A Domain Adaptive Convolutional Neural Network for sEMG-based Gait Phase Recognition Against to Speed Changes

• Published in: IEEE sensors journal Vol. 23; no. 3; p. 1
• Main Authors: Ling, Zi-Qin, Chen, Jiang-Cheng, Cao, Guang-Zhong, Zhang, Yue-Peng, Li, Ling-Long, Xu, Wen-Xin, Cao, Sheng-Bin
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.02.2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Adaptation models; Algorithms; Artificial neural networks; Data models; deep learning; domain adaptation; Domains; Electromyography; Force; Gait; Gait phase recognition; Low-pass filters; Machine learning; Matrix converters; Muscles; Neural networks; Pattern recognition; Pattern recognition systems; Rehabilitation robots; Sensors; Speed changes; surface electromyography
• ISSN: 1530-437X; 1558-1748
• DOI: 10.1109/JSEN.2022.3228320

Gait phase recognition based on surface electromyography (sEMG) signals provides a human-robot interface for rehabilitation robots. However, a high-performance deep learning model requires to be calibrated on a large amount of data and frequently recalibrated for new gait speeds, which raises a burden on users. In this paper, a domain adaptive convolutional neural network (DACNN) model for sEMG-based gait phase recognition against speed changes is proposed, which only needs to be pre-trained on a comfortable gait speed and quick recalibrated on new gait speeds. Specifically, a CNN-based backbone model (BM) is first constructed and pre-trained using comfortable gait speed data. Then, a Domain Adaptation (DA) algorithm is applied for recalibration, which allows the BM to learn speed-invariant representation with a small amount of new speed data. Four subjects participated in the experiment and were required to walk at four gait speeds, three BMs based on AlexNet, LeNet and FDCNN were set as architectural options. Compared with the fine-tuning (FT) and non-recalibration (NoRC) strategies, the results show that the proposed method significantly outperforms both baseline methods. By recalibrating on 6s data, the proposed method can achieve an accuracy improvement of 1.5% to 6.1% compared to FT, and 58% to 81% compared to NoRC. The average gait phase recognition accuracies are 58.13% for AlexNet, 81.56% for LeNet, and 81.53% for FDCNN, respectively. The average mean absolute errors for gait event identification are 48ms, 85ms, and 66ms, respectively. These results indicate that DACNN with low recalibration burden can improve the usability of sEMG-based pattern recognition systems.