Pub Date : 2021-07-03DOI: 10.1109/ICARM52023.2021.9536140
Yang Liu, Ziyu Chen, Xiaodong Zhang, Jie Gao
The compliant peg-in-hole assembly method based on Attractive Region in Environment has achieved great performance in high-precision assembly for convex components. However, for nonconvex parts, there may be local minima in the constraint region, and jamming in the assembly process may arise by using the existing method. To solve the problem, this paper propose a compliant assembly method based on Attractive Region in Environment combined with geometric features. First, according to the geometric characteristics of the constraint region, sub-targets during the assembly process are designed. Then, low dimensional attractive regions are utilized to realize the phased directional movement and solve the jamming problem caused by the grooves. Furthermore, Impedance control is applied to guarantee the compliance of assembly. The experimental results of the nonconvex peg-in-hole assembly with clearance of 0.02 mm are presented and show the validity of the proposed method.
{"title":"Compliant Peg-in-Hole Assembly for Components with Grooves Based on Attractive Region in Environment","authors":"Yang Liu, Ziyu Chen, Xiaodong Zhang, Jie Gao","doi":"10.1109/ICARM52023.2021.9536140","DOIUrl":"https://doi.org/10.1109/ICARM52023.2021.9536140","url":null,"abstract":"The compliant peg-in-hole assembly method based on Attractive Region in Environment has achieved great performance in high-precision assembly for convex components. However, for nonconvex parts, there may be local minima in the constraint region, and jamming in the assembly process may arise by using the existing method. To solve the problem, this paper propose a compliant assembly method based on Attractive Region in Environment combined with geometric features. First, according to the geometric characteristics of the constraint region, sub-targets during the assembly process are designed. Then, low dimensional attractive regions are utilized to realize the phased directional movement and solve the jamming problem caused by the grooves. Furthermore, Impedance control is applied to guarantee the compliance of assembly. The experimental results of the nonconvex peg-in-hole assembly with clearance of 0.02 mm are presented and show the validity of the proposed method.","PeriodicalId":367307,"journal":{"name":"2021 6th IEEE International Conference on Advanced Robotics and Mechatronics (ICARM)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124977404","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this paper, the development of a novel dual-arm robot is introduced. With the use of modular actuator technology, a dual-arm robot consists of two 7-degree-of-freedom (7-DOF) robotic arms. The robot's high degree of freedom enables it to effectively avoid the joint limitations and singularities of the arm. A solo controller for both arms is proposed and this dual-arm robot has the human sized body and arms. Firstly, a modular actuator is presented and a dual-arm robot is introduced. Then, the kinematic and dynamic analysis for this dual-arm robot is presented. Finally, the prototype of this dual-arm robot is presented and discussed.
{"title":"Development of A Novel Dual-arm Robot via Modular Actuator","authors":"Weijun Wang, Jiangtao Hu, Xiaofeng Yang, Tian Xie, Chaoyang Ma, Wenjie Li","doi":"10.1109/ICARM52023.2021.9536159","DOIUrl":"https://doi.org/10.1109/ICARM52023.2021.9536159","url":null,"abstract":"In this paper, the development of a novel dual-arm robot is introduced. With the use of modular actuator technology, a dual-arm robot consists of two 7-degree-of-freedom (7-DOF) robotic arms. The robot's high degree of freedom enables it to effectively avoid the joint limitations and singularities of the arm. A solo controller for both arms is proposed and this dual-arm robot has the human sized body and arms. Firstly, a modular actuator is presented and a dual-arm robot is introduced. Then, the kinematic and dynamic analysis for this dual-arm robot is presented. Finally, the prototype of this dual-arm robot is presented and discussed.","PeriodicalId":367307,"journal":{"name":"2021 6th IEEE International Conference on Advanced Robotics and Mechatronics (ICARM)","volume":"51 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129763475","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-03DOI: 10.1109/ICARM52023.2021.9536176
Qimin Li, L. Ke, Huayan Pu, Jin Yi, Jie Ma, Ruqing Bai, Jinglei Zhao, Shilong Wang, Jun Luo, Tao Zhu
Vibration affects the function of precision instruments and equipment, causing major structural deformation and damage. Strong vibration and noise cause serious public hazards. Therefore, vibration is considered to be a negative factor. There are many methods for designing vibration controllers, such as machine learning algorithms and artificial intelligence algorithms. In order to reduce the harm of vibration, this paper proposes an improved Youla parameterized adaptive controller to suppress fixed or time-varying sinusoidal disturbances. The algorithm uses fuzzy control to adjust the forgetting factor, so that the system quickly reaches a steady-state and keeps the steady-state error unchanged. The paper illustrates that the improved Youla parameterized adaptive controller can suppress the disturbance. The convergence speed of this algorithm is better than the original Youla parameterized adaptive controller, and the steady-state error is basically unchanged.
