sEMG-Based Gesture Classifier for a Rehabilitation Glove

Human hand gesture recognition from surface electromyography (sEMG) signals is one of the main paradigms for prosthetic and rehabilitation device control. The accuracy of gesture recognition is correlated with the control mechanism. In this work, a new classifier based on the Bayesian neural network, pattern recognition networks, and layer recurrent network is presented. The online results obtained with this architecture represent a promising solution for hand gesture recognition (98.7% accuracy) in sEMG signal classification. For real time classification performance with rehabilitation devices, a new simple and efficient interface is developed in which users can re-train the classification algorithm with their own sEMG gesture data in a few minutes while enables shape memory alloy-based rehabilitation device connection and control. The position of reference for the rehabilitation device is generated by the algorithm based on the classifier, which is capable of detecting user movement intention in real time. The main aim of this study is to prove that the device control algorithm is adapted to the characteristics and necessities of the user through the proposed classifier with high accuracy in hand gesture recognition.