Classification of EMG signals through wavelet neural network for Finger-Robot Interface

Wavelet Neural Network (WNN) with Particle Swarm Optimization (PSO) is presented in this study as a classification method for surface electromyogram (sEMG) pattern classification. A change in the spectrum of surface electromyogram has largely been attributed to the change in muscle conduction velocity. We have investigated this theory by using these changes to run robot. In the experiment, the subjects were instructed by an auditory cue to elicit a contraction from rest and hold that finger posture for a period of 5 s.For this purpose, two EMG electrodes located on the human forearm are utilized to collect the EMG data. Time and frequency sets such as Number of Zero Crossings (ZC), Autoregressive (AR) and wavelet coefficients are considered as features. Finally, WNN is used as a classification method and its performance is improved through utilizing optimization algorithm. Accuracy of WNN with PSO is compared to Artificial Neural Network (ANN) and results show accuracy of proposed approach is ≈%09 and it is better than ANN. Finally outputs of best classification method are implemented on a robot named Tabriz-Puma.