate = {published}, tppubtype = {patent} } An autorotating unmanned aerial vehicle (UAV) has a data acquisition system and a rotor assembly including a hub that couples the rotor assembly to the UAV. Although not limited thereto, the UAV is suitable for collecting data about the inside of a cavity. The data acquisition system includes a processor and one or more sensors that obtain data about motion of the UAV and at least one parameter of the cavity as the UAV descends though the cavity. Features of the cavity may be mapped by generating a 3D point cloud from the data. The cavity may be natural or man-made, such as a mine. |
Sreedharan, Sarath; Chakraborti, Tathagata; Muise, Christian; Khazaeni, Yasaman; Kambhampati, Subbarao --D3WA+--A Case Study of XAIP in a Model Acquisition Task for Dialogue Planning Inproceedings Proceedings of the International Conference on Automated Planning and Scheduling, pp. 488–497, 2020. Abstract | Links | BibTeX | Tags: Artificial Intelligence, Human Machine Interaction, Natural Language Processing, Planning @inproceedings{sreedharan2020d3wa+, title = {--D3WA+--A Case Study of XAIP in a Model Acquisition Task for Dialogue Planning}, author = { Sarath Sreedharan and Tathagata Chakraborti and Christian Muise and Yasaman Khazaeni and Subbarao Kambhampati}, url = {https://www.aaai.org/ojs/index.php/ICAPS/article/view/6744}, year = {2020}, date = {2020-06-01}, booktitle = {Proceedings of the International Conference on Automated Planning and Scheduling}, volume = {30}, pages = {488--497}, abstract = {Recently, the D3WA system was proposed as a paradigm shift in how complex goal-oriented dialogue agents can be specified by taking a declarative view of design. However, it turns out actual users of the system have a hard time evolving their mental model and grasping the imperative consequences of declarative design. In this paper, we adopt ideas from existing works in the field of Explainable AI Planning (XAIP) to provide guidance to the dialogue designer during the model acquisition process. We will highlight in the course of this discussion how the setting presents unique challenges to the XAIP setting, including having to deal with the user persona of a domain modeler rather than the end-user of the system, and consequently having to deal with the unsolvability of models in addition to explaining generated plans.}, keywords = {Artificial Intelligence, Human Machine Interaction, Natural Language Processing, Planning}, pubstate = {published}, tppubtype = {inproceedings} } Recently, the D3WA system was proposed as a paradigm shift in how complex goal-oriented dialogue agents can be specified by taking a declarative view of design. However, it turns out actual users of the system have a hard time evolving their mental model and grasping the imperative consequences of declarative design. In this paper, we adopt ideas from existing works in the field of Explainable AI Planning (XAIP) to provide guidance to the dialogue designer during the model acquisition process. We will highlight in the course of this discussion how the setting presents unique challenges to the XAIP setting, including having to deal with the user persona of a domain modeler rather than the end-user of the system, and consequently having to deal with the unsolvability of models in addition to explaining generated plans. |
von Tiessenhausen, Johann; Artan, Unal; Marshall, Joshua; Li, Qingguo Hand gesture-based control of a front-end loader Journal Article 2020. Abstract | Links | BibTeX | Tags: Human Machine Interaction, Mining, Wearables @article{vonhand, title = {Hand gesture-based control of a front-end loader}, author = { Johann von Tiessenhausen and Unal Artan and Joshua Marshall and Qingguo Li}, url = {http://hdl.handle.net/1974/27726}, year = {2020}, date = {2020-05-01}, publisher = {IEEE}, abstract = {In this paper, we present the design and use of an instrumented glove consisting of a 9-DOF inertial measurement unit (IMU) and resistive flex sensors. The glove is used as a unique human-machine interface to control a Kubota R520s front-end loader, through input gestures, for the excavation of a fragmented rock pile. Raw sensor data from the glove is recorded and transmitted to a computer for gesture recognition. Recognized gestures are then used to command the loader to switch between dig states and control the excavation process. The system allows an operator to observe the entire process from beside the loader, providing them with valuable information about interactions between the loader bucket and rock pile not usually available when seated in the vehicle's cab. Preliminary experiments show that a novice operator was able to improve their performance by using the proposed system, evaluated based on metrics of total and dig completion times, as well as payload.}, keywords = {Human Machine Interaction, Mining, Wearables}, pubstate = {published}, tppubtype = {article} } In this paper, we present the design and use of an instrumented glove consisting of a 9-DOF inertial measurement unit (IMU) and resistive flex sensors. The glove is used as a unique human-machine interface to control a Kubota R520s front-end loader, through input gestures, for the excavation of a fragmented rock pile. Raw sensor data from the glove is recorded and transmitted to a computer for gesture recognition. Recognized gestures are then used to command the loader to switch between dig states and control the excavation process. The system allows an operator to observe the entire process from beside the loader, providing them with valuable information about interactions between the loader bucket and rock pile not usually available when seated in the vehicle's cab. Preliminary experiments show that a novice operator was able to improve their performance by using the proposed system, evaluated based on metrics of total and dig completion times, as well as payload. |
Houmanfar, Roshanak; Etemad, Seyed Ali; MacEachern, Leonard; Klibanov, Mark Systems, methods and devices for activity recognition Patent 2020, (US Patent 10,575,760). Abstract | Links | BibTeX | Tags: Artificial Intelligence, Predictive Model, Wearables @patent{houmanfar2020systems, title = {Systems, methods and devices for activity recognition}, author = { Roshanak Houmanfar and Seyed Ali Etemad and Leonard MacEachern and Mark Klibanov}, url = {https://patents.google.com/patent/US10575760B2/en}, year = {2020}, date = {2020-03-03}, publisher = {Google Patents}, abstract = {Systems, methods and devices for recognizing user activity using data from accelerometer or other sensors. “Feature-based” approach and “model-based” approaches are described. In a feature-based approach, various values are extracted from input signals and projected onto a space that is selected to facilitate better segregation of data points. Classifiers identify the regions in this projected space in which the data points fall to distinguish between the different activity types. In a “model-based” approach, a generative model is trained for each activity type. Different activity types can be distinguished by identifying similarities between the input data with the generative models.}, note = {US Patent 10,575,760}, keywords = {Artificial Intelligence, Predictive Model, Wearables}, pubstate = {published}, tppubtype = {patent} } Systems, methods and devices for recognizing user activity using data from accelerometer or other sensors. “Feature-based” approach and “model-based” approaches are described. In a feature-based approach, various values are extracted from input signals and projected onto a space that is selected to facilitate better segregation of data points. Classifiers identify the regions in this projected space in which the data points fall to distinguish between the different activity types. In a “model-based” approach, a generative model is trained for each activity type. Different activity types can be distinguished by identifying similarities between the input data with the generative models. |
Salari, Omid; Zaad, Keyvan Hashtrudi; Bakhshai, Alireza; Jain, Praveen Reconfigurable Hybrid Energy Storage System for an Electric Vehicle DC/AC Inverter Journal Article IEEE Transactions on Power Electronics, 2020, ISSN: 1941-0107. Abstract | Links | BibTeX | Tags: Controls, Energy Storage @article{salari2020reconfigurable, title = {Reconfigurable Hybrid Energy Storage System for an Electric Vehicle DC/AC Inverter}, author = { Omid Salari and Keyvan Hashtrudi Zaad and Alireza Bakhshai and Praveen Jain}, doi = {10.1109/TPEL.2020.2993783}, issn = {1941-0107}, year = {2020}, date = {2020-01-01}, journal = {IEEE Transactions on Power Electronics}, publisher = {IEEE}, abstract = {Hybrid energy storage systems using battery packs and super capacitor (SC) banks are gaining considerable attraction in Electric Vehicle (EV) applications. Conventionally, such hybrid systems have been exploited using massive high-power DC/DC converters with bulk magnetic elements. Recently, switch-capacitor Multi-Source Inverters (MSIs) have been introduced for such applications. In this paper a new modular reconfigurable MSI is proposed for active control of energy storage systems in EV applications. Utilizing MSIs, reduces the weight and the volume of the power electronics interface and offers simple control. Along with the proposed MSI, a Space Vector Modulation (SVM) technique and a deterministic State of Charge (SOC) controller are also introduced for control of the switching actions and the operation of the SC bank. Simulations using MATLAB/Simulink and experimental results on a scaled down lab prototype are studied to assess the concepts.}, keywords = {Controls, Energy Storage}, pubstate = {published}, tppubtype = {article} } Hybrid energy storage systems using battery packs and super capacitor (SC) banks are gaining considerable attraction in Electric Vehicle (EV) applications. Conventionally, such hybrid systems have been exploited using massive high-power DC/DC converters with bulk magnetic elements. Recently, switch-capacitor Multi-Source Inverters (MSIs) have been introduced for such applications. In this paper a new modular reconfigurable MSI is proposed for active control of energy storage systems in EV applications. Utilizing MSIs, reduces the weight and the volume of the power electronics interface and offers simple control. Along with the proposed MSI, a Space Vector Modulation (SVM) technique and a deterministic State of Charge (SOC) controller are also introduced for control of the switching actions and the operation of the SC bank. Simulations using MATLAB/Simulink and experimental results on a scaled down lab prototype are studied to assess the concepts. |
Agarwal, Shubham; Muise, Christian; Agarwal, Mayank; Upadhyay, Sohini; Tang, Zilu; Zeng, Zhongshen; Khazaeni, Yasaman TraceHub-A Platform to Bridge the Gap between State-of-the-Art Time-Series Analytics and Datasets. Inproceedings AAAI, pp. 13600–13601, 2020. Links | BibTeX | Tags: Artificial Intelligence, Planning, Signal Processing @inproceedings{agarwal2020tracehub, title = {TraceHub-A Platform to Bridge the Gap between State-of-the-Art Time-Series Analytics and Datasets.}, author = { Shubham Agarwal and Christian Muise and Mayank Agarwal and Sohini Upadhyay and Zilu Tang and Zhongshen Zeng and Yasaman Khazaeni}, url = {https://www.aaai.org/Papers/AAAI/2020GB/DEMO-AgarwalS.556.pdf}, year = {2020}, date = {2020-01-01}, booktitle = {AAAI}, pages = {13600--13601}, keywords = {Artificial Intelligence, Planning, Signal Processing}, pubstate = {published}, tppubtype = {inproceedings} } |
Asai, Masataro; Muise, Christian Learning Neural-Symbolic Descriptive Planning Models via Cube-Space Priors: The Voyage Home (to STRIPS) Journal Article arXiv preprint arXiv:2004.12850, 2020. Abstract | Links | BibTeX | Tags: Artificial Intelligence, Planning @article{asai2020learning, title = {Learning Neural-Symbolic Descriptive Planning Models via Cube-Space Priors: The Voyage Home (to STRIPS)}, author = { Masataro Asai and Christian Muise}, url = {https://arxiv.org/abs/2004.12850}, year = {2020}, date = {2020-01-01}, journal = {arXiv preprint arXiv:2004.12850}, abstract = {We achieved a new milestone in the difficult task of enabling agents to learn about their environment autonomously. Our neuro-symbolic architecture is trained end-to-end to produce a succinct and effective discrete state transition model from images alone. Our target representation (the Planning Domain Definition Language) is already in a form that off-the-shelf solvers can consume, and opens the door to the rich array of modern heuristic search capabilities. We demonstrate how the sophisticated innate prior we place on the learning process significantly reduces the complexity of the learned representation, and reveals a connection to the graph-theoretic notion of "cube-like graphs", thus opening the door to a deeper understanding of the ideal properties for learned symbolic representations. We show that the powerful domain-independent heuristics allow our system to solve visual 15-Puzzle instances which are beyond the reach of blind search, without resorting to the Reinforcement Learning approach that requires a huge amount of training on the domain-dependent reward information. }, keywords = {Artificial Intelligence, Planning}, pubstate = {published}, tppubtype = {article} } We achieved a new milestone in the difficult task of enabling agents to learn about their environment autonomously. Our neuro-symbolic architecture is trained end-to-end to produce a succinct and effective discrete state transition model from images alone. Our target representation (the Planning Domain Definition Language) is already in a form that off-the-shelf solvers can consume, and opens the door to the rich array of modern heuristic search capabilities. We demonstrate how the sophisticated innate prior we place on the learning process significantly reduces the complexity of the learned representation, and reveals a connection to the graph-theoretic notion of "cube-like graphs", thus opening the door to a deeper understanding of the ideal properties for learned symbolic representations. We show that the powerful domain-independent heuristics allow our system to solve visual 15-Puzzle instances which are beyond the reach of blind search, without resorting to the Reinforcement Learning approach that requires a huge amount of training on the domain-dependent reward information. |
Wollenstein-Betech, Salomón; Muise, Christian; Cassandras, Christos G; Paschalidis, Ioannis Ch; Khazaeni, Yasaman Explainability of Intelligent Transportation Systems using Knowledge Compilation: a Traffic Light Controller Case Journal Article arXiv preprint arXiv:2007.04916, 2020. Abstract | Links | BibTeX | Tags: Artificial Intelligence, Civil Infrastructure, Planning, Transportation @article{wollenstein2020explainability, title = {Explainability of Intelligent Transportation Systems using Knowledge Compilation: a Traffic Light Controller Case}, author = { Salomón Wollenstein-Betech and Christian Muise and Christos G Cassandras and Ioannis Ch Paschalidis and Yasaman Khazaeni}, url = {https://arxiv.org/abs/2007.04916}, year = {2020}, date = {2020-01-01}, journal = {arXiv preprint arXiv:2007.04916}, abstract = {sage of automated controllers which make decisions on an environment are widespread and are often based on black-box models. We use Knowledge Compilation theory to bring explainability to the controller's decision given the state of the system. For this, we use simulated historical state-action data as input and build a compact and structured representation which relates states with actions. We implement this method in a Traffic Light Control scenario where the controller selects the light cycle by observing the presence (or absence) of vehicles in different regions of the incoming roads. }, keywords = {Artificial Intelligence, Civil Infrastructure, Planning, Transportation}, pubstate = {published}, tppubtype = {article} } sage of automated controllers which make decisions on an environment are widespread and are often based on black-box models. We use Knowledge Compilation theory to bring explainability to the controller's decision given the state of the system. For this, we use simulated historical state-action data as input and build a compact and structured representation which relates states with actions. We implement this method in a Traffic Light Control scenario where the controller selects the light cycle by observing the presence (or absence) of vehicles in different regions of the incoming roads. |