{"title":"An Improved Vibration Controller for Precision Manufacture Based on Youla Parametrization and Fuzzy Logic","authors":"Qimin Li, L. Ke, Huayan Pu, Jin Yi, Jie Ma, Ruqing Bai, Jinglei Zhao, Shilong Wang, Jun Luo, Tao Zhu","doi":"10.1109/ICARM52023.2021.9536176","DOIUrl":"https://doi.org/10.1109/ICARM52023.2021.9536176","url":null,"abstract":"Vibration affects the function of precision instruments and equipment, causing major structural deformation and damage. Strong vibration and noise cause serious public hazards. Therefore, vibration is considered to be a negative factor. There are many methods for designing vibration controllers, such as machine learning algorithms and artificial intelligence algorithms. In order to reduce the harm of vibration, this paper proposes an improved Youla parameterized adaptive controller to suppress fixed or time-varying sinusoidal disturbances. The algorithm uses fuzzy control to adjust the forgetting factor, so that the system quickly reaches a steady-state and keeps the steady-state error unchanged. The paper illustrates that the improved Youla parameterized adaptive controller can suppress the disturbance. The convergence speed of this algorithm is better than the original Youla parameterized adaptive controller, and the steady-state error is basically unchanged.","PeriodicalId":367307,"journal":{"name":"2021 6th IEEE International Conference on Advanced Robotics and Mechatronics (ICARM)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129790515","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-03DOI: 10.1109/ICARM52023.2021.9536203
Zhijie Zhou, Wenbin Chen, Hao Fu, Xiang Fang, C. Xiong
Soldiers are often required to carry heavy loads during long distance march. Such load carriage can easily induce joint injuries and foot blisters, which further reduce the wearer’s task performance. To assist human walking with load carriage, various types of robotic devices are proposed, such as powered exoskeletons, supernumerary robotic limbs and suspended backpacks. However, these devices have individual shortcomings. This paper proposes a non-anthropomorphic passive load-carrying exoskeleton, which can dynamically support the carried load during the walking rhythm via a passive legged structure. This exoskeleton can reduce the load borne by human without energy input. The simple and passive structure design brings the highest robustness and flexibility. The simulation based on the mathematic model shows that the exoskeleton can reduce the foot pressure of the users. Such analysis results are also verified by the walking experiment. The experiment results show that the exoskeleton can transfer on average 68.0% of the load to the ground while standing, and 24.6% of the load while walking. The maximum load is reduced by 22.1% during walking.
{"title":"Design and Experimental Evaluation of a Non-anthropomorphic Passive Load-carrying Exoskeleton","authors":"Zhijie Zhou, Wenbin Chen, Hao Fu, Xiang Fang, C. Xiong","doi":"10.1109/ICARM52023.2021.9536203","DOIUrl":"https://doi.org/10.1109/ICARM52023.2021.9536203","url":null,"abstract":"Soldiers are often required to carry heavy loads during long distance march. Such load carriage can easily induce joint injuries and foot blisters, which further reduce the wearer’s task performance. To assist human walking with load carriage, various types of robotic devices are proposed, such as powered exoskeletons, supernumerary robotic limbs and suspended backpacks. However, these devices have individual shortcomings. This paper proposes a non-anthropomorphic passive load-carrying exoskeleton, which can dynamically support the carried load during the walking rhythm via a passive legged structure. This exoskeleton can reduce the load borne by human without energy input. The simple and passive structure design brings the highest robustness and flexibility. The simulation based on the mathematic model shows that the exoskeleton can reduce the foot pressure of the users. Such analysis results are also verified by the walking experiment. The experiment results show that the exoskeleton can transfer on average 68.0% of the load to the ground while standing, and 24.6% of the load while walking. The maximum load is reduced by 22.1% during walking.","PeriodicalId":367307,"journal":{"name":"2021 6th IEEE International Conference on Advanced Robotics and Mechatronics (ICARM)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125284286","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-03DOI: 10.1109/ICARM52023.2021.9536086
Yufeng Li, Xingquan Wang, Yan He, Fei Ren, Yuling Wang
Tool wear and breakage are inevitable due to the severe stress and high temperature in the cutting zone. A highly reliable tool condition monitoring system is vital to maintain the quality of tool and workpiece during machining process. Many studies for tool condition monitoring via monitoring signals based deep learning have been conducted. Each signal has a different sensitivity to a different status of tool wear. It is a key problem that how to combine the advantages of various signals and fuse the sensor signals to improve the accuracy of monitoring. This paper proposes a multiple signals fusing framework(MSFF) for tool condition monitoring based on deep learning. The monitoring signals in machining processes, including force signal, vibration signal, and acoustic emission signal, are collected and analyzed. Then, features related to tool wear on the collected signals are extracted based on deep learning and realize the mapping between the extracted features and tool condition through linear regression. The advantages and the disadvantages of different signal selection schemes based on deep learning are compared and analyzed. The experimental results show that the performance of the proposed MSFF is superior compared to other schemes for tool condition monitoring.
