Impact of Transient EMG Signals on the Performance of a Gesture Recognition System

An Electromyography (EMG) based gesture recognition system takes myoelectric signals as input and uses these signals to predict hand movements. To predict the hand movements, three basic procedures are involved: signal processing, feature extraction, and classification. An EMG signal is combination of transient and static part observed during onset and hold state in a typical movement cycle. In this work, we investigated the impact of transient signals detected at the movement onset on the performance of a gesture-controlled system. We used transient, static, and complete signals to calculate the accuracy using LDA and kNN classifiers to evaluate the performance of the proposed study in different cases. The results indicate maximum accuracy with only static EMG signals and minimum accuracy while selecting only transient signals. The average accuracy for the LDA classifier with static signals is 83.29 %, and with transient signals, it is 61.53 %, signifying the impact of static signals in identifying movements.