Joshi, Keyur D; Chauhan, Vedang; Surgenor, Brian A flexible machine vision system for small part inspection based on a hybrid SVM/ANN approach Journal Article Journal of Intelligent Manufacturing, 31 (1), pp. 103–125, 2020. Abstract | Links | BibTeX | Tags: Computer Vision @article{joshi2020flexible, title = {A flexible machine vision system for small part inspection based on a hybrid SVM/ANN approach}, author = { Keyur D Joshi and Vedang Chauhan and Brian Surgenor}, doi = {10.1007/s10845-018-1438-3}, year = {2020}, date = {2020-01-01}, journal = {Journal of Intelligent Manufacturing}, volume = {31}, number = {1}, pages = {103--125}, publisher = {Springer}, abstract = {Machine vision inspection systems are often used for part classification applications to confirm that correct parts are available in manufacturing or assembly operations. Support vector machines (SVMs) and artificial neural networks (ANNs) are popular choices for classifiers. These supervised classifiers perform well when developed for specific applications and trained with known class images. Their drawback is that they cannot be easily applied to different applications without extensive retuning. Moreover, for the same application, they do not perform well if there are unknown class images. This paper proposes a novel solution to the above limitations of SVMs and ANNs, with the development of a hybrid approach that combines supervised and semi-supervised layers. To illustrate its performance, the system is applied to three different small part identification and sorting applications: (1) solid plastic gears, (2) clear plastic wire connectors and (3) metallic Indian coins. The ability of the system to work with different applications with minimal tuning and user inputs illustrates its flexibility. The robustness of the system is demonstrated by its ability to reject unknown class images. Four hybrid classification methods were developed and tested: (1) SSVM–USVM, (2) USVM–SSVM, (3) USVM–SANN and (4) SANN–USVM. It was found that SANN–USVM gave the best results with an accuracy of over 95% for all three applications. A software package known as FlexMVS for flexible machine vision system was written to illustrate the hybrid approach that enabled easy execution of the image conditioning, feature extraction and classification steps. The image library and database used in this study is available at http://my.me.queensu.ca/People/Surgenor/Laboratory/Database.html.}, keywords = {Computer Vision}, pubstate = {published}, tppubtype = {article} } Machine vision inspection systems are often used for part classification applications to confirm that correct parts are available in manufacturing or assembly operations. Support vector machines (SVMs) and artificial neural networks (ANNs) are popular choices for classifiers. These supervised classifiers perform well when developed for specific applications and trained with known class images. Their drawback is that they cannot be easily applied to different applications without extensive retuning. Moreover, for the same application, they do not perform well if there are unknown class images. This paper proposes a novel solution to the above limitations of SVMs and ANNs, with the development of a hybrid approach that combines supervised and semi-supervised layers. To illustrate its performance, the system is applied to three different small part identification and sorting applications: (1) solid plastic gears, (2) clear plastic wire connectors and (3) metallic Indian coins. The ability of the system to work with different applications with minimal tuning and user inputs illustrates its flexibility. The robustness of the system is demonstrated by its ability to reject unknown class images. Four hybrid classification methods were developed and tested: (1) SSVM–USVM, (2) USVM–SSVM, (3) USVM–SANN and (4) SANN–USVM. It was found that SANN–USVM gave the best results with an accuracy of over 95% for all three applications. A software package known as FlexMVS for flexible machine vision system was written to illustrate the hybrid approach that enabled easy execution of the image conditioning, feature extraction and classification steps. The image library and database used in this study is available at http://my.me.queensu.ca/People/Surgenor/Laboratory/Database.html. |
Jardine, Peter T; Givigi, Sidney N; Yousefi, Shahram Leveraging Data Engineering to Improve Unmanned Aerial Vehicle Control Design Journal Article IEEE Systems Journal, 2020, ISSN: 1937-9234 . Abstract | Links | BibTeX | Tags: Controls, Predictive Model, Robotics, Unmanned Aerial Vehicles @article{jardine2020leveraging, title = {Leveraging Data Engineering to Improve Unmanned Aerial Vehicle Control Design}, author = { Peter T Jardine and Sidney N Givigi and Shahram Yousefi}, doi = {10.1109/JSYST.2020.3003203}, issn = {1937-9234 }, year = {2020}, date = {2020-01-01}, journal = {IEEE Systems Journal}, publisher = {IEEE}, abstract = {The potential benefits of big data and machine learning techniques are yet to be fully realized in real-time, safety-critical applications like unmanned aerial vehicle control. This is because of challenges related to interpretation, error susceptibility, and resources requirements. Due to their robustness and reliability, traditional model-based design techniques still dominate this landscape. However, a growing body of research in adaptive control has demonstrated the potential benefits of merging these two distinct design philosophies. This article investigates the benefits of using a combination of machine learning techniques to automatically tune parameters within a strictly defined model predictive control architecture. Fast orthogonal search and finite action-set learning automata are used to tune model coefficients and objective function weights, respectively. The strategy is validated experimentally on an actual Quanser Qball2 quadcopter and through several simulations of a Parrot AR.drone. Results demonstrate that the proposed approach improves performance while reducing design effort.}, keywords = {Controls, Predictive Model, Robotics, Unmanned Aerial Vehicles}, pubstate = {published}, tppubtype = {article} } The potential benefits of big data and machine learning techniques are yet to be fully realized in real-time, safety-critical applications like unmanned aerial vehicle control. This is because of challenges related to interpretation, error susceptibility, and resources requirements. Due to their robustness and reliability, traditional model-based design techniques still dominate this landscape. However, a growing body of research in adaptive control has demonstrated the potential benefits of merging these two distinct design philosophies. This article investigates the benefits of using a combination of machine learning techniques to automatically tune parameters within a strictly defined model predictive control architecture. Fast orthogonal search and finite action-set learning automata are used to tune model coefficients and objective function weights, respectively. The strategy is validated experimentally on an actual Quanser Qball2 quadcopter and through several simulations of a Parrot AR.drone. Results demonstrate that the proposed approach improves performance while reducing design effort. |
Lins, Romulo Gonçalves; Givigi, Sidney N FPGA-Based Design Optimization in Autonomous Robot Systems for Inspection of Civil Infrastructure Journal Article IEEE Systems Journal, 14 (2), pp. 2961–2964, 2020, ISSN: 1937-9234. Abstract | Links | BibTeX | Tags: Autonomous Vehicles, Civil Infrastructure, Robotics @article{lins2020fpga, title = {FPGA-Based Design Optimization in Autonomous Robot Systems for Inspection of Civil Infrastructure}, author = { Romulo Gonçalves Lins and Sidney N Givigi}, doi = {10.1109/JSYST.2019.2960309}, issn = {1937-9234}, year = {2020}, date = {2020-01-01}, journal = {IEEE Systems Journal}, volume = {14}, number = {2}, pages = {2961--2964}, publisher = {IEEE}, abstract = {Recently, Lins et al. have developed an autonomous robot system for inspection of defects in civil infrastructure. The proposed autonomous system is a unique design because it is comprised of commercial off-the-shelf products that integrate into one fully automated system. The design can be easily deployed to different environments, such as nuclear power plants and bridges, while also being less expensive to create and maintain. However, a crucial point to be improved is the system performance, which is related to the low image processing frequency that largely increases the time needed to autonomously inspect a real structure. A new embedded and more efficient crack detection and crack measurement apparatus based on a field-programmable gate array was developed to optimize the on-board image processing. Thereby, the hardware architecture, software design, and database schematic were redesigned in order to update the autonomous robot system for inspection. Finally, some discussion and data are provided to validate the technical feasibility of the proposed optimization.}, keywords = {Autonomous Vehicles, Civil Infrastructure, Robotics}, pubstate = {published}, tppubtype = {article} } Recently, Lins et al. have developed an autonomous robot system for inspection of defects in civil infrastructure. The proposed autonomous system is a unique design because it is comprised of commercial off-the-shelf products that integrate into one fully automated system. The design can be easily deployed to different environments, such as nuclear power plants and bridges, while also being less expensive to create and maintain. However, a crucial point to be improved is the system performance, which is related to the low image processing frequency that largely increases the time needed to autonomously inspect a real structure. A new embedded and more efficient crack detection and crack measurement apparatus based on a field-programmable gate array was developed to optimize the on-board image processing. Thereby, the hardware architecture, software design, and database schematic were redesigned in order to update the autonomous robot system for inspection. Finally, some discussion and data are provided to validate the technical feasibility of the proposed optimization. |
Ragab, Hany; Elhabiby, Mohamed; Givigi, Sidney; Noureldin, Aboelmagd 2020 IEEE/ION Position, Location and Navigation Symposium (PLANS), pp. 960–966, IEEE 2020, ISSN: 2153-3598. Abstract | Links | BibTeX | Tags: Artificial Intelligence, Autonomous Vehicles, Computer Vision, Navigation @inproceedings{ragab2020utilization, title = {The Utilization of DNN-based Semantic Segmentation for Improving Low-Cost Integrated Stereo Visual Odometry in Challenging Urban Environments}, author = { Hany Ragab and Mohamed Elhabiby and Sidney Givigi and Aboelmagd Noureldin}, doi = {10.1109/PLANS46316.2020.9110144}, issn = {2153-3598}, year = {2020}, date = {2020-01-01}, booktitle = {2020 IEEE/ION Position, Location and Navigation Symposium (PLANS)}, pages = {960--966}, organization = {IEEE}, abstract = {Positioning and Navigation (PN) is one of the most important topics in the world of Autonomous Vehicles (AVs). Being equipped with a suite of sensors and high-performance computers, self-driving cars are designed to perceive its surrounding environment prior to planning and control. Among the observations are the semantics of the objects appearing in the scene. While PN is very challenging for extended GNSS outages, vision sensors can also exhibit failures in the case of highly dynamic scenes and lack of texture between consecutive image frames. To overcome this problem, we propose a stereo visual odometry scheme and advocate the use of a pretrained state-of-the-art Semantic Segmentation (SS) Deep Convolutional Neural Networks (CNN) model to forcefully remove features belonging to objects that most likely behave dynamically in the scene prior to egomotion estimation and integration with inertial sensors. When loosely coupled with inertial sensors, the proposed method was able to outperform the integrated algorithm without SS-based outlier rejection during natural GNSS outages.}, keywords = {Artificial Intelligence, Autonomous Vehicles, Computer Vision, Navigation}, pubstate = {published}, tppubtype = {inproceedings} } Positioning and Navigation (PN) is one of the most important topics in the world of Autonomous Vehicles (AVs). Being equipped with a suite of sensors and high-performance computers, self-driving cars are designed to perceive its surrounding environment prior to planning and control. Among the observations are the semantics of the objects appearing in the scene. While PN is very challenging for extended GNSS outages, vision sensors can also exhibit failures in the case of highly dynamic scenes and lack of texture between consecutive image frames. To overcome this problem, we propose a stereo visual odometry scheme and advocate the use of a pretrained state-of-the-art Semantic Segmentation (SS) Deep Convolutional Neural Networks (CNN) model to forcefully remove features belonging to objects that most likely behave dynamically in the scene prior to egomotion estimation and integration with inertial sensors. When loosely coupled with inertial sensors, the proposed method was able to outperform the integrated algorithm without SS-based outlier rejection during natural GNSS outages. |
Wang, Cunxiang; Liang, Shuailong; Jin, Yili; Wang, Yilong; Zhu, Xiaodan; Zhang, Yue SemEval-2020 task 4: Commonsense validation and explanation Journal Article arXiv preprint arXiv:2007.00236, 2020. Abstract | Links | BibTeX | Tags: Artificial Intelligence, Natural Language Processing @article{wang2020semeval, title = {SemEval-2020 task 4: Commonsense validation and explanation}, author = { Cunxiang Wang and Shuailong Liang and Yili Jin and Yilong Wang and Xiaodan Zhu and Yue Zhang}, url = {https://arxiv.org/abs/2007.00236}, year = {2020}, date = {2020-01-01}, journal = {arXiv preprint arXiv:2007.00236}, abstract = {In this paper, we present SemEval-2020 Task 4, Commonsense Validation and Explanation (ComVE), which includes three subtasks, aiming to evaluate whether a system can distinguish a natural language statement that makes sense to human from one that does not, and provide the reasons. Specifically, in our first subtask, the participating systems are required to choose from two natural language statements of similar wording the one that makes sense and the one does not. The second subtask additionally asks a system to select the key reason from three options why a given statement does not make sense. In the third subtask, a participating system needs to generate the reason automatically. }, keywords = {Artificial Intelligence, Natural Language Processing}, pubstate = {published}, tppubtype = {article} } In this paper, we present SemEval-2020 Task 4, Commonsense Validation and Explanation (ComVE), which includes three subtasks, aiming to evaluate whether a system can distinguish a natural language statement that makes sense to human from one that does not, and provide the reasons. Specifically, in our first subtask, the participating systems are required to choose from two natural language statements of similar wording the one that makes sense and the one does not. The second subtask additionally asks a system to select the key reason from three options why a given statement does not make sense. In the third subtask, a participating system needs to generate the reason automatically. |
Sarkar, Pritam; Etemad, Ali Self-supervised learning for ecg-based emotion recognition Inproceedings ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 3217–3221, IEEE 2020, ISBN: 2379-190X . Abstract | Links | BibTeX | Tags: Affective Computing, Artificial Intelligence, Human Machine Interaction, Neural Networks @inproceedings{sarkar2020self, title = {Self-supervised learning for ecg-based emotion recognition}, author = { Pritam Sarkar and Ali Etemad}, doi = {10.1109/ICASSP40776.2020.9053985}, isbn = {2379-190X }, year = {2020}, date = {2020-01-01}, booktitle = {ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, pages = {3217--3221}, organization = {IEEE}, abstract = {We present an electrocardiogram (ECG) -based emotion recognition system using self-supervised learning. Our proposed architecture consists of two main networks, a signal transformation recognition network and an emotion recognition network. First, unlabelled data are used to successfully train the former network to detect specific pre-determined signal transformations in the self-supervised learning step. Next, the weights of the convolutional layers of this network are transferred to the emotion recognition network, and two dense layers are trained in order to classify arousal and valence scores. We show that our self-supervised approach helps the model learn the ECG feature manifold required for emotion recognition, performing equal or better than the fully-supervised version of the model. Our proposed method outperforms the state-of-the-art in ECG-based emotion recognition with two publicly available datasets, SWELL and AMIGOS. Further analysis highlights the advantage of our self-supervised approach in requiring significantly less data to achieve acceptable results.}, keywords = {Affective Computing, Artificial Intelligence, Human Machine Interaction, Neural Networks}, pubstate = {published}, tppubtype = {inproceedings} } We present an electrocardiogram (ECG) -based emotion recognition system using self-supervised learning. Our proposed architecture consists of two main networks, a signal transformation recognition network and an emotion recognition network. First, unlabelled data are used to successfully train the former network to detect specific pre-determined signal transformations in the self-supervised learning step. Next, the weights of the convolutional layers of this network are transferred to the emotion recognition network, and two dense layers are trained in order to classify arousal and valence scores. We show that our self-supervised approach helps the model learn the ECG feature manifold required for emotion recognition, performing equal or better than the fully-supervised version of the model. Our proposed method outperforms the state-of-the-art in ECG-based emotion recognition with two publicly available datasets, SWELL and AMIGOS. Further analysis highlights the advantage of our self-supervised approach in requiring significantly less data to achieve acceptable results. |