{"title":"A Multiple Signals Fusing Framework for Tool Condition Monitoring Based on Deep Learning","authors":"Yufeng Li, Xingquan Wang, Yan He, Fei Ren, Yuling Wang","doi":"10.1109/ICARM52023.2021.9536086","DOIUrl":"https://doi.org/10.1109/ICARM52023.2021.9536086","url":null,"abstract":"Tool wear and breakage are inevitable due to the severe stress and high temperature in the cutting zone. A highly reliable tool condition monitoring system is vital to maintain the quality of tool and workpiece during machining process. Many studies for tool condition monitoring via monitoring signals based deep learning have been conducted. Each signal has a different sensitivity to a different status of tool wear. It is a key problem that how to combine the advantages of various signals and fuse the sensor signals to improve the accuracy of monitoring. This paper proposes a multiple signals fusing framework(MSFF) for tool condition monitoring based on deep learning. The monitoring signals in machining processes, including force signal, vibration signal, and acoustic emission signal, are collected and analyzed. Then, features related to tool wear on the collected signals are extracted based on deep learning and realize the mapping between the extracted features and tool condition through linear regression. The advantages and the disadvantages of different signal selection schemes based on deep learning are compared and analyzed. The experimental results show that the performance of the proposed MSFF is superior compared to other schemes for tool condition monitoring.","PeriodicalId":367307,"journal":{"name":"2021 6th IEEE International Conference on Advanced Robotics and Mechatronics (ICARM)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128315581","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Event camera conveys dynamic visual information in the format of asynchronous digital events, resulting to the disability of detectors developed for RGB images. Previous methods of event-based object detection mainly rely on simple template matching and encoded maps with deep learning, which sacrifices the spatial sparsity of events and achieves a weak performance in the noisy environment. This paper proposes a miniature event-based spatial attention mechanism of the one-stage detector to reduce the noise of events and to enrich the multi-scale feature maps by merging the shallow features. Maintaining the sparse property of events to the maximum degree, this paper transplants the model from convolution neural network to sparse convolution network and trains it in two ways (by its own and with knowledge distillation). Results show that the lightweight spatial attention mechanism is compatible with one-stage detectors and convolution neural network outperforms sparse convolution network in the event-based object detection.
{"title":"Event-based Object Detection with Lightweight Spatial Attention Mechanism","authors":"Zichen Liang, Guang Chen, Zhijun Li, Peigen Liu, Alois Knoll","doi":"10.1109/ICARM52023.2021.9536146","DOIUrl":"https://doi.org/10.1109/ICARM52023.2021.9536146","url":null,"abstract":"Event camera conveys dynamic visual information in the format of asynchronous digital events, resulting to the disability of detectors developed for RGB images. Previous methods of event-based object detection mainly rely on simple template matching and encoded maps with deep learning, which sacrifices the spatial sparsity of events and achieves a weak performance in the noisy environment. This paper proposes a miniature event-based spatial attention mechanism of the one-stage detector to reduce the noise of events and to enrich the multi-scale feature maps by merging the shallow features. Maintaining the sparse property of events to the maximum degree, this paper transplants the model from convolution neural network to sparse convolution network and trains it in two ways (by its own and with knowledge distillation). Results show that the lightweight spatial attention mechanism is compatible with one-stage detectors and convolution neural network outperforms sparse convolution network in the event-based object detection.","PeriodicalId":367307,"journal":{"name":"2021 6th IEEE International Conference on Advanced Robotics and Mechatronics (ICARM)","volume":"EM-30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126527740","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-03DOI: 10.1109/ICARM52023.2021.9536212
Chongshan Wang, Bin Li, Jiaqi Zhu, Qi Li, Yuan Zhang
The 2-RPU&2-SPS four degree of freedom parallel mechanism is a mechanism with two-by-two symmetrical motion branch chains. Some design parameters of the mechanism can be changed within a certain range. These variable parameters have a certain influence on the design goal. How to choose variable parameter values in order to optimize the design goals. These problems can be solved through parametric design and optimization analysis. This article uses the parametric modeling and optimization analysis functions provided by Adams/View to establish a parametric model of the mechanism by creating four design variables, and the effective value of transmission efficiency index, driving speed stability, total kinetic energy and driving force is four Objective function, introduce weighting factor, transform multiple objective functions into a total objective function for mechanism optimization analysis.
