SEMG Based Recognition of Hand Motions for Lower Limb Prostheses

Aim: On multiple muscle locations, Surface Electromyography (sEMG) signals were recorded to predict the effect of different hand movements. Background: Myoelectric information is a non-stationary signal, so extracting correct features is important to boost any myoelectric control devices' performance. Myoelectric signal is an electrical activity recorded by a surface electrode at various movements of the muscles. Objective: The study presented pattern recognition classification methods to select an excellent algorithm for controlling the SEMG signal. Methods: Various time domain and frequency domain parameters were extracted prior to conducting the classifier test. Results: For the evaluation of the results for the recorded data (of all six movements), confusion matrix for neural network, Support Vector Machine (SVM), DT, and Linear Discriminant Analysis (LDA) classifiers is presented. Conclusion: This present study will be a step ahead in analyzing different problems for developing lower limb prosthesis.