Best, Aaron N; Wu, Amy R Upper body and ankle strategies compensate for reduced lateral stability at very slow walking speeds Journal Article bioRxiv, 2020. Abstract | Links | BibTeX | Tags: Biomechanics @article{best2020upper, title = {Upper body and ankle strategies compensate for reduced lateral stability at very slow walking speeds}, author = { Aaron N Best and Amy R Wu}, doi = {10.1101/2020.07.16.207092 }, year = {2020}, date = {2020-01-01}, journal = {bioRxiv}, publisher = {Cold Spring Harbor Laboratory}, abstract = {At the typical walking speeds of healthy humans, step placement seems to be the primary strategy to maintain gait stability, with ankle torques and upper body momentum providing additional compensation. The average walking speeds of populations with an increased risk of falling, however, are much slower and may require differing control strategies. The purpose of this study was to analyze mediolateral gait stability and the contributions of the different control strategies at very slow walking speeds. We analyzed an open dataset including kinematics and kinetics from eight healthy subjects walking at speeds from 0.1 to 0.6 m/s as well as a self-selected speed. As gait speed slowed, we found that the margin of stability decreased linearly. Increased lateral excursions of the extrapolated centre of mass, caused by increased lateral excursions of the trunk, were not compensated for by an equivalent increase in the lateral centre of pressure, leading to decreased margin of stability. Additionally, both the ankle eversion torque and hip abduction torque at the minimum margin of stability event increased at the same rate as gait speed slowed. These results suggest that the contributions of both the ankle and the upper body to stability are more crucial than stepping at slow speeds, which have important implications for populations with slow gait and limited motor function.}, keywords = {Biomechanics}, pubstate = {published}, tppubtype = {article} } At the typical walking speeds of healthy humans, step placement seems to be the primary strategy to maintain gait stability, with ankle torques and upper body momentum providing additional compensation. The average walking speeds of populations with an increased risk of falling, however, are much slower and may require differing control strategies. The purpose of this study was to analyze mediolateral gait stability and the contributions of the different control strategies at very slow walking speeds. We analyzed an open dataset including kinematics and kinetics from eight healthy subjects walking at speeds from 0.1 to 0.6 m/s as well as a self-selected speed. As gait speed slowed, we found that the margin of stability decreased linearly. Increased lateral excursions of the extrapolated centre of mass, caused by increased lateral excursions of the trunk, were not compensated for by an equivalent increase in the lateral centre of pressure, leading to decreased margin of stability. Additionally, both the ankle eversion torque and hip abduction torque at the minimum margin of stability event increased at the same rate as gait speed slowed. These results suggest that the contributions of both the ankle and the upper body to stability are more crucial than stepping at slow speeds, which have important implications for populations with slow gait and limited motor function. |
Dai, Yinpei; Li, Hangyu; Tang, Chengguang; Li, Yongbin; Sun, Jian; Zhu, Xiaodan Learning Low-Resource End-To-End Goal-Oriented Dialog for Fast and Reliable System Deployment Inproceedings Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pp. 609–618, 2020. Abstract | Links | BibTeX | Tags: Natural Language Processing @inproceedings{dai2020learning, title = {Learning Low-Resource End-To-End Goal-Oriented Dialog for Fast and Reliable System Deployment}, author = { Yinpei Dai and Hangyu Li and Chengguang Tang and Yongbin Li and Jian Sun and Xiaodan Zhu}, doi = {10.18653/v1/2020.acl-main.57}, year = {2020}, date = {2020-01-01}, booktitle = {Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics}, pages = {609--618}, abstract = {Existing end-to-end dialog systems perform less effectively when data is scarce. To obtain an acceptable success in real-life online services with only a handful of training examples, both fast adaptability and reliable performance are highly desirable for dialog systems. In this paper, we propose the Meta-Dialog System (MDS), which combines the advantages of both meta-learning approaches and human-machine collaboration. We evaluate our methods on a new extended-bAbI dataset and a transformed MultiWOZ dataset for low-resource goal-oriented dialog learning. Experimental results show that MDS significantly outperforms non-meta-learning baselines and can achieve more than 90% per-turn accuracies with only 10 dialogs on the extended-bAbI dataset.}, keywords = {Natural Language Processing}, pubstate = {published}, tppubtype = {inproceedings} } Existing end-to-end dialog systems perform less effectively when data is scarce. To obtain an acceptable success in real-life online services with only a handful of training examples, both fast adaptability and reliable performance are highly desirable for dialog systems. In this paper, we propose the Meta-Dialog System (MDS), which combines the advantages of both meta-learning approaches and human-machine collaboration. We evaluate our methods on a new extended-bAbI dataset and a transformed MultiWOZ dataset for low-resource goal-oriented dialog learning. Experimental results show that MDS significantly outperforms non-meta-learning baselines and can achieve more than 90% per-turn accuracies with only 10 dialogs on the extended-bAbI dataset. |
Ruan, Yu-Ping; Ling, Zhen-Hua; Zhu, Xiaodan; Liu, Quan; Gu, Jia-Chen Generating diverse conversation responses by creating and ranking multiple candidates Journal Article Computer Speech & Language, 62 , pp. 101071, 2020. Abstract | Links | BibTeX | Tags: Natural Language Processing @article{ruan2020generating, title = {Generating diverse conversation responses by creating and ranking multiple candidates}, author = { Yu-Ping Ruan and Zhen-Hua Ling and Xiaodan Zhu and Quan Liu and Jia-Chen Gu}, doi = {10.1016/j.csl.2020.101071}, year = {2020}, date = {2020-01-01}, journal = {Computer Speech & Language}, volume = {62}, pages = {101071}, publisher = {Elsevier}, abstract = {This paper introduces our systems built for Track 2 of Dialog System Technology Challenge 7 (DSTC7). This challenge track aimed to evaluate the response generation methods using fully data-driven conversation models in a knowledge-grounded setting, where textual facts were provided as the knowledge for each context-response pair. The sequence-to-sequence models have achieved impressive results in machine translation and have also been widely used for end-to-end generative conversation modelling. However, they tended to output dull and repeated responses in previous studies. Our work aims to promote the diversity of end-to-end conversation response generation by adopting a two-stage pipeline. 1) Create multiple responses for an input context together with its textual facts. At this stage, two different models are designed, i.e., a variational generative (VariGen) model and a retrieval-based (Retrieval) model. 2) Rank and return the most relevant response by training a topic coherence discrimination (TCD) model for calculating ranking scores. In our experiments, we demonstrated the effectiveness of the response ranking strategy and the external textual knowledge for generating better responses. According to the official evaluation results, our Retrieval and VariGen systems ranked first and second respectively among all participant systems on Entropy metrics which measured the objective diversity of generated responses. Besides, the VariGen system ranked second on NIST and METEOR metrics which measured the objective quality of generated responses.}, keywords = {Natural Language Processing}, pubstate = {published}, tppubtype = {article} } This paper introduces our systems built for Track 2 of Dialog System Technology Challenge 7 (DSTC7). This challenge track aimed to evaluate the response generation methods using fully data-driven conversation models in a knowledge-grounded setting, where textual facts were provided as the knowledge for each context-response pair. The sequence-to-sequence models have achieved impressive results in machine translation and have also been widely used for end-to-end generative conversation modelling. However, they tended to output dull and repeated responses in previous studies. Our work aims to promote the diversity of end-to-end conversation response generation by adopting a two-stage pipeline. 1) Create multiple responses for an input context together with its textual facts. At this stage, two different models are designed, i.e., a variational generative (VariGen) model and a retrieval-based (Retrieval) model. 2) Rank and return the most relevant response by training a topic coherence discrimination (TCD) model for calculating ranking scores. In our experiments, we demonstrated the effectiveness of the response ranking strategy and the external textual knowledge for generating better responses. According to the official evaluation results, our Retrieval and VariGen systems ranked first and second respectively among all participant systems on Entropy metrics which measured the objective diversity of generated responses. Besides, the VariGen system ranked second on NIST and METEOR metrics which measured the objective quality of generated responses. |
Li, Xiaoyan; Kiringa, Iluju; Yeap, Tet; Zhu, Xiaodan; Li, Yifeng Exploring deep anomaly detection methods based on capsule net Inproceedings Canadian Conference on Artificial Intelligence, pp. 375–387, Springer 2020, ISBN: 978-3-030-47358-7. Abstract | Links | BibTeX | Tags: Artificial Intelligence, Natural Language Processing @inproceedings{li2020exploring, title = {Exploring deep anomaly detection methods based on capsule net}, author = { Xiaoyan Li and Iluju Kiringa and Tet Yeap and Xiaodan Zhu and Yifeng Li}, doi = {10.1007/978-3-030-47358-7_39}, isbn = {978-3-030-47358-7}, year = {2020}, date = {2020-01-01}, booktitle = {Canadian Conference on Artificial Intelligence}, pages = {375--387}, organization = {Springer}, abstract = {In this paper, we develop and explore deep anomaly detection techniques based on capsule network (named AnoCapsNet) for image data. Being able to encode intrinsic spatial relationship between parts and a whole, CapsNet has been applied as both a classifier and deep autoencoder. This inspires us to design three normality score functions: prediction-probability-based (PP-based), reconstruction-error-based (RE-based), and combination of both (PP+RE-based) for evaluating the “outlierness” of unseen images. Our results on four datasets demonstrate that PP-based and RE-based methods outperform the principled benchmark methods in many cases and the pp-based method performs consistently well, while the RE-based approach is relatively sensitive to the similarity between labeled and unlabeled images. The PP+RE-based approach effectively takes advantages of both methods and achieves state-of-the-art results. }, keywords = {Artificial Intelligence, Natural Language Processing}, pubstate = {published}, tppubtype = {inproceedings} } In this paper, we develop and explore deep anomaly detection techniques based on capsule network (named AnoCapsNet) for image data. Being able to encode intrinsic spatial relationship between parts and a whole, CapsNet has been applied as both a classifier and deep autoencoder. This inspires us to design three normality score functions: prediction-probability-based (PP-based), reconstruction-error-based (RE-based), and combination of both (PP+RE-based) for evaluating the “outlierness” of unseen images. Our results on four datasets demonstrate that PP-based and RE-based methods outperform the principled benchmark methods in many cases and the pp-based method performs consistently well, while the RE-based approach is relatively sensitive to the similarity between labeled and unlabeled images. The PP+RE-based approach effectively takes advantages of both methods and achieves state-of-the-art results. |
Gu, Jia-Chen; Li, Tianda; Liu, Quan; Zhu, Xiaodan; Ling, Zhen-Hua; Ruan, Yu-Ping Pre-Trained and Attention-Based Neural Networks for Building Noetic Task-Oriented Dialogue Systems Journal Article arXiv preprint arXiv:2004.01940, 2020. Abstract | Links | BibTeX | Tags: Artificial Intelligence, Natural Language Processing @article{gu2020pre, title = {Pre-Trained and Attention-Based Neural Networks for Building Noetic Task-Oriented Dialogue Systems}, author = { Jia-Chen Gu and Tianda Li and Quan Liu and Xiaodan Zhu and Zhen-Hua Ling and Yu-Ping Ruan}, url = {https://arxiv.org/abs/2004.01940}, year = {2020}, date = {2020-01-01}, journal = {arXiv preprint arXiv:2004.01940}, abstract = {The NOESIS II challenge, as the Track 2 of the 8th Dialogue System Technology Challenges (DSTC 8), is the extension of DSTC 7. This track incorporates new elements that are vital for the creation of a deployed task-oriented dialogue system. This paper describes our systems that are evaluated on all subtasks under this challenge. We study the problem of employing pre-trained attention-based network for multi-turn dialogue systems. Meanwhile, several adaptation methods are proposed to adapt the pre-trained language models for multi-turn dialogue systems, in order to keep the intrinsic property of dialogue systems. In the released evaluation results of Track 2 of DSTC 8, our proposed models ranked fourth in subtask 1, third in subtask 2, and first in subtask 3 and subtask 4 respectively. }, keywords = {Artificial Intelligence, Natural Language Processing}, pubstate = {published}, tppubtype = {article} } The NOESIS II challenge, as the Track 2 of the 8th Dialogue System Technology Challenges (DSTC 8), is the extension of DSTC 7. This track incorporates new elements that are vital for the creation of a deployed task-oriented dialogue system. This paper describes our systems that are evaluated on all subtasks under this challenge. We study the problem of employing pre-trained attention-based network for multi-turn dialogue systems. Meanwhile, several adaptation methods are proposed to adapt the pre-trained language models for multi-turn dialogue systems, in order to keep the intrinsic property of dialogue systems. In the released evaluation results of Track 2 of DSTC 8, our proposed models ranked fourth in subtask 1, third in subtask 2, and first in subtask 3 and subtask 4 respectively. |
Hedayati, Mohammadali; Kim, Il-Min CoMP-NOMA in the SWIPT Networks Journal Article IEEE Transactions on Wireless Communications, 2020, ISBN: 1558-2248. Abstract | Links | BibTeX | Tags: Telecommunication @article{hedayati2020comp, title = {CoMP-NOMA in the SWIPT Networks}, author = { Mohammadali Hedayati and Il-Min Kim}, doi = {10.1109/TWC.2020.2985038}, isbn = {1558-2248}, year = {2020}, date = {2020-01-01}, journal = {IEEE Transactions on Wireless Communications}, publisher = {IEEE}, abstract = {This paper studies coordinated multipoint (CoMP) transmission with non-orthogonal multiple access (NOMA) for a simultaneous wireless information and power transfer (SWIPT) network. We investigate two different CoMP-NOMA schemes, namely joint transmission-NOMA (JT-NOMA) and Alamouti NOMA (A-NOMA). Also, as the benchmark, joint transmission orthogonal multiple access (JT-OMA) is considered. Expressions are derived for the outage probabilities achieved by JT-NOMA, A-NOMA, and JT-OMA for cell edge users with quality of service (QoS) requirements. To compare these schemes in terms of spectral efficiency, we use the ϵ -outage rate region that is defined to be the set of all achievable rate pairs at which the outage probabilities are equal to ϵ . It is shown that A-NOMA is superior to both JT-NOMA and JT-OMA in high SNR. Also, it is analytically proved that JT-NOMA outperforms JT-OMA, when the circuit energy consumption is negligible. A particularly interesting observation is that when the circuit energy consumption is non-negligible, JT-NOMA is not always better than JT-OMA. It is shown that when the difference between the expectations of effective channel power gains is not sufficiently large, JT-OMA is superior to JT-NOMA for the same circuit energy consumption.}, keywords = {Telecommunication}, pubstate = {published}, tppubtype = {article} } This paper studies coordinated multipoint (CoMP) transmission with non-orthogonal multiple access (NOMA) for a simultaneous wireless information and power transfer (SWIPT) network. We investigate two different CoMP-NOMA schemes, namely joint transmission-NOMA (JT-NOMA) and Alamouti NOMA (A-NOMA). Also, as the benchmark, joint transmission orthogonal multiple access (JT-OMA) is considered. Expressions are derived for the outage probabilities achieved by JT-NOMA, A-NOMA, and JT-OMA for cell edge users with quality of service (QoS) requirements. To compare these schemes in terms of spectral efficiency, we use the ϵ -outage rate region that is defined to be the set of all achievable rate pairs at which the outage probabilities are equal to ϵ . It is shown that A-NOMA is superior to both JT-NOMA and JT-OMA in high SNR. Also, it is analytically proved that JT-NOMA outperforms JT-OMA, when the circuit energy consumption is negligible. A particularly interesting observation is that when the circuit energy consumption is non-negligible, JT-NOMA is not always better than JT-OMA. It is shown that when the difference between the expectations of effective channel power gains is not sufficiently large, JT-OMA is superior to JT-NOMA for the same circuit energy consumption. |