{"title":"Multi-objective Function Optimization of 2-RPU&2-SPS Parallel Mechanism Based on Adams","authors":"Chongshan Wang, Bin Li, Jiaqi Zhu, Qi Li, Yuan Zhang","doi":"10.1109/ICARM52023.2021.9536212","DOIUrl":"https://doi.org/10.1109/ICARM52023.2021.9536212","url":null,"abstract":"The 2-RPU&2-SPS four degree of freedom parallel mechanism is a mechanism with two-by-two symmetrical motion branch chains. Some design parameters of the mechanism can be changed within a certain range. These variable parameters have a certain influence on the design goal. How to choose variable parameter values in order to optimize the design goals. These problems can be solved through parametric design and optimization analysis. This article uses the parametric modeling and optimization analysis functions provided by Adams/View to establish a parametric model of the mechanism by creating four design variables, and the effective value of transmission efficiency index, driving speed stability, total kinetic energy and driving force is four Objective function, introduce weighting factor, transform multiple objective functions into a total objective function for mechanism optimization analysis.","PeriodicalId":367307,"journal":{"name":"2021 6th IEEE International Conference on Advanced Robotics and Mechatronics (ICARM)","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122278416","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-03DOI: 10.1109/ICARM52023.2021.9536118
Yipeng Zhu, Tao Wang, Shiqiang Zhu
Localizing people in 3D space, rather than in original 2D image plane, provides a more comprehensive understanding of the scene and brings up more potential applications. However, inferring 3D locations usually requires stereo camera or additional sensors since deriving depth information from single image is regarded as an ill-posed problem. With recent progress in deep learning methods, depth estimation neural network can provide convincing depth map by a single RGB image. This work develops a people localization and tracking method based on a monocular camera. Specifically, an efficient self-supervised monocular depth estimation method is adopted to generate pseudo depth map. Afterwards, 2D object detection results are adopted for finding accurate people location. Finally, a filter based tracking method is adopted to fuse temporal information and improve the accuracy. Aiming to provide a real time solution for people tracking on embedded system, our methods are deployed and tested on a NVIDIA Jetson Xavier NX develop kit. The proposed efficient localization and tracking method is validated by a group of field tests. The overall performance reaches 12 fps with an acceptable accuracy compared to ground truth.
{"title":"Real-time Monocular 3D People Localization and Tracking on Embedded System","authors":"Yipeng Zhu, Tao Wang, Shiqiang Zhu","doi":"10.1109/ICARM52023.2021.9536118","DOIUrl":"https://doi.org/10.1109/ICARM52023.2021.9536118","url":null,"abstract":"Localizing people in 3D space, rather than in original 2D image plane, provides a more comprehensive understanding of the scene and brings up more potential applications. However, inferring 3D locations usually requires stereo camera or additional sensors since deriving depth information from single image is regarded as an ill-posed problem. With recent progress in deep learning methods, depth estimation neural network can provide convincing depth map by a single RGB image. This work develops a people localization and tracking method based on a monocular camera. Specifically, an efficient self-supervised monocular depth estimation method is adopted to generate pseudo depth map. Afterwards, 2D object detection results are adopted for finding accurate people location. Finally, a filter based tracking method is adopted to fuse temporal information and improve the accuracy. Aiming to provide a real time solution for people tracking on embedded system, our methods are deployed and tested on a NVIDIA Jetson Xavier NX develop kit. The proposed efficient localization and tracking method is validated by a group of field tests. The overall performance reaches 12 fps with an acceptable accuracy compared to ground truth.","PeriodicalId":367307,"journal":{"name":"2021 6th IEEE International Conference on Advanced Robotics and Mechatronics (ICARM)","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125614278","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-03DOI: 10.1109/ICARM52023.2021.9536126
Haitao Wang, Guang Chen, Zhijun Li, Zhengfa Liu
Traffic sign detection based on image and video data is critical, which captures real-time traffic road information for autonomous vehicle. With the rapid development of CNN, more and more CNN-based detectors have promoted general object detection. However, these mainstream detectors still suffer from small object detection task because of small size and fuzzy representation. Traffic signs are representative small object on road scenes causing a rigid challenge for autonomous driving perception system. In this paper, traffic sign detection (TSD) is regard as a small object detection task. We propose a feature fusion method via cross-connection to enhance feature representation. In addition, contextual information searched by dilated convolution is also used to support small traffic sign detection. We have implemented our modules into Faster R-CNN and evaluated effectiveness of proposed method on Tsinghua-Tencent 100K dataset. Our experimental results prove that the feature fusion method via cross connection and contextual information improve detection result of small traffic sign.
