Feature extraction and pattern recognition of EMG-based signal for hand movements

EMG pattern recognition has been developed to interpret the performance of different functional movements. It can be used to develop the movement control techniques of assistive devices for people who are physically disabled. Suitable features in time domain are extracted from three different hand movements. The EMG signal was recorded from Cubitus (Elbow) bending, Carpus (Wrist) twist, Brachium (Arm) twist and Palm contraction of forty eight healthy subjects (male and female) by two pairs of Ag-AgCl surface electrodes on the right and left antebrachium. The performance of the classifier indicates EMG-based recognition accuracy for similar movement of right vs. left is found to be more than 95% and for different movement of around 90%. Also, the classification accuracy for average data set is achieved around 93-97% when compared for left and right hand.