Kang, Jae-Mo; Chun, Chang-Jae; Kim, Il-Min; Kim, Dong In Dynamic Power Splitting for SWIPT With Nonlinear Energy Harvesting in Ergodic Fading Channel Journal Article IEEE Internet of Things Journal, 2020, ISSN: 2327-4662. Abstract | Links | BibTeX | Tags: Energy Harvesting, Telecommunication @article{kang2020dynamic, title = {Dynamic Power Splitting for SWIPT With Nonlinear Energy Harvesting in Ergodic Fading Channel}, author = { Jae-Mo Kang and Chang-Jae Chun and Il-Min Kim and Dong In Kim}, doi = { 10.1109/JIOT.2020.2980328}, issn = {2327-4662}, year = {2020}, date = {2020-01-01}, journal = {IEEE Internet of Things Journal}, publisher = {IEEE}, abstract = {Simultaneous wireless information and power transfer (SWIPT) is very promising for various applications with the Internet of Things (IoT). In this article, we study dynamic power splitting for the SWIPT in an ergodic fading channel. Considering nonlinearity of practical energy harvesting (EH) circuits, we adopt the realistic nonlinear EH model rather than the idealistic linear EH model. To characterize the optimal rate-energy (R-E) tradeoff, we consider the problem of maximizing the R-E region, which is nonconvex. We solve this challenging problem for two different cases of the channel state information (CSI): 1) when the CSI is known only at the receiver (the CSIR case) and 2) when the CSI is known at both the transmitter and the receiver (the CSI case). For these two cases, we develop the corresponding optimal dynamic power-splitting schemes. To address the complexity issue, we also propose the suboptimal schemes with low complexities. Comparing the proposed schemes to the existing schemes, we provide various useful insights into the dynamic power splitting with nonlinear EH. Furthermore, we extend the analysis to the scenarios of the partial CSI at the transmitter and the harvested energy maximization. The numerical results demonstrate that the proposed schemes significantly outperform the existing schemes and the proposed suboptimal scheme works very close to the optimal scheme at a much lower complexity.}, keywords = {Energy Harvesting, Telecommunication}, pubstate = {published}, tppubtype = {article} } Simultaneous wireless information and power transfer (SWIPT) is very promising for various applications with the Internet of Things (IoT). In this article, we study dynamic power splitting for the SWIPT in an ergodic fading channel. Considering nonlinearity of practical energy harvesting (EH) circuits, we adopt the realistic nonlinear EH model rather than the idealistic linear EH model. To characterize the optimal rate-energy (R-E) tradeoff, we consider the problem of maximizing the R-E region, which is nonconvex. We solve this challenging problem for two different cases of the channel state information (CSI): 1) when the CSI is known only at the receiver (the CSIR case) and 2) when the CSI is known at both the transmitter and the receiver (the CSI case). For these two cases, we develop the corresponding optimal dynamic power-splitting schemes. To address the complexity issue, we also propose the suboptimal schemes with low complexities. Comparing the proposed schemes to the existing schemes, we provide various useful insights into the dynamic power splitting with nonlinear EH. Furthermore, we extend the analysis to the scenarios of the partial CSI at the transmitter and the harvested energy maximization. The numerical results demonstrate that the proposed schemes significantly outperform the existing schemes and the proposed suboptimal scheme works very close to the optimal scheme at a much lower complexity. |
Kang, Jae-Mo; Chun, Chang-Jae; Kim, Il-Min; Kim, Dong In Deep RNN-Based Channel Tracking for Wireless Energy Transfer System Journal Article IEEE Systems Journal, 2020, ISSN: 1937-9234. Abstract | Links | BibTeX | Tags: Neural Networks, Telecommunication @article{kang2020deep, title = {Deep RNN-Based Channel Tracking for Wireless Energy Transfer System}, author = { Jae-Mo Kang and Chang-Jae Chun and Il-Min Kim and Dong In Kim}, doi = {10.1109/JSYST.2020.2975188}, issn = {1937-9234}, year = {2020}, date = {2020-01-01}, journal = {IEEE Systems Journal}, publisher = {IEEE}, abstract = {In this article, we study channel tracking for a wireless energy transfer (WET) system. This problem is practically very important, but challenging. Regarding time-varying channels as a sequence to be predicted, we exploit the deep learning technique for channel tracking. Particularly, by constructing a recurrent neural network (RNN) architecture based on long short-term memory (LSTM) and feedforward neural network (FNN), we develop a novel channel tracking scheme for the WET system. This scheme sequentially estimates the channel state information (CSI) at the ET based on the previous CSI estimates and the harvested energy feedback information from the ER. Numerical results demonstrate the superior performance and effectiveness of the proposed scheme.}, keywords = {Neural Networks, Telecommunication}, pubstate = {published}, tppubtype = {article} } In this article, we study channel tracking for a wireless energy transfer (WET) system. This problem is practically very important, but challenging. Regarding time-varying channels as a sequence to be predicted, we exploit the deep learning technique for channel tracking. Particularly, by constructing a recurrent neural network (RNN) architecture based on long short-term memory (LSTM) and feedforward neural network (FNN), we develop a novel channel tracking scheme for the WET system. This scheme sequentially estimates the channel state information (CSI) at the ET based on the previous CSI estimates and the harvested energy feedback information from the ER. Numerical results demonstrate the superior performance and effectiveness of the proposed scheme. |
Yun, Sangseok; Kang, Jae-Mo; Kim, Il-Min; Ha, Jeongseok Deep Artificial Noise: Deep Learning-Based Precoding Optimization for Artificial Noise Scheme Journal Article IEEE Transactions on Vehicular Technology, 69 (3), pp. 3465–3469, 2020, ISSN: 1939-9359. Abstract | Links | BibTeX | Tags: Deep Learning, Signal Processing, Telecommunication @article{yun2020deep, title = {Deep Artificial Noise: Deep Learning-Based Precoding Optimization for Artificial Noise Scheme}, author = { Sangseok Yun and Jae-Mo Kang and Il-Min Kim and Jeongseok Ha}, doi = {10.1109/TVT.2020.2965959}, issn = {1939-9359}, year = {2020}, date = {2020-01-01}, journal = {IEEE Transactions on Vehicular Technology}, volume = {69}, number = {3}, pages = {3465--3469}, publisher = {IEEE}, abstract = {In this work, we consider a secure precoding optimization problem for the artificial noise (AN) scheme in multiple-input single-output (MISO) wiretap channels. In previous researches (Lin et al., 2013), it was proved that the generalized AN scheme which allows some portion of AN signal to be injected to the legitimate receiver's channel is the optimal precoding scheme for MISO wiretap channels. However, the optimality is valid only under some ideal assumptions such as perfect channel estimation and spatially uncorrelated channels. To break through this limitation, in this paper, we propose a novel deep neural network (DNN)-based secure precoding scheme, called the deep AN scheme. To the best of the authors' knowledge, the deep AN scheme is the first secure precoding scheme which exploits a DNN to jointly design and optimize the precoders for the information signal and the AN signal. From the numerical experiments, it is demonstrated that the proposed deep AN scheme outperforms the generalized AN scheme under various practical wireless environments.}, keywords = {Deep Learning, Signal Processing, Telecommunication}, pubstate = {published}, tppubtype = {article} } In this work, we consider a secure precoding optimization problem for the artificial noise (AN) scheme in multiple-input single-output (MISO) wiretap channels. In previous researches (Lin et al., 2013), it was proved that the generalized AN scheme which allows some portion of AN signal to be injected to the legitimate receiver's channel is the optimal precoding scheme for MISO wiretap channels. However, the optimality is valid only under some ideal assumptions such as perfect channel estimation and spatially uncorrelated channels. To break through this limitation, in this paper, we propose a novel deep neural network (DNN)-based secure precoding scheme, called the deep AN scheme. To the best of the authors' knowledge, the deep AN scheme is the first secure precoding scheme which exploits a DNN to jointly design and optimize the precoders for the information signal and the AN signal. From the numerical experiments, it is demonstrated that the proposed deep AN scheme outperforms the generalized AN scheme under various practical wireless environments. |
Schneider, Adrian Jessup L; Keiver, Kathy; Orr, Alison Pritchard; Reynolds, James N; Golubovich, Neven; Graham, TC Nicholas Toward the Design of Enjoyable Games for Children with Fetal Alcohol Spectrum Disorder Inproceedings Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, pp. 1–13, 2020. Abstract | Links | BibTeX | Tags: Game Design, Healthcare @inproceedings{schneider2020toward, title = {Toward the Design of Enjoyable Games for Children with Fetal Alcohol Spectrum Disorder}, author = { Adrian L Jessup Schneider and Kathy Keiver and Alison Pritchard Orr and James N Reynolds and Neven Golubovich and TC Nicholas Graham}, doi = {10.1145/3313831.3376480}, year = {2020}, date = {2020-01-01}, booktitle = {Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems}, pages = {1--13}, abstract = {Fetal Alcohol Spectrum Disorder (FASD) is a heterogeneous and complex set of disorders caused by prenatal alcohol exposure, estimated to affect 2-5% of the North American population. Deficits associated with FASD affect social skill development and executive function, including emotional regulation and impulse control. These deficits can increase the difficulty of playing digital games. While considerable research has been performed in understanding how to design games for people with neurodevelopmental disorders in general, there is little data on how to design engaging games for children with FASD. We conducted a ten-week in-school gaming trial with eleven elementary-aged children with diagnosed or suspected FASD. Participants enjoyed playing together and responded well to the in-game reward system, while some game elements caused unexpected frustration. Based on our observations, we advise that games for FASD be designed to have low cost of failure, avoid retracting options, account for taking breaks when needed, show progression in rewards, and enable cooperative play.}, keywords = {Game Design, Healthcare}, pubstate = {published}, tppubtype = {inproceedings} } Fetal Alcohol Spectrum Disorder (FASD) is a heterogeneous and complex set of disorders caused by prenatal alcohol exposure, estimated to affect 2-5% of the North American population. Deficits associated with FASD affect social skill development and executive function, including emotional regulation and impulse control. These deficits can increase the difficulty of playing digital games. While considerable research has been performed in understanding how to design games for people with neurodevelopmental disorders in general, there is little data on how to design engaging games for children with FASD. We conducted a ten-week in-school gaming trial with eleven elementary-aged children with diagnosed or suspected FASD. Participants enjoyed playing together and responded well to the in-game reward system, while some game elements caused unexpected frustration. Based on our observations, we advise that games for FASD be designed to have low cost of failure, avoid retracting options, account for taking breaks when needed, show progression in rewards, and enable cooperative play. |
Uppal, Hardik; Sepas-Moghaddam, Alireza; Greenspan, Michael; Etemad, Ali Attention-aware fusion RGB-D face recognition Journal Article arXiv preprint arXiv:2003.00168, 2020. Abstract | Links | BibTeX | Tags: Artificial Intelligence, Computer Vision, Neural Networks @article{uppal2020attention, title = {Attention-aware fusion RGB-D face recognition}, author = { Hardik Uppal and Alireza Sepas-Moghaddam and Michael Greenspan and Ali Etemad}, url = {https://arxiv.org/abs/2003.00168}, year = {2020}, date = {2020-01-01}, journal = {arXiv preprint arXiv:2003.00168}, abstract = {With recent advances in RGB-D sensing technologies as well as improvements in machine learning and fusion techniques, RGB-D facial recognition has become an active area of research. A novel attention aware method is proposed to fuse two image modalities, RGB and depth, for enhanced RGB-D facial recognition. The proposed method first extracts features from both modalities using a convolutional feature extractor. These features are then fused using a two-layer attention mechanism. The first layer focuses on the fused feature maps generated by the feature extractor, exploiting the relationship between feature maps using LSTM recurrent learning. The second layer focuses on the spatial features of those maps using convolution. The training database is preprocessed and augmented through a set of geometric transformations, and the learning process is further aided using transfer learning from a pure 2D RGB image training process. Comparative evaluations demonstrate that the proposed method outperforms other state-of-the-art approaches, including both traditional and deep neural network-based methods, on the challenging CurtinFaces and IIIT-D RGB-D benchmark databases, achieving classification accuracies over 98.2% and 99.3% respectively. The proposed attention mechanism is also compared with other attention mechanisms, demonstrating more accurate results. }, keywords = {Artificial Intelligence, Computer Vision, Neural Networks}, pubstate = {published}, tppubtype = {article} } With recent advances in RGB-D sensing technologies as well as improvements in machine learning and fusion techniques, RGB-D facial recognition has become an active area of research. A novel attention aware method is proposed to fuse two image modalities, RGB and depth, for enhanced RGB-D facial recognition. The proposed method first extracts features from both modalities using a convolutional feature extractor. These features are then fused using a two-layer attention mechanism. The first layer focuses on the fused feature maps generated by the feature extractor, exploiting the relationship between feature maps using LSTM recurrent learning. The second layer focuses on the spatial features of those maps using convolution. The training database is preprocessed and augmented through a set of geometric transformations, and the learning process is further aided using transfer learning from a pure 2D RGB image training process. Comparative evaluations demonstrate that the proposed method outperforms other state-of-the-art approaches, including both traditional and deep neural network-based methods, on the challenging CurtinFaces and IIIT-D RGB-D benchmark databases, achieving classification accuracies over 98.2% and 99.3% respectively. The proposed attention mechanism is also compared with other attention mechanisms, demonstrating more accurate results. |