{"title":"Traffic Sign Detection using Feature Fusion and Contextual Information","authors":"Haitao Wang, Guang Chen, Zhijun Li, Zhengfa Liu","doi":"10.1109/ICARM52023.2021.9536126","DOIUrl":"https://doi.org/10.1109/ICARM52023.2021.9536126","url":null,"abstract":"Traffic sign detection based on image and video data is critical, which captures real-time traffic road information for autonomous vehicle. With the rapid development of CNN, more and more CNN-based detectors have promoted general object detection. However, these mainstream detectors still suffer from small object detection task because of small size and fuzzy representation. Traffic signs are representative small object on road scenes causing a rigid challenge for autonomous driving perception system. In this paper, traffic sign detection (TSD) is regard as a small object detection task. We propose a feature fusion method via cross-connection to enhance feature representation. In addition, contextual information searched by dilated convolution is also used to support small traffic sign detection. We have implemented our modules into Faster R-CNN and evaluated effectiveness of proposed method on Tsinghua-Tencent 100K dataset. Our experimental results prove that the feature fusion method via cross connection and contextual information improve detection result of small traffic sign.","PeriodicalId":367307,"journal":{"name":"2021 6th IEEE International Conference on Advanced Robotics and Mechatronics (ICARM)","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134054200","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-03DOI: 10.1109/ICARM52023.2021.9536067
Jiahui Yu, Hongwei Gao, Qing Gao, Dalin Zhou, Zhaojie Ju
Compared with RGB-D-based human action analysis, skeleton-based works reach higher robustness and better performance, which are widely applied in the real world. However, the diversity of action observation perspectives hinders the improvement of recognition accuracy. Most of the existing works solve this problem by increasing the amount of training data, which brings a huge computational cost and cannot improve the robustness of the models. This paper proposes an adaptive model to obtain high-performance representations to improve human action recognition accuracy. First, a skeleton representation transfer scheme is proposed to transform the input skeleton-based body model to the best perspective, in which all parameters can be adaptively learned. This is more robust and cost-effective than hand-crafted features. Next, a re-designed backbone is proposed to train the model with a small computational cost based on the 3D-CNN. In the training process, a data enhancement method is also introduced to enhance robustness. Finally, extensive experimental evaluations are conducted on two benchmarks. The results show that this deep model can effectively and adaptively obtain high-performance skeleton representation and its performance is better than other state-of-the-art methods.
{"title":"Skeleton-based Human Activity Analysis Using Deep Neural Networks with Adaptive Representation Transformation","authors":"Jiahui Yu, Hongwei Gao, Qing Gao, Dalin Zhou, Zhaojie Ju","doi":"10.1109/ICARM52023.2021.9536067","DOIUrl":"https://doi.org/10.1109/ICARM52023.2021.9536067","url":null,"abstract":"Compared with RGB-D-based human action analysis, skeleton-based works reach higher robustness and better performance, which are widely applied in the real world. However, the diversity of action observation perspectives hinders the improvement of recognition accuracy. Most of the existing works solve this problem by increasing the amount of training data, which brings a huge computational cost and cannot improve the robustness of the models. This paper proposes an adaptive model to obtain high-performance representations to improve human action recognition accuracy. First, a skeleton representation transfer scheme is proposed to transform the input skeleton-based body model to the best perspective, in which all parameters can be adaptively learned. This is more robust and cost-effective than hand-crafted features. Next, a re-designed backbone is proposed to train the model with a small computational cost based on the 3D-CNN. In the training process, a data enhancement method is also introduced to enhance robustness. Finally, extensive experimental evaluations are conducted on two benchmarks. The results show that this deep model can effectively and adaptively obtain high-performance skeleton representation and its performance is better than other state-of-the-art methods.","PeriodicalId":367307,"journal":{"name":"2021 6th IEEE International Conference on Advanced Robotics and Mechatronics (ICARM)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130753771","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：