EMG signal recognition based on EMD sample entropy

Considering the nonstationary and nonlinear of surface electromyography(sEMG),a new method for sEMG signal recognition based on wavelet packet entropy and support vector machines was proposed.The uncertainty and complexity of sEMG were analyzed using wavelet packet entropy.According to the sEMG wavelet packet transform,extracted the wavelet packet coefficient matrix and calculate wavelet packet entropy.A 3-dimension eigenvector which was constructed with three-sEMG signal wavelet packet entropy was inputted Support vector machines(SVM) to classify the hand actions.Experimental results show that six movements(wrist spreads,wrist bends,hand extension,hand grasps,wrist external rotation,wrist internal rotation) are successfully identified by the method of SVM combined with the wavelet packet entropy.Compared with neural network sorter,the more precise classified results can be get.