Sepas-Moghaddam, Alireza; Etemad, Ali; Pereira, Fernando; Correia, Paulo Lobato Facial emotion recognition using light field images with deep attention-based bidirectional LSTM Inproceedings ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 3367–3371, IEEE 2020. Abstract | Links | BibTeX | Tags: Affective Computing, Artificial Intelligence, Computer Vision, Neural Networks @inproceedings{sepas2020facial, title = {Facial emotion recognition using light field images with deep attention-based bidirectional LSTM}, author = { Alireza Sepas-Moghaddam and Ali Etemad and Fernando Pereira and Paulo Lobato Correia}, doi = {10.1109/ICASSP40776.2020.9053919}, year = {2020}, date = {2020-01-01}, booktitle = {ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, pages = {3367--3371}, organization = {IEEE}, abstract = {Light field cameras are able to capture the intensity of light rays coming from multiple directions, thus representing the visual scene from multiple viewpoints. This paper exploits the rich spatio-angular information available in light field images for facial emotion recognition. In this context, a new deep network is proposed that first extracts spatial features using a VGG16 convolutional neural network. Then, a Bidirectional Long Short-Term Memory (Bi-LSTM) recurrent neural network is used to learn spatio-angular features from viewpoint feature sequences, exploring both forward and backward angular relationships. Additionally, an attention mechanism allows our model to selectively focus on the most important spatio-angular features, thus enabling a more effective learning outcome. Finally, a fusion scheme is adopted to obtain the emotion recognition classification results. Comprehensive experiments have been conducted on the IST-EURECOM Light Field Face database using two challenging evaluation protocols, showing the superiority of our method over the state-of-the-art.}, keywords = {Affective Computing, Artificial Intelligence, Computer Vision, Neural Networks}, pubstate = {published}, tppubtype = {inproceedings} } Light field cameras are able to capture the intensity of light rays coming from multiple directions, thus representing the visual scene from multiple viewpoints. This paper exploits the rich spatio-angular information available in light field images for facial emotion recognition. In this context, a new deep network is proposed that first extracts spatial features using a VGG16 convolutional neural network. Then, a Bidirectional Long Short-Term Memory (Bi-LSTM) recurrent neural network is used to learn spatio-angular features from viewpoint feature sequences, exploring both forward and backward angular relationships. Additionally, an attention mechanism allows our model to selectively focus on the most important spatio-angular features, thus enabling a more effective learning outcome. Finally, a fusion scheme is adopted to obtain the emotion recognition classification results. Comprehensive experiments have been conducted on the IST-EURECOM Light Field Face database using two challenging evaluation protocols, showing the superiority of our method over the state-of-the-art. |
Sarkar, Pritam; Etemad, Ali Self-supervised ecg representation learning for emotion recognition Journal Article arXiv preprint arXiv:2002.03898, 2020. Abstract | Links | BibTeX | Tags: Affective Computing, Artificial Intelligence, Neural Networks @article{sarkar2020selfb, title = {Self-supervised ecg representation learning for emotion recognition}, author = { Pritam Sarkar and Ali Etemad}, url = {https://arxiv.org/abs/2002.03898}, year = {2020}, date = {2020-01-01}, journal = {arXiv preprint arXiv:2002.03898}, abstract = {We present a self-supervised deep multi-task learning framework for electrocardiogram (ECG) -based emotion recognition. The proposed framework consists of two stages of learning a) learning ECG representations and b) learning to classify emotions. ECG representations are learned by a signal transformation recognition network. The network learns high-level abstract representations from unlabeled ECG data. Six different signal transformations are applied to the ECG signals, and transformation recognition is performed as pretext tasks. Training the model on pretext tasks helps our network learn spatiotemporal representations that generalize well across different datasets and different emotion categories. We transfer the weights of the self-supervised network to an emotion recognition network, where the convolutional layers are kept frozen and the dense layers are trained with labelled ECG data. We show that our proposed method considerably improves the performance compared to a network trained using fully-supervised learning. New state-of-the-art results are set in classification of arousal, valence, affective states, and stress for the four utilized datasets. Extensive experiments are performed, providing interesting insights into the impact of using a multi-task self-supervised structure instead of a single-task model, as well as the optimum level of difficulty required for the pretext self-supervised tasks. }, keywords = {Affective Computing, Artificial Intelligence, Neural Networks}, pubstate = {published}, tppubtype = {article} } We present a self-supervised deep multi-task learning framework for electrocardiogram (ECG) -based emotion recognition. The proposed framework consists of two stages of learning a) learning ECG representations and b) learning to classify emotions. ECG representations are learned by a signal transformation recognition network. The network learns high-level abstract representations from unlabeled ECG data. Six different signal transformations are applied to the ECG signals, and transformation recognition is performed as pretext tasks. Training the model on pretext tasks helps our network learn spatiotemporal representations that generalize well across different datasets and different emotion categories. We transfer the weights of the self-supervised network to an emotion recognition network, where the convolutional layers are kept frozen and the dense layers are trained with labelled ECG data. We show that our proposed method considerably improves the performance compared to a network trained using fully-supervised learning. New state-of-the-art results are set in classification of arousal, valence, affective states, and stress for the four utilized datasets. Extensive experiments are performed, providing interesting insights into the impact of using a multi-task self-supervised structure instead of a single-task model, as well as the optimum level of difficulty required for the pretext self-supervised tasks. |
Kourkounakis, Tedd; Hajavi, Amirhossein; Etemad, Ali Detecting Multiple Speech Disfluencies Using a Deep Residual Network with Bidirectional Long Short-Term Memory Inproceedings ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 6089–6093, IEEE 2020, ISSN: 2379-190X. Abstract | Links | BibTeX | Tags: Artificial Intelligence, Healthcare, Neural Networks, Signal Processing @inproceedings{kourkounakis2020detecting, title = {Detecting Multiple Speech Disfluencies Using a Deep Residual Network with Bidirectional Long Short-Term Memory}, author = { Tedd Kourkounakis and Amirhossein Hajavi and Ali Etemad}, doi = {10.1109/ICASSP40776.2020.9053893}, issn = {2379-190X}, year = {2020}, date = {2020-01-01}, booktitle = {ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, pages = {6089--6093}, organization = {IEEE}, abstract = {Stuttering is a speech impediment affecting tens of millions of people on an everyday basis. Even with its commonality, there is minimal data and research on the identification and classification of stuttered speech. This paper tackles the problem of detection and classification of different forms of stutters. As opposed to most existing works that identify stutters with language models, our work proposes a model that relies solely on acoustic features, allowing for identification of several variations of stutter disfluencies without the need for speech recognition. Our model uses a deep residual network and bidirectional long short-term memory layers to classify different types of stutters and achieves an average miss rate of 10.03%, outperforming the state-of-the-art by almost 27%.}, keywords = {Artificial Intelligence, Healthcare, Neural Networks, Signal Processing}, pubstate = {published}, tppubtype = {inproceedings} } Stuttering is a speech impediment affecting tens of millions of people on an everyday basis. Even with its commonality, there is minimal data and research on the identification and classification of stuttered speech. This paper tackles the problem of detection and classification of different forms of stutters. As opposed to most existing works that identify stutters with language models, our work proposes a model that relies solely on acoustic features, allowing for identification of several variations of stutter disfluencies without the need for speech recognition. Our model uses a deep residual network and bidirectional long short-term memory layers to classify different types of stutters and achieves an average miss rate of 10.03%, outperforming the state-of-the-art by almost 27%. |
Mitchell, Jordan; Marshall, Joshua A Towards a novel auto-rotating lidar platform for cavity surveying Journal Article Tunnelling and Underground Space Technology, 97 , pp. 103260, 2020. Abstract | Links | BibTeX | Tags: Mapping, Mining, Robotics @article{mitchell2020towards, title = {Towards a novel auto-rotating lidar platform for cavity surveying}, author = { Jordan Mitchell and Joshua A Marshall}, doi = {10.1016/j.tust.2019.103260}, year = {2020}, date = {2020-01-01}, journal = {Tunnelling and Underground Space Technology}, volume = {97}, pages = {103260}, publisher = {Elsevier}, abstract = {This paper presents the conceptual design, construction, and preliminary testing of a novel auto-rotating platform for cavity surveying. Cavity surveying involves generating a 3D model of an opening from acquired point cloud data. 3D models are used for volume estimation, stope reconciliation, dilution control, convergence monitoring, and rock mass stability analysis. In contrast to conventional cavity monitoring systems (CMS), which employ a rotating light detection and ranging (LiDAR) sensor attached to a stationary boom, the presented novel platform consists of a single point LiDAR sensor attached to an auto-rotating body as a unique solution to the cavity surveying problem. A proof of concept device consisting of a four-blade rotor and a data acquisition system was constructed and dropped from a five-metre overhang. Although auto-rotation was not reached, the set of indoor drop tests provide preliminary results that illustrate the design concept and describes the impact of design parameters.}, keywords = {Mapping, Mining, Robotics}, pubstate = {published}, tppubtype = {article} } This paper presents the conceptual design, construction, and preliminary testing of a novel auto-rotating platform for cavity surveying. Cavity surveying involves generating a 3D model of an opening from acquired point cloud data. 3D models are used for volume estimation, stope reconciliation, dilution control, convergence monitoring, and rock mass stability analysis. In contrast to conventional cavity monitoring systems (CMS), which employ a rotating light detection and ranging (LiDAR) sensor attached to a stationary boom, the presented novel platform consists of a single point LiDAR sensor attached to an auto-rotating body as a unique solution to the cavity surveying problem. A proof of concept device consisting of a four-blade rotor and a data acquisition system was constructed and dropped from a five-metre overhang. Although auto-rotation was not reached, the set of indoor drop tests provide preliminary results that illustrate the design concept and describes the impact of design parameters. |
Mazumder, Joy; Zand, Mohsen; Ziauddin, Sheikh; Greenspan, Michael ValidNet: A Deep Learning Network for Validation of Surface Registration Inproceedings pp. 389-397, 2020. Abstract | Links | BibTeX | Tags: Computer Vision, Neural Networks @inproceedings{Mazumder2020, title = {ValidNet: A Deep Learning Network for Validation of Surface Registration}, author = {Joy Mazumder and Mohsen Zand and Sheikh Ziauddin and Michael Greenspan}, doi = {10.5220/0008988103890397}, year = {2020}, date = {2020-01-01}, pages = {389-397}, abstract = {This paper proposes a novel deep learning architecture called ValidNet to automatically validate 3D surface registration algorithms for object recognition and pose determination tasks. The performance of many tasks such as object detection mainly depends on the applied registration algorithms, which themselves are susceptible to local minima. Revealing this tendency and verifying the success of registration algorithms is a difficult task. We treat this as a classification problem, and propose a two-class classifier to distinguish clearly between true positive and false positive instances. Our proposed ValidNet deploys a shared mlp architecture which works on the raw and unordered numeric data of scene and model points. This network is able to perform two fundamental tasks of feature extraction and similarity matching using the powerful capability of deep neural network. Experiments on a large synthetic datasets show that the proposed method can effectively be used in automatic validation of registration.}, keywords = {Computer Vision, Neural Networks}, pubstate = {published}, tppubtype = {inproceedings} } This paper proposes a novel deep learning architecture called ValidNet to automatically validate 3D surface registration algorithms for object recognition and pose determination tasks. The performance of many tasks such as object detection mainly depends on the applied registration algorithms, which themselves are susceptible to local minima. Revealing this tendency and verifying the success of registration algorithms is a difficult task. We treat this as a classification problem, and propose a two-class classifier to distinguish clearly between true positive and false positive instances. Our proposed ValidNet deploys a shared mlp architecture which works on the raw and unordered numeric data of scene and model points. This network is able to perform two fundamental tasks of feature extraction and similarity matching using the powerful capability of deep neural network. Experiments on a large synthetic datasets show that the proposed method can effectively be used in automatic validation of registration. |
Wang, Lei; Sun, Yun; Li, Qingguo; Liu, Tao; Yi, Jingang Two Shank-Mounted IMUs-Based Gait Analysis and Classification for Neurological Disease Patients Journal Article IEEE Robotics and Automation Letters, 5 (2), pp. 1970–1976, 2020, ISSN: 2377-3766. Abstract | Links | BibTeX | Tags: Biomechanics, Classifier @article{wang2020two, title = {Two Shank-Mounted IMUs-Based Gait Analysis and Classification for Neurological Disease Patients}, author = { Lei Wang and Yun Sun and Qingguo Li and Tao Liu and Jingang Yi}, doi = {10.1109/LRA.2020.2970656}, issn = {2377-3766}, year = {2020}, date = {2020-01-01}, journal = {IEEE Robotics and Automation Letters}, volume = {5}, number = {2}, pages = {1970--1976}, publisher = {IEEE}, abstract = {Automatic gait measurement and analysis is an enabling tool for intelligent healthcare and robotics-assisted rehabilitation. This letter proposes a novel two shank-mounted inertial measurement units (IMU)-based method on gait analysis and classification for three different neurological diseases. The IMU-based gait analysis and design aims to be applied in personal daily activities and environment for remote diagnosis and rehabilitation guidance. In the design, eight spatial-temporal and kinematic gait parameters are extracted from two shank-mounted IMUs. A support vector machine-based classifier is developed to classify four types of gait patterns with different neurological diseases (healthy control, peripheral neuropathy, post-stroke and Parkinson's disease). A total of 49 subjects are recruited and 93.9% of them are assigned to the right group after the leave-one-subject-out cross validation. The results demonstrate that the proposed IMU-based gait parameters and classifier are capable of differentiating the four types of gait patterns. The analysis and design have great potentials for use in clinical applications.}, keywords = {Biomechanics, Classifier}, pubstate = {published}, tppubtype = {article} } Automatic gait measurement and analysis is an enabling tool for intelligent healthcare and robotics-assisted rehabilitation. This letter proposes a novel two shank-mounted inertial measurement units (IMU)-based method on gait analysis and classification for three different neurological diseases. The IMU-based gait analysis and design aims to be applied in personal daily activities and environment for remote diagnosis and rehabilitation guidance. In the design, eight spatial-temporal and kinematic gait parameters are extracted from two shank-mounted IMUs. A support vector machine-based classifier is developed to classify four types of gait patterns with different neurological diseases (healthy control, peripheral neuropathy, post-stroke and Parkinson's disease). A total of 49 subjects are recruited and 93.9% of them are assigned to the right group after the leave-one-subject-out cross validation. The results demonstrate that the proposed IMU-based gait parameters and classifier are capable of differentiating the four types of gait patterns. The analysis and design have great potentials for use in clinical applications. |
