Abstract

Info

Author(s):: ZHANG Sihe; CAO Le; WANG Jinwei; XU Haoyang; ZHANG Feng; School of Electronic and Electrical Engineering, Shanghai University of Engineering Science, Shanghai 201620, China

Keywords:: Keywords: bidirectional long short-term memory network self-attention mechanism surface electromyography signal independent component analysis

Abstract:: A muscle strength estimation model incorporating bidirectional long short-term memory network (BiLSTM) and self-attention mechanism (SA) is presented for addressing the problem of low estimation accuracy caused by the continuous change of arms. After collecting the muscle strength and surface electromyography signal during isometric muscle contraction of arms with a self-developed platform, independent component analysis and wavelet threshold denoising are used to preprocess the collected signals. With the root-mean-square value of the extracted signals as the characteristic value, the muscle strength is estimated using BiLSTM-SA model. The experimental results show that BiLSTM-SA model has a high accuracy for muscle strength estimation, with an average value of R2 above 0.97 for the muscle strength estimation during isometric contraction.