Article
Engineering, Biomedical
Jena L. Nawfel, Kevin B. Englehart, Erik J. Scheme
Summary: Research indicates that the standard offline metric, classification accuracy, is not a good indicator of usability and other metrics are needed for prediction. Combining offline metrics leads to more accurate predictions, with feature efficiency being the best indicator for predicting usability metric throughput.
IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING
(2021)
Article
Engineering, Biomedical
Zhicheng Teng, Guanghua Xu, Renghao Liang, Min Li, Sicong Zhang
Summary: This study proposed a force-invariant intent recognition method based on muscle synergy analysis, which was found to significantly outperform traditional pattern recognition methods in accuracy under different force levels. Although there was a drop in classification accuracy among amputee subjects, the proposed method still holds potential for clinical applications.
IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING
(2021)
Article
Engineering, Biomedical
Gang Seo, Ameen Kishta, Emily Mugler, Marc W. Slutzky, Jinsook Roh
Summary: This study found that MyoCI training can reduce arm impairment in stroke survivors by decoupling the trained muscles while leaving other muscles relatively unaffected.
JOURNAL OF NEUROENGINEERING AND REHABILITATION
(2022)
Article
Physiology
Kunkun Zhao, Chuan He, Wentao Xiang, Yuxuan Zhou, Zhisheng Zhang, Jianqing Li, Alessandro Scano
Summary: This study investigated the alterations of muscle synergies in post-stroke patients during upper-limb movements. It found that stroke led to changes in synergy structure and increased variability. Four synergy coordination patterns were identified, providing insights into the neurophysiological mechanisms and motor control strategies in post-stroke patients.
FRONTIERS IN PHYSIOLOGY
(2023)
Article
Engineering, Biomedical
Philip P. Vu, Alex K. Vaskov, Christina Lee, Ritvik R. Jillala, Dylan M. Wallace, Alicia J. Davis, Theodore A. Kung, Stephen W. P. Kemp, Deanna H. Gates, Cynthia A. Chestek, Paul S. Cederna
Summary: By implanting electrodes in RPNI and muscles, stable and high amplitude signals can be obtained for long-term prosthetic control. The signal quality remained consistent for up to 276 and 1054 days in P1 and P2 respectively. P2 maintained high accuracy for real-time prosthetic control for 604 days and performed a real-world multi-sequence coffee task with 99% accuracy for 611 days.
JOURNAL OF NEURAL ENGINEERING
(2023)
Article
Engineering, Biomedical
Ruochen Hu, Xiang Chen, Haotian Zhang, Xu Zhang, Xun Chen
Summary: This study proposes a novel myoelectric control scheme that supports gesture recognition and muscle force estimation. The scheme utilizes multi-task learning technique to achieve synchronous prediction of gesture category and instantaneous force. A post-processing algorithm based on threshold method is used to improve the accuracy of gesture recognition. Experimental results show that the proposed scheme reduces the overall gesture classification error and meets the real-time requirement of EMG control system.
IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING
(2022)
Article
Engineering, Biomedical
Xinhui Li, Xu Zhang, Liwei Zhang, Xiang Chen, Ping Zhou
Summary: To improve the accuracy of myoelectric pattern recognition and muscle force estimation, a novel method using transformer-based multi-task learning was proposed. The method achieved high classification accuracy and low estimation error by capturing the inherent characteristics of gesture patterns and long-term temporal correlations. The proposed method outperformed other temporally modeling methods in terms of both gesture recognition accuracies and muscle force estimation errors.
IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING
(2023)
Article
Neurosciences
Xiaojun Wang, Junlin Wang, Ningbo Fei, Dehao Duanmu, Beibei Feng, Xiaodong Li, Wing-Yuk Ip, Yong Hu
Summary: This study observed the long-term effect of motor skill learning on amputees who developed unique muscle activation patterns to control their prostheses. The results showed that with proper feedback training, amputees could learn distinct muscle activation patterns and improve myoelectric prosthetic control performance. The effect of motor skill learning has a lasting impact on sEMG pattern classification accuracy.