Fan, Bingfei; Li, Qingguo; Liu, Tao Accurate foot clearance estimation during level and uneven ground walking using inertial sensors Journal Article Measurement Science and Technology, 31 (5), pp. 055106, 2020. Abstract | Links | BibTeX | Tags: Biomechanics @article{fan2020accurate, title = {Accurate foot clearance estimation during level and uneven ground walking using inertial sensors}, author = { Bingfei Fan and Qingguo Li and Tao Liu}, doi = {10.1088/1361-6501/ab6917}, year = {2020}, date = {2020-01-01}, journal = {Measurement Science and Technology}, volume = {31}, number = {5}, pages = {055106}, publisher = {IOP Publishing}, abstract = {Foot clearance during walking is considered as a key indicator for assessing fall risk, obstacle negotiation strategies and energy expenditure. Foot clearance estimation using inertial measurement units (IMUs) has the advantages of small size, low cost and user-friendliness. However, its application is still limited due to issues with accuracy and reliability. In this paper, we aimed at understanding the limiting factors in foot clearance estimation using low-cost IMUs and proposed a foot clearance estimation method with millimeter-level accuracy. We first analyzed each component in conventional double-integration-based foot clearance estimation, and then proposed a set of new procedures for foot trajectory estimation, including a gait-adaptive complementary filter for orientation estimation, a two-IMU configuration and a shock absorber. Finally, we extracted the foot clearance from the estimated foot trajectory. In the experiments, we recruited eight healthy subjects and instructed them to walk under level and uneven ground conditions; moreover, to validate the applicability of the proposed method, we also instructed the subjects to mimic pathological gaits, including ataxic gait (zigzag walking), waddling gait and Parkinsonian gait. A total of 2640 gait cycles were collected and the extracted foot clearances were benchmarked with the optical motion capture system. With the proposed method, the average mean and standard deviation of the extracted maximal heel clearance and minimal toe clearance in all the gait cycles were −0.34 ± 0.24 cm and 0.02 ± 0.26 cm. The results are more accurate than in previous studies. Most importantly, the proposed method does not require any post correction and flat-floor assumption. The presented foot clearance estimation method provides an applicable and practical clinical solution not only for heel and toe clearance estimation but also for foot trajectory estimation.}, keywords = {Biomechanics}, pubstate = {published}, tppubtype = {article} } Foot clearance during walking is considered as a key indicator for assessing fall risk, obstacle negotiation strategies and energy expenditure. Foot clearance estimation using inertial measurement units (IMUs) has the advantages of small size, low cost and user-friendliness. However, its application is still limited due to issues with accuracy and reliability. In this paper, we aimed at understanding the limiting factors in foot clearance estimation using low-cost IMUs and proposed a foot clearance estimation method with millimeter-level accuracy. We first analyzed each component in conventional double-integration-based foot clearance estimation, and then proposed a set of new procedures for foot trajectory estimation, including a gait-adaptive complementary filter for orientation estimation, a two-IMU configuration and a shock absorber. Finally, we extracted the foot clearance from the estimated foot trajectory. In the experiments, we recruited eight healthy subjects and instructed them to walk under level and uneven ground conditions; moreover, to validate the applicability of the proposed method, we also instructed the subjects to mimic pathological gaits, including ataxic gait (zigzag walking), waddling gait and Parkinsonian gait. A total of 2640 gait cycles were collected and the extracted foot clearances were benchmarked with the optical motion capture system. With the proposed method, the average mean and standard deviation of the extracted maximal heel clearance and minimal toe clearance in all the gait cycles were −0.34 ± 0.24 cm and 0.02 ± 0.26 cm. The results are more accurate than in previous studies. Most importantly, the proposed method does not require any post correction and flat-floor assumption. The presented foot clearance estimation method provides an applicable and practical clinical solution not only for heel and toe clearance estimation but also for foot trajectory estimation. |
Li, Tong; Li, Qingguo; Liu, Tao; Yi, Jingang How to Carry Loads Economically: Analysis Based on a Predictive Biped Model Journal Article Journal of Biomechanical Engineering, 142 (4), 2020, ISSN: 0148-0731. Abstract | Links | BibTeX | Tags: Biomechanics, Predictive Model @article{li2020carry, title = {How to Carry Loads Economically: Analysis Based on a Predictive Biped Model}, author = { Tong Li and Qingguo Li and Tao Liu and Jingang Yi}, doi = {10.1115/1.4044505}, issn = {0148-0731}, year = {2020}, date = {2020-01-01}, journal = {Journal of Biomechanical Engineering}, volume = {142}, number = {4}, publisher = {American Society of Mechanical Engineers Digital Collection}, abstract = {Carrying heavy loads costs additional energy during walking and leads to fatigue of the user. Conventionally, the load is fixed on the body. Some recent studies showed energy cost reduction when the relative motion of the load with respect to the body was allowed. However, the influences of the load's relative motion on the user are still not fully understood. We employed an optimization-based biped model, which can generate human-like walking motion to study the load–carrier interaction. The relative motion can be achieved by a passive mechanism (such as springs) or a powered mechanism (such as actuators), and the relative motion can occur in the vertical or fore-aft directions. The connection between the load and body is added to the biped model in four scenarios (two types × two directions). The optimization results indicate that the stiffness values affect energy cost differently and the same stiffness value in different directions may have opposite effects. Powered relative motion in either direction can potentially reduce energy cost but the vertical relative motion can achieve a higher reduction than fore-aft relative motion. Surprisingly, powered relative motion only performs marginally better than the passive conditions at similar peak interaction force levels. This work provides insights into developing more economical load-carrying methods and the model presented may be applied to the design and control of wearable load-carrying devices.}, keywords = {Biomechanics, Predictive Model}, pubstate = {published}, tppubtype = {article} } Carrying heavy loads costs additional energy during walking and leads to fatigue of the user. Conventionally, the load is fixed on the body. Some recent studies showed energy cost reduction when the relative motion of the load with respect to the body was allowed. However, the influences of the load's relative motion on the user are still not fully understood. We employed an optimization-based biped model, which can generate human-like walking motion to study the load–carrier interaction. The relative motion can be achieved by a passive mechanism (such as springs) or a powered mechanism (such as actuators), and the relative motion can occur in the vertical or fore-aft directions. The connection between the load and body is added to the biped model in four scenarios (two types × two directions). The optimization results indicate that the stiffness values affect energy cost differently and the same stiffness value in different directions may have opposite effects. Powered relative motion in either direction can potentially reduce energy cost but the vertical relative motion can achieve a higher reduction than fore-aft relative motion. Surprisingly, powered relative motion only performs marginally better than the passive conditions at similar peak interaction force levels. This work provides insights into developing more economical load-carrying methods and the model presented may be applied to the design and control of wearable load-carrying devices. |
Moulton, Richard Hugh; Marasco, Anthony J; Rudie, Karen Limited Lookahead Policies for the Control of Discrete-Event Systems: A Tutorial Journal Article arXiv preprint arXiv:2006.06514, 2020. Abstract | Links | BibTeX | Tags: Discrete Event Systems @article{moulton2020limited, title = {Limited Lookahead Policies for the Control of Discrete-Event Systems: A Tutorial}, author = { Richard Hugh Moulton and Anthony J Marasco and Karen Rudie}, url = {https://arxiv.org/abs/2006.06514}, year = {2020}, date = {2020-01-01}, journal = {arXiv preprint arXiv:2006.06514}, abstract = {Some problems in discrete-event systems (DES) model large, time-varying state spaces with complex legal languages. To address these problems, Chung et al. introduced limited lookahead policies (LLP) to provide online supervisory control for discrete-event systems. This seminal paper, along with an addendum of technical results, provided the field with a series of very important and powerful results, but in a notationally- and conceptually-dense manner. In this tutorial, we present Chung et al.'s problem formulation for online control and unravel the formal definitions and proofs from their original work with the aim of making the ideas behind limited lookahead accessible to all DES researchers. Finally, we introduce the Air Traffic Control problem as an example of an online control problem and demonstrate the synthesis of LLP supervisors. }, keywords = {Discrete Event Systems}, pubstate = {published}, tppubtype = {article} } Some problems in discrete-event systems (DES) model large, time-varying state spaces with complex legal languages. To address these problems, Chung et al. introduced limited lookahead policies (LLP) to provide online supervisory control for discrete-event systems. This seminal paper, along with an addendum of technical results, provided the field with a series of very important and powerful results, but in a notationally- and conceptually-dense manner. In this tutorial, we present Chung et al.'s problem formulation for online control and unravel the formal definitions and proofs from their original work with the aim of making the ideas behind limited lookahead accessible to all DES researchers. Finally, we introduce the Air Traffic Control problem as an example of an online control problem and demonstrate the synthesis of LLP supervisors. |
Adebola, Titilope O; Hoult, Neil; Moore, Ian D Distributed Strain Sensing to Study a Composite Liner for Cast Iron Water Pipe Rehabilitation Journal Article Journal of Testing and Evaluation, 48 (6), 2020, ISSN: 0090-3973. Abstract | Links | BibTeX | Tags: Civil Infrastructure, Fibre Optic Sensors @article{adebola2020distributed, title = {Distributed Strain Sensing to Study a Composite Liner for Cast Iron Water Pipe Rehabilitation}, author = { Titilope O Adebola and Neil Hoult and Ian D Moore}, doi = {10.1520/JTE20170497}, issn = {0090-3973}, year = {2020}, date = {2020-01-01}, journal = {Journal of Testing and Evaluation}, volume = {48}, number = {6}, publisher = {ASTM International}, abstract = {The use of cured-in-place pipe lining systems for the rehabilitation of cast iron water pressure pipe has grown significantly in the past 20 years. With limited information on the performance of these lining systems in service, strain measurements are needed to evaluate theoretical findings about three-dimensional strength limits. The application of distributed optical fiber sensing to study the properties of a polymer composite liner was investigated in this research. Polyimide optical fibers were installed into flat liner coupons fabricated in the laboratory, and the specimens were subsequently tested in uniaxial tension and four-point bending. Using optical frequency domain reflectometry, Rayleigh backscatter was monitored from preinstalled optical fiber sensors and strains of up to 4,000 μϵ were obtained. Comparisons between experimental optical fiber measurements and results obtained from conventional strain measurement techniques (i.e., extensometer and crosshead displacement) showed a strong correlation with an average difference of 3 % in four-point bending and 6 % in tension. However, the orientation and diameter of the composite fiber reinforcements were observed to affect distributed sensing performance.}, keywords = {Civil Infrastructure, Fibre Optic Sensors}, pubstate = {published}, tppubtype = {article} } The use of cured-in-place pipe lining systems for the rehabilitation of cast iron water pressure pipe has grown significantly in the past 20 years. With limited information on the performance of these lining systems in service, strain measurements are needed to evaluate theoretical findings about three-dimensional strength limits. The application of distributed optical fiber sensing to study the properties of a polymer composite liner was investigated in this research. Polyimide optical fibers were installed into flat liner coupons fabricated in the laboratory, and the specimens were subsequently tested in uniaxial tension and four-point bending. Using optical frequency domain reflectometry, Rayleigh backscatter was monitored from preinstalled optical fiber sensors and strains of up to 4,000 μϵ were obtained. Comparisons between experimental optical fiber measurements and results obtained from conventional strain measurement techniques (i.e., extensometer and crosshead displacement) showed a strong correlation with an average difference of 3 % in four-point bending and 6 % in tension. However, the orientation and diameter of the composite fiber reinforcements were observed to affect distributed sensing performance. |
Liu, Yuchen; Moore, Ian D; Hoult, Neil A Field Monitoring of a Corrugated Steel Culvert Using Multiple Sensing Technologies Journal Article Journal of Pipeline Systems Engineering and Practice, 11 (3), pp. 04020030, 2020. Abstract | Links | BibTeX | Tags: @article{liu2020field, title = {Field Monitoring of a Corrugated Steel Culvert Using Multiple Sensing Technologies}, author = { Yuchen Liu and Ian D Moore and Neil A Hoult}, doi = {10.1061/(ASCE)PS.1949-1204.0000477}, year = {2020}, date = {2020-01-01}, journal = {Journal of Pipeline Systems Engineering and Practice}, volume = {11}, number = {3}, pages = {04020030}, publisher = {American Society of Civil Engineers}, abstract = {Quantitative measurements are necessary to support accurate assessments of culverts; however, the choice of which sensor to employ to support these assessments is often not clear. An in-service corrugated steel culvert with a 3.1-m span was tested under static truck loading and monitored with fiber-optic strain sensors, a total station, digital image correlation (DIC), and lidar. Distributions of strains around the circumference were successfully captured by fiber-optic sensors, showing how the position of the peak thrust depended on the truck location, and that the peak flexural stress was double the peak hoop stress (contrary to strength limits considered within North American design codes). The peak thrust and moment occurred when the truck wheel pair was directly over the crown, as inferred in previous studies. The live load was found to be distributed nonuniformly along the longitudinal axis. Both the total station and DIC were feasible options for measuring displacements at the crown, while lidar was highly effective at recording the overall geometry.}, keywords = {}, pubstate = {published}, tppubtype = {article} } Quantitative measurements are necessary to support accurate assessments of culverts; however, the choice of which sensor to employ to support these assessments is often not clear. An in-service corrugated steel culvert with a 3.1-m span was tested under static truck loading and monitored with fiber-optic strain sensors, a total station, digital image correlation (DIC), and lidar. Distributions of strains around the circumference were successfully captured by fiber-optic sensors, showing how the position of the peak thrust depended on the truck location, and that the peak flexural stress was double the peak hoop stress (contrary to strength limits considered within North American design codes). The peak thrust and moment occurred when the truck wheel pair was directly over the crown, as inferred in previous studies. The live load was found to be distributed nonuniformly along the longitudinal axis. Both the total station and DIC were feasible options for measuring displacements at the crown, while lidar was highly effective at recording the overall geometry. |
Broth, Zachary E; Hoult, Neil A Field Monitoring of RC-Structures under Dynamic Loading Using Distributed Fiber-Optic Sensors Journal Article Journal of Performance of Constructed Facilities, 34 (4), pp. 04020070, 2020. Abstract | Links | BibTeX | Tags: Civil Infrastructure, Fibre Optic Sensors @article{broth2020field, title = {Field Monitoring of RC-Structures under Dynamic Loading Using Distributed Fiber-Optic Sensors}, author = { Zachary E Broth and Neil A Hoult}, doi = {10.1061/(ASCE)CF.1943-5509.0001488}, year = {2020}, date = {2020-01-01}, journal = {Journal of Performance of Constructed Facilities}, volume = {34}, number = {4}, pages = {04020070}, publisher = {American Society of Civil Engineers}, abstract = {The ability to properly assess existing reinforced concrete (RC) structures provides an opportunity to prevent costly rehabilitations or replacements and aid in the optimization of future designs. Distributed fiber-optic sensing (DFOS) is a promising option in the assessment of these complex structures. Previous uses of this technology, however, have been limited to static measurements, prohibiting the assessment of structures exposed to dynamic loads. This research intends to assess the dynamic sensing capabilities of a DFOS system and, in turn, improve the current understanding of the dynamic behavior of an existing RC beam tested in situ through a case study. Based on the data provided by the dynamic distributed fiber-optic sensors (DDFOS), dynamic measurements of distributed strains, distributed deflections, and crack widths can all be provided. Further details regarding the use of the DDFOS system in assessment, including the determination of support conditions and dynamic response factors, are presented and discussed.}, keywords = {Civil Infrastructure, Fibre Optic Sensors}, pubstate = {published}, tppubtype = {article} } The ability to properly assess existing reinforced concrete (RC) structures provides an opportunity to prevent costly rehabilitations or replacements and aid in the optimization of future designs. Distributed fiber-optic sensing (DFOS) is a promising option in the assessment of these complex structures. Previous uses of this technology, however, have been limited to static measurements, prohibiting the assessment of structures exposed to dynamic loads. This research intends to assess the dynamic sensing capabilities of a DFOS system and, in turn, improve the current understanding of the dynamic behavior of an existing RC beam tested in situ through a case study. Based on the data provided by the dynamic distributed fiber-optic sensors (DDFOS), dynamic measurements of distributed strains, distributed deflections, and crack widths can all be provided. Further details regarding the use of the DDFOS system in assessment, including the determination of support conditions and dynamic response factors, are presented and discussed. |
