SA-SVM-Based Locomotion Pattern Recognition for Exoskeleton Robot

• DOI: 10.3390/app11125573
• Publication Date: 2021-06-16T14:27:07Z
• Authors: Zeyu Yin; Jianbin Zheng; Liping Huang; Yifan Gao; Huihui Peng; Linghan Yin
• Keywords: Torso; Powered exoskeleton

An exoskeleton robot is a kind of wearable mechanical instrument designed according to the shape and function human body. The main purpose its design manufacture enhance strength, assist walking help patients recover. state should be highly consistent with human, so accurate locomotion pattern recognition premise flexible control robot. In this paper, simulated annealing (SA) algorithm-based support vector machine model proposed for different patterns. order improve overall performance (SVM), algorithm adopted obtain optimal parameters machine. pressure signal measured by force sensing resistors integrated on sole shoe fused position pose information inertial measurement units attached thigh, shank foot, which are used as input max-relevance min-redundancy was selected feature extraction based window size 300 ms sampling frequency 100 Hz. Since signals come from types sensors, normalization required scale interval (0,1). prevent classifier overfitting, five layers cross validation train classifier. obtained offline in MATLAB. finite limit transition accuracy. Experiments patterns show that accuracy 97.47% ± 1.16%. SA-SVM method can extended industrial robots rehabilitation robots.