文献类型：期刊文献

英文题名：Comparison of Swarm Intelligence Algorithms Optimized Classification Algorithms For Mechanomyography-Based Hand Movements Recognition

作者：Zhang, Yue[1]; Cao, Gangsheng[2]; Sun, Maoxun[3]; Xia, Chunming[2,4]

机构：[1] School of Mechanical Engineering, Nantong University, Nantong, 226019, China; [2] School of Mechanical and Power Engineering, East China University of Science and Technology, Shanghai, 200237, China; [3] School of Mechanical Engineering, University of Shanghai for Science and Technology, Shanghai, 200093, China; [4] School of Mechanical and Automotive Engineering, Shanghai University of Engineering Science, Shanghai, 201620, China

年份：2023

外文期刊名：SSRN

收录：EI(收录号：20230070687)

语种：英文

外文关键词：Classification (of information) - Exoskeleton (Robotics) - Finite difference method - Frequency domain analysis - Knowledge acquisition - Learning algorithms - Learning systems - Palmprint recognition - Particle swarm optimization (PSO) - Principal component analysis - Swarm intelligence - Time domain analysis - Virtual reality

摘要：On the basis of extracting mechanomyography (MMG) features, classification of hand movements has certain application values in human machine interaction systems and wearable devices. Pattern recognition of hand movements based on MMG can be applied in controlling exoskeletons, prosthetics and virtual reality system. In this study, when seven subjects performed hand movements, eight channels of MMG were collected in three condition methods to acquire 21 data sets. Ten time domain (TD) features, eleven wavelet packet node energy (WPNE) features, and two frequency domain (FD) features were extracted from each channel to constitute six feature sets. Each feature sets were reduce the dimension by principal component analysis (PCA) and ReliefF selection algorithm. Four swarm intelligence algorithms, i.e. particle swarm optimization (PSO), sparrow search algorithm (SSA), grey wolf optimization (GWO) and bald eagle search (BES), were utilized to optimize extreme learning machine (ELM) and support vector machine (SVM), then models are trained to the recognition of hand movements from each feature set. For ELM and SVM, using PSO as the optimization algorithm, time consumption is less than using the three other swarm algorithms. For SVM, using BES as the optimization algorithm can generate larger descend of fitness than others, while using PSO lead to the smallest descend. Though BES optimized SVM and ELM consume longer training time than other swarm algorithms, adopting BES-SVM can obtain the classification accuracy up to 97.33% after reducing the dimension of TD+FD hybrid feature set by ReliefF. ? 2023, The Authors. All rights reserved.