Pannese, Eric M; Take, Andy W; Hoult, Neil A; Yan, Ruobing; Le, Hoat Bridge transition monitoring: Interpretation of track defects using digital image correlation and distributed fiber optic strain sensing Journal Article Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 234 (6), pp. 616–637, 2020. Abstract | Links | BibTeX | Tags: Civil Infrastructure, Fibre Optic Sensors, Railway @article{pannese2020bridge, title = {Bridge transition monitoring: Interpretation of track defects using digital image correlation and distributed fiber optic strain sensing}, author = { Eric M Pannese and W Andy Take and Neil A Hoult and Ruobing Yan and Hoat Le}, doi = {10.1177/0954409719851626}, year = {2020}, date = {2020-01-01}, journal = {Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit}, volume = {234}, number = {6}, pages = {616--637}, publisher = {SAGE Publications Sage UK: London, England}, abstract = {Railway bridge transitions represent locations of vertical stiffness variations that are believed to amplify the dynamic wheel loads which contribute to the development of both differential track settlement and hanging sleeper issues that are difficult to resolve despite the current knowledge of bridge transition behaviour. This paper presents a field monitoring study of a railway bridge transition in which track defects – including both gaps between the rail and sleeper plates, i.e. “rail–sleeper gaps,” and rail flange scrape marks exposing bare steel – were observed during a visual inspection of the transition. Trains were monitored by measuring both track displacements using Digital Image Correlation and distributed rail strains using a Rayleigh-based fiber optic analyzer. Through analysis and interpretation of the collected monitoring data, it was found that measurements of rail–sleeper gaps could be used to obtain a first-order estimate of the shape of the differential track settlement profile. Additionally, it was found that measurements of scrape marks on the rail flange could be used to estimate the extent of longitudinal rail movement that could occur during train passage, and that the loads applied to the bridge structure were influenced by the nature of the rail–sleeper gaps at the monitoring site.}, keywords = {Civil Infrastructure, Fibre Optic Sensors, Railway}, pubstate = {published}, tppubtype = {article} } Railway bridge transitions represent locations of vertical stiffness variations that are believed to amplify the dynamic wheel loads which contribute to the development of both differential track settlement and hanging sleeper issues that are difficult to resolve despite the current knowledge of bridge transition behaviour. This paper presents a field monitoring study of a railway bridge transition in which track defects – including both gaps between the rail and sleeper plates, i.e. “rail–sleeper gaps,” and rail flange scrape marks exposing bare steel – were observed during a visual inspection of the transition. Trains were monitored by measuring both track displacements using Digital Image Correlation and distributed rail strains using a Rayleigh-based fiber optic analyzer. Through analysis and interpretation of the collected monitoring data, it was found that measurements of rail–sleeper gaps could be used to obtain a first-order estimate of the shape of the differential track settlement profile. Additionally, it was found that measurements of scrape marks on the rail flange could be used to estimate the extent of longitudinal rail movement that could occur during train passage, and that the loads applied to the bridge structure were influenced by the nature of the rail–sleeper gaps at the monitoring site. |
Liu, Yuchen; Hoult, Neil A; Moore, Ian D Structural performance of in-service corrugated steel culvert under vehicle loading Journal Article Journal of Bridge Engineering, 25 (3), pp. 04019142, 2020. Abstract | Links | BibTeX | Tags: Civil Infrastructure, Fibre Optic Sensors @article{liu2020structural, title = {Structural performance of in-service corrugated steel culvert under vehicle loading}, author = { Yuchen Liu and Neil A Hoult and Ian D Moore}, doi = {10.1061/(ASCE)BE.1943-5592.0001524}, year = {2020}, date = {2020-01-01}, journal = {Journal of Bridge Engineering}, volume = {25}, number = {3}, pages = {04019142}, publisher = {American Society of Civil Engineers}, abstract = {Dynamic distributed fiber optic strain sensors were used, for the first time, to monitor the behavior of an in-service corrugated steel culvert with a span of 2.75 m and a skew angle of 45° under static and dynamic vehicle loading. These sensors enabled thrust and moment distributions to be measured around the circumference of the culvert. The peak thrusts were not found at the springlines as assumed by North American design codes. Additionally, the flexural stresses were of a similar magnitude to the hoop stresses, contrary to the assumption of ring compression used in many design codes. Dynamic truck loading resulted in higher thrusts and lower moments than did static loading. For this culvert, the deteriorated asphalt pavement had little influence on thrusts but did reduce the magnitude of the moments. Maximum thrust envelopes were developed using the distributed data, and demonstrated that positions near the crown experienced thrusts higher than those at the springlines. The peak measured strain was 9.3% of the yield strain, suggesting that the culvert had adequate live load capacity.}, keywords = {Civil Infrastructure, Fibre Optic Sensors}, pubstate = {published}, tppubtype = {article} } Dynamic distributed fiber optic strain sensors were used, for the first time, to monitor the behavior of an in-service corrugated steel culvert with a span of 2.75 m and a skew angle of 45° under static and dynamic vehicle loading. These sensors enabled thrust and moment distributions to be measured around the circumference of the culvert. The peak thrusts were not found at the springlines as assumed by North American design codes. Additionally, the flexural stresses were of a similar magnitude to the hoop stresses, contrary to the assumption of ring compression used in many design codes. Dynamic truck loading resulted in higher thrusts and lower moments than did static loading. For this culvert, the deteriorated asphalt pavement had little influence on thrusts but did reduce the magnitude of the moments. Maximum thrust envelopes were developed using the distributed data, and demonstrated that positions near the crown experienced thrusts higher than those at the springlines. The peak measured strain was 9.3% of the yield strain, suggesting that the culvert had adequate live load capacity. |
Hoult, Neil A; Glisic, Branko Structural Health Monitoring of Bridges Journal Article Frontiers in Built Environment, 6 , pp. 17, 2020, ISSN: 2297-3362. @article{hoult2020structural, title = {Structural Health Monitoring of Bridges}, author = { Neil A Hoult and Branko Glisic}, doi = {10.3389/fbuil.2020.00017}, issn = {2297-3362}, year = {2020}, date = {2020-01-01}, journal = {Frontiers in Built Environment}, volume = {6}, pages = {17}, publisher = {Frontiers}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Broth, Zachary; Hoult, Neil A Dynamic distributed strain sensing to assess reinforced concrete behaviour Journal Article Engineering Structures, 204 , pp. 110036, 2020, ISSN: 0141-0296. Abstract | Links | BibTeX | Tags: Civil Infrastructure, Fibre Optic Sensors @article{broth2020dynamic, title = {Dynamic distributed strain sensing to assess reinforced concrete behaviour}, author = { Zachary Broth and Neil A Hoult}, doi = {10.1016/j.engstruct.2019.110036}, issn = {0141-0296}, year = {2020}, date = {2020-01-01}, journal = {Engineering Structures}, volume = {204}, pages = {110036}, publisher = {Elsevier}, abstract = {To investigate the potential of using dynamic distributed fibre optic sensing (DDFOS), a series of slender and deep reinforced concrete beams were tested. The beams were loaded cyclically using combinations of three and four-point bending and then tested to ultimate while DDFOS was used to measure the reinforcement strains. The DDFOS data was used to assess the increase in deflections due to cyclic loading as well as the change from beam action to arching action in some specimens. The strain profiles were able to capture the moment of failure including critical location in terms of peak reinforcement strains.}, keywords = {Civil Infrastructure, Fibre Optic Sensors}, pubstate = {published}, tppubtype = {article} } To investigate the potential of using dynamic distributed fibre optic sensing (DDFOS), a series of slender and deep reinforced concrete beams were tested. The beams were loaded cyclically using combinations of three and four-point bending and then tested to ultimate while DDFOS was used to measure the reinforcement strains. The DDFOS data was used to assess the increase in deflections due to cyclic loading as well as the change from beam action to arching action in some specimens. The strain profiles were able to capture the moment of failure including critical location in terms of peak reinforcement strains. |
Tetreault, Jacob; Moore, Ian D; Hoult, Neil A Laboratory Study on Effect of Grout Choice on Culvert Rehabilitation Using Sliplining Journal Article Journal of Pipeline Systems Engineering and Practice, 11 (1), pp. 04019044, 2020. Abstract | Links | BibTeX | Tags: Civil Infrastructure @article{tetreault2020laboratory, title = {Laboratory Study on Effect of Grout Choice on Culvert Rehabilitation Using Sliplining}, author = { Jacob Tetreault and Ian D Moore and Neil A Hoult}, doi = {10.1061/(ASCE)PS.1949-1204.0000420}, year = {2020}, date = {2020-01-01}, journal = {Journal of Pipeline Systems Engineering and Practice}, volume = {11}, number = {1}, pages = {04019044}, publisher = {American Society of Civil Engineers}, abstract = {Two corrugated steel pipes were repaired with grouted high-density polyethylene (HDPE) slipliners, one with low-density (i.e., low-modulus and low-strength) grout, and the other with high-density (i.e., high-modulus and high-strength) grout. The responses of the rehabilitated culverts were measured under service loads and fully factored design loads, as well as even higher load levels, to establish the ultimate limit states. Finally, the test culverts were exhumed and the damage was observed. Sliplining increased the stiffness of both culverts, and their strength then exceeded the required load carrying capacity by more than a factor of 2. Although the lined pipe with low-density grout behaved like a rigid pipe during service loading, it began to behave more flexibly at higher loads. At very high loads, the liner deflections exceeded levels permitted for HDPE pipes. Extensive shear failure and crushing of the grout were observed after the test. The sliplined pipe with high-density grout behaved like a rigid pipe during both service load and ultimate load experiments. Distributions of hoop strain through the liner–grout–steel pipe system demonstrated bonded, composite behavior at some sections of the system repaired using high-density grout, and noncomposite behavior for the system repaired using low-density grout.}, keywords = {Civil Infrastructure}, pubstate = {published}, tppubtype = {article} } Two corrugated steel pipes were repaired with grouted high-density polyethylene (HDPE) slipliners, one with low-density (i.e., low-modulus and low-strength) grout, and the other with high-density (i.e., high-modulus and high-strength) grout. The responses of the rehabilitated culverts were measured under service loads and fully factored design loads, as well as even higher load levels, to establish the ultimate limit states. Finally, the test culverts were exhumed and the damage was observed. Sliplining increased the stiffness of both culverts, and their strength then exceeded the required load carrying capacity by more than a factor of 2. Although the lined pipe with low-density grout behaved like a rigid pipe during service loading, it began to behave more flexibly at higher loads. At very high loads, the liner deflections exceeded levels permitted for HDPE pipes. Extensive shear failure and crushing of the grout were observed after the test. The sliplined pipe with high-density grout behaved like a rigid pipe during both service load and ultimate load experiments. Distributions of hoop strain through the liner–grout–steel pipe system demonstrated bonded, composite behavior at some sections of the system repaired using high-density grout, and noncomposite behavior for the system repaired using low-density grout. |
Ramos, Andrés; Hashtrudi-Zaad, Keyvan Estimation of Upper-Limb Energy Absorption Capabilities for Stable Human-Robot Interactions Inproceedings 2020 IEEE Haptics Symposium (HAPTICS), pp. 115–120, IEEE 2020, ISSN: 2324-7355. Abstract | Links | BibTeX | Tags: Biomechanics, Healthcare, Robotics @inproceedings{ramos2020estimation, title = {Estimation of Upper-Limb Energy Absorption Capabilities for Stable Human-Robot Interactions}, author = { Andrés Ramos and Keyvan Hashtrudi-Zaad}, doi = {10.1109/HAPTICS45997.2020.ras.HAP20.20.95bee409}, issn = {2324-7355}, year = {2020}, date = {2020-01-01}, booktitle = {2020 IEEE Haptics Symposium (HAPTICS)}, pages = {115--120}, organization = {IEEE}, abstract = {Human-robot interactions are becoming more and more prevalent in various aspects of life, enhancing humans' mobility, accessibility and health. However, safety measures need to be addressed when applying robotic-generated forces that put human users at risk. One way to improve safety and performance in robotic tasks is to include physiological information, such as damping properties of the arm, in the control system to help regulate the energy that is delivered to the user. In this work, we estimated the energy absorbing capabilities of the human arm, based on the metric Excess of Passivity (EOP), originally defined in [1]. We used data from healthy subjects to generate models that fit different levels of safety and stability. Variability in subjects' EOP was a major finding in this study. For demanding applications such as robotic rehabilitation therapy, we suggest using a linear model with two EOP points. Such points are the mean values of EOP estimations at relaxed and rigid levels of hand-grasp forces. Two standard deviations were subtracted from each EOP point to consider the variability due to the neuromuscular changes in the human arm.}, keywords = {Biomechanics, Healthcare, Robotics}, pubstate = {published}, tppubtype = {inproceedings} } Human-robot interactions are becoming more and more prevalent in various aspects of life, enhancing humans' mobility, accessibility and health. However, safety measures need to be addressed when applying robotic-generated forces that put human users at risk. One way to improve safety and performance in robotic tasks is to include physiological information, such as damping properties of the arm, in the control system to help regulate the energy that is delivered to the user. In this work, we estimated the energy absorbing capabilities of the human arm, based on the metric Excess of Passivity (EOP), originally defined in [1]. We used data from healthy subjects to generate models that fit different levels of safety and stability. Variability in subjects' EOP was a major finding in this study. For demanding applications such as robotic rehabilitation therapy, we suggest using a linear model with two EOP points. Such points are the mean values of EOP estimations at relaxed and rigid levels of hand-grasp forces. Two standard deviations were subtracted from each EOP point to consider the variability due to the neuromuscular changes in the human arm. |
