A Novel Postprocessing Method for Robust Myoelectric Pattern-Recognition Control Through Movement Pattern Transition Detection

Bin Yu,Xu Zhang,Le Wu,Xiang Chen,Xun Chen

DOI: https://doi.org/10.1109/THMS.2019.2953262

2020-01-01

IEEE Transactions on Human-Machine Systems

Abstract:Pattern-recognition-based myoelectric control systems are not yet widely available due to their limited robustness in real-life situations. Some postprocessing methods were introduced to improve the robustness in previous studies, but there is lack of investigation into movement transition phases. This article presents a novel postprocessing method based on movement pattern transition (MPT) detection. An image-based index is used to quantify the similarity of adjacent feature matrices from high-density surface electromyogram (EMG) signals. MPT detection is implemented by applying a double threshold to the calculated index. The proposed postprocessing method is used to rectify the EMG pattern recognition decisions from the classifier by incorporating the detected information. Two representative testing schemes are used to verify the robustness of the proposed method against force level variation and consecutive nonstop task performance. The proposed method achieved mean classification accuracy improvements of 7.33 and 10.91 with respect to the baseline performance of a raw classifier (without any postprocessing) in the two testing schemes. It also outperformed other common postprocessing methods (p < 0.05). Considering both the accuracy improvement and time efficiency for rapid responses to MPT, the proposed method could be a potential option for postprocessing to enhance the robustness of myoelectric control.