Machine-Learning based Wearable Multi-Channel sEMG Biometrics Modality for User's Identification

Recently, wearable technologies have several bio-engineering applications. In this paper, a Multi-channel surface electromyography (sEMG) wearable armband has been used for an access control system in biometrics applications. A set of experiments have been conducted to explore the ability of sEMG signal to be used for user's identification system. Features are extracted from EMG signals in both frequency and time domains. Three classifiers have been used, namely: K-nearest Neighbors (KNN), Linear Discernment Analysis (LDA), and Ensemble of Classifiers. Results show that the KNN classifier allows performance of 86.01 % in the user's identification system.