Laija, Victor Luna A; Ozdil, Gizem P; Hashtrudi-Zaad, Keyvan Effect of Direct Velocity Measurement on the Stability of Haptic Simulation Systems Inproceedings 2020 IEEE Haptics Symposium (HAPTICS), pp. 712–717, IEEE 2020, ISBN: 2324-7355 . Abstract | Links | BibTeX | Tags: Haptics @inproceedings{laija2020effect, title = {Effect of Direct Velocity Measurement on the Stability of Haptic Simulation Systems}, author = { Victor A Luna Laija and P Gizem Ozdil and Keyvan Hashtrudi-Zaad}, doi = {10.1109/HAPTICS45997.2020.ras.HAP20.27.fb93aa04}, isbn = {2324-7355 }, year = {2020}, date = {2020-01-01}, booktitle = {2020 IEEE Haptics Symposium (HAPTICS)}, pages = {712--717}, organization = {IEEE}, abstract = {The quality of the velocity signal required to implement linear viscoelastic virtual environments is of great importance for the stability of haptic simulation systems. In a typical haptic simulation system, the velocity is numerically estimated from position samples. The range of the implementable environment stiffness and damping for uncoupled stability is limited by quantization. In this paper, we analytically and numerically studied the effect of direct velocity measurement on the virtual environment viscoelastic dynamic range for guaranteed uncoupled stability and the passivity of the dynamics projected to the user. The analytical results are assessed through experiments conducted in a QET haptic knob testbed.}, keywords = {Haptics}, pubstate = {published}, tppubtype = {inproceedings} } The quality of the velocity signal required to implement linear viscoelastic virtual environments is of great importance for the stability of haptic simulation systems. In a typical haptic simulation system, the velocity is numerically estimated from position samples. The range of the implementable environment stiffness and damping for uncoupled stability is limited by quantization. In this paper, we analytically and numerically studied the effect of direct velocity measurement on the virtual environment viscoelastic dynamic range for guaranteed uncoupled stability and the passivity of the dynamics projected to the user. The analytical results are assessed through experiments conducted in a QET haptic knob testbed. |
Wiercigroch, Julia; Hashtrudi-Zaad, Keyvan; Ungi, Tamas; Bisleri, Gianluigi; Fichtinger, Gabor Force and torque feedback in endoscopic vessel harvesting Inproceedings Fei, Baowei; Linte, Cristian A (Ed.): Medical Imaging 2020: Image-Guided Procedures, Robotic Interventions, and Modeling, pp. 113152Q, International Society for Optics and Photonics 2020. Abstract | Links | BibTeX | Tags: Haptics, Healthcare, Robotics @inproceedings{wiercigroch2020force, title = {Force and torque feedback in endoscopic vessel harvesting}, author = { Julia Wiercigroch and Keyvan Hashtrudi-Zaad and Tamas Ungi and Gianluigi Bisleri and Gabor Fichtinger}, editor = {Baowei Fei and Cristian A. Linte}, doi = {10.1117/12.2550160}, year = {2020}, date = {2020-01-01}, booktitle = {Medical Imaging 2020: Image-Guided Procedures, Robotic Interventions, and Modeling}, volume = {11315}, pages = {113152Q}, organization = {International Society for Optics and Photonics}, abstract = {PURPOSE: Endoscopic vessel harvesting is the preferred minimally invasive approach to obtain grafts for coronary bypass surgeries, however it requires extensive practice to minimize vessel damage. We propose to create a surgical training simulation with visual and haptic feedback. In this study, we focus on analyzing the force and torque peaks on the surgical retractor during the procedure. METHODS: The original retractor handle was 3D scanned and modified to attach an ATI Mini40 force-torque transducer. The forces and torques in two radial artery and two saphenous vein procedures in human cadavers were recorded. The measurements, endoscopic video and surgical surface video were collected. The median and interquartile range of the force and torque peaks were calculated for the artery and vein harvesting procedures. RESULTS: The median and interquartile range for saphenous vein harvests was larger than radial artery harvests. The largest median force and torque generated in the vein was 11.654 N [posterior] and 0.661 Nm [- frontal], whereas in the artery was 6.163 N [anterior] and 0.381 Nm [+ frontal], respectively. CONCLUSION: The distribution of force and torque peaks in the retractor was found for endoscopic vessel harvests. This data can be used to design a haptic user interface, and to establish expert benchmarks for learning curve evaluation.}, keywords = {Haptics, Healthcare, Robotics}, pubstate = {published}, tppubtype = {inproceedings} } PURPOSE: Endoscopic vessel harvesting is the preferred minimally invasive approach to obtain grafts for coronary bypass surgeries, however it requires extensive practice to minimize vessel damage. We propose to create a surgical training simulation with visual and haptic feedback. In this study, we focus on analyzing the force and torque peaks on the surgical retractor during the procedure. METHODS: The original retractor handle was 3D scanned and modified to attach an ATI Mini40 force-torque transducer. The forces and torques in two radial artery and two saphenous vein procedures in human cadavers were recorded. The measurements, endoscopic video and surgical surface video were collected. The median and interquartile range of the force and torque peaks were calculated for the artery and vein harvesting procedures. RESULTS: The median and interquartile range for saphenous vein harvests was larger than radial artery harvests. The largest median force and torque generated in the vein was 11.654 N [posterior] and 0.661 Nm [- frontal], whereas in the artery was 6.163 N [anterior] and 0.381 Nm [+ frontal], respectively. CONCLUSION: The distribution of force and torque peaks in the retractor was found for endoscopic vessel harvests. This data can be used to design a haptic user interface, and to establish expert benchmarks for learning curve evaluation. |
Qiu, Yubo; Malcolm, Jonathon; Ziauddin, Sheikh; Greenspan, Michael; Vatoo, Abhay Inverse Rectification for Efficient Procam Pattern Correspondence Inproceedings The IEEE Winter Conference on Applications of Computer Vision, pp. 710–718, 2020. Abstract | Links | BibTeX | Tags: Computer Vision @inproceedings{qiu2020inverse, title = {Inverse Rectification for Efficient Procam Pattern Correspondence}, author = { Yubo Qiu and Jonathon Malcolm and Sheikh Ziauddin and Michael Greenspan and Abhay Vatoo}, url = {https://openaccess.thecvf.com/content_WACV_2020/html/Qiu_Inverse_Rectification_for_Efficient_Procam_Pattern_Correspondence_WACV_2020_paper.html}, year = {2020}, date = {2020-01-01}, booktitle = {The IEEE Winter Conference on Applications of Computer Vision}, pages = {710--718}, abstract = {A method called inverse rectification, is proposed which facilitates the establishment of correspondences across a projected pattern and an acquired image. A pattern of features comprising vertical dashes is warped by the inverse of the rectifying homography of the projector-camera pair, prior to projection. This warping imparts upon the system the property that projected features will fall on distinct conjugate epipolar lines of the rectified projector and acquired camera images. This reduces the correspondence search to a trivial constant-time table lookup once a feature is found in the camera image, and leads to robust, accurate, and extremely efficient disparity calculations. A projector-camera range sensor is developed based on this method, and is shown experimentally to be effective, with bandwidth exceeding some existing consumer-level range sensors. }, keywords = {Computer Vision}, pubstate = {published}, tppubtype = {inproceedings} } A method called inverse rectification, is proposed which facilitates the establishment of correspondences across a projected pattern and an acquired image. A pattern of features comprising vertical dashes is warped by the inverse of the rectifying homography of the projector-camera pair, prior to projection. This warping imparts upon the system the property that projected features will fall on distinct conjugate epipolar lines of the rectified projector and acquired camera images. This reduces the correspondence search to a trivial constant-time table lookup once a feature is found in the camera image, and leads to robust, accurate, and extremely efficient disparity calculations. A projector-camera range sensor is developed based on this method, and is shown experimentally to be effective, with bandwidth exceeding some existing consumer-level range sensors. |
Muise, Christian; Lipovetzky, Nir KEPS Book: Planning. Domains Incollection Knowledge Engineering Tools and Techniques for AI Planning, pp. 91–105, Springer, 2020, ISBN: 978-3-030-38561-3. Abstract | Links | BibTeX | Tags: Artificial Intelligence, Modelling @incollection{muise2020keps, title = {KEPS Book: Planning. Domains}, author = { Christian Muise and Nir Lipovetzky}, doi = {10.1007/978-3-030-38561-3_5}, isbn = {978-3-030-38561-3}, year = {2020}, date = {2020-01-01}, booktitle = {Knowledge Engineering Tools and Techniques for AI Planning}, pages = {91--105}, publisher = {Springer}, abstract = {In this chapter we describe the main pillars of the Planning.Domains initiative (API, Solver, Editor, and Education), detail some of the current use-cases for them, and outline the future path of the initiative. We further dive into some of the most recent developments of Planning.Domains, and shed light on what is next for the platform.}, keywords = {Artificial Intelligence, Modelling}, pubstate = {published}, tppubtype = {incollection} } In this chapter we describe the main pillars of the Planning.Domains initiative (API, Solver, Editor, and Education), detail some of the current use-cases for them, and outline the future path of the initiative. We further dive into some of the most recent developments of Planning.Domains, and shed light on what is next for the platform. |
2020 |
What lies beneath: Material classification for autonomous excavators using proprioceptive force sensing and machine learning Journal Article 119 , pp. 103374, 2020, ISSN: 0926-5805. |
Wearable Sensor-Based Sign Language Recognition: A Comprehensive Review Journal Article IEEE Reviews in Biomedical Engineering, pp. 1-1, 2020, ISSN: 1941-1189. |
Deep multitask learning for pervasive BMI estimation and identity recognition in smart beds Journal Article Journal of Ambient Intelligence and Humanized Computing, pp. 1–15, 2020. |
Autorotating unmanned aerial vehicle surveying platform Patent 2020, (US Patent App. 16/046,436). |
--D3WA+--A Case Study of XAIP in a Model Acquisition Task for Dialogue Planning Inproceedings Proceedings of the International Conference on Automated Planning and Scheduling, pp. 488–497, 2020. |
Hand gesture-based control of a front-end loader Journal Article 2020. |
Systems, methods and devices for activity recognition Patent 2020, (US Patent 10,575,760). |
Reconfigurable Hybrid Energy Storage System for an Electric Vehicle DC/AC Inverter Journal Article IEEE Transactions on Power Electronics, 2020, ISSN: 1941-0107. |
TraceHub-A Platform to Bridge the Gap between State-of-the-Art Time-Series Analytics and Datasets. Inproceedings AAAI, pp. 13600–13601, 2020. |
Learning Neural-Symbolic Descriptive Planning Models via Cube-Space Priors: The Voyage Home (to STRIPS) Journal Article arXiv preprint arXiv:2004.12850, 2020. |
Explainability of Intelligent Transportation Systems using Knowledge Compilation: a Traffic Light Controller Case Journal Article arXiv preprint arXiv:2007.04916, 2020. |
A flexible machine vision system for small part inspection based on a hybrid SVM/ANN approach Journal Article Journal of Intelligent Manufacturing, 31 (1), pp. 103–125, 2020. |
Leveraging Data Engineering to Improve Unmanned Aerial Vehicle Control Design Journal Article IEEE Systems Journal, 2020, ISSN: 1937-9234 . |
FPGA-Based Design Optimization in Autonomous Robot Systems for Inspection of Civil Infrastructure Journal Article IEEE Systems Journal, 14 (2), pp. 2961–2964, 2020, ISSN: 1937-9234. |
2020 IEEE/ION Position, Location and Navigation Symposium (PLANS), pp. 960–966, IEEE 2020, ISSN: 2153-3598. |
SemEval-2020 task 4: Commonsense validation and explanation Journal Article arXiv preprint arXiv:2007.00236, 2020. |
Self-supervised learning for ecg-based emotion recognition Inproceedings ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 3217–3221, IEEE 2020, ISBN: 2379-190X . |
Upper body and ankle strategies compensate for reduced lateral stability at very slow walking speeds Journal Article bioRxiv, 2020. |
Learning Low-Resource End-To-End Goal-Oriented Dialog for Fast and Reliable System Deployment Inproceedings Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pp. 609–618, 2020. |
Generating diverse conversation responses by creating and ranking multiple candidates Journal Article Computer Speech & Language, 62 , pp. 101071, 2020. |
Exploring deep anomaly detection methods based on capsule net Inproceedings Canadian Conference on Artificial Intelligence, pp. 375–387, Springer 2020, ISBN: 978-3-030-47358-7. |
Pre-Trained and Attention-Based Neural Networks for Building Noetic Task-Oriented Dialogue Systems Journal Article arXiv preprint arXiv:2004.01940, 2020. |
CoMP-NOMA in the SWIPT Networks Journal Article IEEE Transactions on Wireless Communications, 2020, ISBN: 1558-2248. |
Dynamic Power Splitting for SWIPT With Nonlinear Energy Harvesting in Ergodic Fading Channel Journal Article IEEE Internet of Things Journal, 2020, ISSN: 2327-4662. |
Deep RNN-Based Channel Tracking for Wireless Energy Transfer System Journal Article IEEE Systems Journal, 2020, ISSN: 1937-9234. |
Deep Artificial Noise: Deep Learning-Based Precoding Optimization for Artificial Noise Scheme Journal Article IEEE Transactions on Vehicular Technology, 69 (3), pp. 3465–3469, 2020, ISSN: 1939-9359. |
Toward the Design of Enjoyable Games for Children with Fetal Alcohol Spectrum Disorder Inproceedings Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, pp. 1–13, 2020. |
Attention-aware fusion RGB-D face recognition Journal Article arXiv preprint arXiv:2003.00168, 2020. |
Facial emotion recognition using light field images with deep attention-based bidirectional LSTM Inproceedings ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 3367–3371, IEEE 2020. |
Self-supervised ecg representation learning for emotion recognition Journal Article arXiv preprint arXiv:2002.03898, 2020. |
Detecting Multiple Speech Disfluencies Using a Deep Residual Network with Bidirectional Long Short-Term Memory Inproceedings ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 6089–6093, IEEE 2020, ISSN: 2379-190X. |
Towards a novel auto-rotating lidar platform for cavity surveying Journal Article Tunnelling and Underground Space Technology, 97 , pp. 103260, 2020. |
ValidNet: A Deep Learning Network for Validation of Surface Registration Inproceedings pp. 389-397, 2020. |
Two Shank-Mounted IMUs-Based Gait Analysis and Classification for Neurological Disease Patients Journal Article IEEE Robotics and Automation Letters, 5 (2), pp. 1970–1976, 2020, ISSN: 2377-3766. |
Accurate foot clearance estimation during level and uneven ground walking using inertial sensors Journal Article Measurement Science and Technology, 31 (5), pp. 055106, 2020. |
How to Carry Loads Economically: Analysis Based on a Predictive Biped Model Journal Article Journal of Biomechanical Engineering, 142 (4), 2020, ISSN: 0148-0731. |
Limited Lookahead Policies for the Control of Discrete-Event Systems: A Tutorial Journal Article arXiv preprint arXiv:2006.06514, 2020. |
Distributed Strain Sensing to Study a Composite Liner for Cast Iron Water Pipe Rehabilitation Journal Article Journal of Testing and Evaluation, 48 (6), 2020, ISSN: 0090-3973. |
Field Monitoring of a Corrugated Steel Culvert Using Multiple Sensing Technologies Journal Article Journal of Pipeline Systems Engineering and Practice, 11 (3), pp. 04020030, 2020. |
Field Monitoring of RC-Structures under Dynamic Loading Using Distributed Fiber-Optic Sensors Journal Article Journal of Performance of Constructed Facilities, 34 (4), pp. 04020070, 2020. |
Bridge transition monitoring: Interpretation of track defects using digital image correlation and distributed fiber optic strain sensing Journal Article Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 234 (6), pp. 616–637, 2020. |
Structural performance of in-service corrugated steel culvert under vehicle loading Journal Article Journal of Bridge Engineering, 25 (3), pp. 04019142, 2020. |
Structural Health Monitoring of Bridges Journal Article Frontiers in Built Environment, 6 , pp. 17, 2020, ISSN: 2297-3362. |
Dynamic distributed strain sensing to assess reinforced concrete behaviour Journal Article Engineering Structures, 204 , pp. 110036, 2020, ISSN: 0141-0296. |
Laboratory Study on Effect of Grout Choice on Culvert Rehabilitation Using Sliplining Journal Article Journal of Pipeline Systems Engineering and Practice, 11 (1), pp. 04019044, 2020. |
Estimation of Upper-Limb Energy Absorption Capabilities for Stable Human-Robot Interactions Inproceedings 2020 IEEE Haptics Symposium (HAPTICS), pp. 115–120, IEEE 2020, ISSN: 2324-7355. |
Effect of Direct Velocity Measurement on the Stability of Haptic Simulation Systems Inproceedings 2020 IEEE Haptics Symposium (HAPTICS), pp. 712–717, IEEE 2020, ISBN: 2324-7355 . |
Force and torque feedback in endoscopic vessel harvesting Inproceedings Fei, Baowei; Linte, Cristian A (Ed.): Medical Imaging 2020: Image-Guided Procedures, Robotic Interventions, and Modeling, pp. 113152Q, International Society for Optics and Photonics 2020. |
Inverse Rectification for Efficient Procam Pattern Correspondence Inproceedings The IEEE Winter Conference on Applications of Computer Vision, pp. 710–718, 2020. |
KEPS Book: Planning. Domains Incollection Knowledge Engineering Tools and Techniques for AI Planning, pp. 91–105, Springer, 2020, ISBN: 978-3-030-38561-3. |