Paper Title
A Comparative Study on Finger-Movement Classification Based on Surface-Electromyography-Peak Parameters

Surface electromyography (SEMG)signals have been used as input to myoelectric controlled robotic prostheses and orthoses. This project was aimed at conducting a comparative study on finger-movement classification using peak feature of SEMG signals.Each fingers’ movement was classified by recording the signals through nine surface electrodes in a four-channel differential setup and one grounding electrode. The signals were filtered through a Notch filter to remove the interfering ground noise and a Butterworth high-pass filter to remove artefacts, and then the signals were smoothed by moving average (MA) filters or root mean square (RMS) filters.We compared the overall success rates of five classification algorithms including the support vector machine, decision tree, ensemble, linear discriminant, and quadratic discriminant algorithms provided by MATLAB Classification Learner app. We found that ensemble algorithm achieves the highest overall success rate of 94.19% and the moving average filter with a window size 10 provides the best smoothing signals for subsequent classification with an overall success rate of 96.78%. This worklays the foundation for SEMG signals to be used for controllinga five-fingered robotic exoskeleton orthosis. Index Terms- Surface electromyography, finger movement, classification, robotic orthoses index finger, and other fingers