Evaluation of LDA, QDA and decision trees for multifunctional controlled below elbow prosthetic limb using EMG signals

The concept of controlling prosthesis through an EMG signal is challenging. This research is based on acquisition of EMG data using surface electrodes. The goal was to design a classifier of higher efficiency for below elbow muscles. The acquired signals are processed in MATLAB for reduction of noise and amplitude amplification through signal processing techniques. The conditioned signal is passed through a number of statistical classifiers such as LDA, QDA, and Decision Tree. The decision tree classifier was found to have more percentage accuracy (77.22%) of true prediction of class as compared to QDA and LDA classifiers in hand open, hand close, wrist rotate and wrist tilt movements.