COGNITIVE NEURODYNAMICS
(2023)
Article
Neurosciences
Andrea Sarasola-Sanz, Eduardo Lopez-Larraz, Nerea Irastorza-Landa, Giulia Rossi, Thiago Figueiredo, Joseph McIntyre, Ander Ramos-Murguialday
Summary: This study aimed to develop a myoelectric interface for multi-degree-of-freedom control of an exoskeleton involving upper limb joints, with the goal of rehabilitation. The results demonstrated successful simultaneous control of multiple upper-limb joints by all participants, indicating the effectiveness of the mirror myoelectric interface.
FRONTIERS IN NEUROSCIENCE
(2022)
Article
Engineering, Biomedical
Miguel Gonzalez, Hao Su, Qiushi Fu
Summary: Research on EMG-based control of prostheses has mainly focused on adult subjects. This study aimed to investigate the ability of children to generate consistent EMG signals for controlling artificial limbs. The experiments demonstrated that children have fewer independent movements and lower performance compared to adults. The study also found that the performance of children was age-dependent.
IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING
(2022)
Article
Neurosciences
Mojisola Grace Asogbon, Oluwarotimi Williams Samuel, Ejay Nsugbe, Yongcheng Li, Frank Kulwa, Deogratias Mzurikwao, Shixiong Chen, Guanglin Li
Summary: This article investigates the impact of myoelectric signal recording duration on finger gesture recognition and finds that a recording duration of 5 to 10 seconds can achieve good decoding accuracy. It also provides guidance for selecting appropriate recording durations.
FRONTIERS IN NEUROSCIENCE
(2023)
Article
Engineering, Electrical & Electronic
Le Wu, Aiping Liu, Xu Zhang, Xiang Chen, Xun Chen
Summary: Under ideal conditions, traditional myoelectric pattern recognition systems can perform well in recognizing motion intents. However, in practical applications, electrode shift can lead to significant performance degradation. Deep learning, particularly convolutional neural networks (CNNs), has shown effective solutions to this issue. In this article, a shift-robust CNN (SR-CNN) is proposed, which replaces the downsampling layer with an anti-aliasing filter and adaptive polyphase sampling (APS) module, to mitigate the aliasing effect caused by electrode shift. Experimental results on two high-density surface electromyography (HD-sEMG) datasets demonstrate that the proposed SR-CNN outperforms existing baselines. This study presents a promising solution for robust myoelectric control against electrode shift.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Computer Science, Information Systems
Md. Johirul Islam, Shamim Ahmad, Fahmida Haque, Mamun Bin Ibne Reaz, Mohammad A. S. Bhuiyan, Md. Rezaul Islam
Summary: The authors proposed a scheme to normalize EMG signals across channels before feature extraction, significantly enhancing force invariant EMG-PR performance compared to previous studies. The proposed method achieved the highest F1 scores when using different classifiers, suggesting its potential for practical application.
Article
Engineering, Biomedical
Xiangxin Li, Lan Tian, Yue Zheng, Oluwarotimi Williams Samuel, Peng Fang, Lin Wang, Guanglin Li
Summary: Surface electromyogram pattern recognition (EMG-PR) is a promising approach for predicting amputees' motion intentions to control myoelectric prostheses. In this study, a feature filtering strategy was proposed and applied to improve the performance of EMG-PR. Experimental results showed that the proposed strategy significantly increased motion classification accuracy and real-time motion completion rate.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2021)
Article
Chemistry, Analytical
Junjun Fan, Jiajun Wen, Zhihui Lai
Summary: In the field of muscle-computer interface, a two-stage architecture called GAF-CNN is proposed to improve the performance of myoelectric pattern recognition. It utilizes Gramian angular field-based 2D representation and convolutional neural network-based classification to extract patterns from complex sEMG signals. Experimental results show that the proposed GAF-CNN method is comparable to the state-of-the-art methods incorporating CNN models.
