Single channel EMG classification using DWT and SVM

In order to develop a prototype of upper limb prosthetic, we present in this paper our contribution to the design of an intelligent classification system for the arm's flexion and extension. The first step, we designed a simple and efficient single channel of electro myogram signal (EMG) acquisition circuit in order to create two databases that contains EMG signals matrices of both flexion and extension of the arm. Our work proves that only one statistical feature, the energy of detail coefficients for the first four decomposition levels, is sufficient to represent these databases. We applied the principal component analysis PCA to reduce the data space and keep the most relevant ones. In order to detect flexion or extension movement, classification by Support Vector Machines (SVM) has made possible for us to achieve recognition rate of 100% using a wise choice of discret wavelet transform (DWT).