Multi-finger myoelectric signals for controlling a virtual robotic prosthetic hand

Abstract

The paper investigates how multi-finger myoelectric signals could be used to control a virtual robotic prosthetic hand created by using robot operating system (ROS). Both off-line and online experiment phases are conducted by using ten electrodes and performing eight selected multi-finger motions on four healthy subjects. Classification accuracy and confusion matrix of eight time domain (TD)-features and two algorithms are compared during the off-line phase. Then the delay time and accuracy of online control of six selected TD-features and support vector machine (SVM) algorithm are presented with and without visual feedback from the virtual robotic prosthetic hand system. The experimental results show that different feature extraction principles have significant influence on online experiment performance when using SVM without visual feedback (SVMO), and the SVM with visual feedback (SVMW) has improved the online classification and recognition accuracy of eight multi-finger motions through all selected TD features.