EMG based neural network and admittance control of an active wrist orthosis

This study proposes an Electromyography (EMG) based neural network and admittance control strategy for an Active wrist orthosis (AWO) that is mobile, powerful and lightweight in order to avoid the occurrence and/or for the treatment of repetitive strain injuries. The device has an EMG based control strategy so that it can track the contraction of the forearm muscles of interest in real time to assist the device user during the extension and flexion wrist movements. Indeed, time-delayed artificial neural networks were trained offline to predict the forearm muscle forces based on features extracted from raw EMG signals. The predicted force values were used to calculate a reference velocity command by using an admittance model with properly adjusted parameters. A closed loop velocity control system was used to perform the desired wrist motion. Some experimental studies were performed to evaluate the assistive performance of the AWO device with the proposed control system under various disturbance loads. The experimental results show that the activation levels of the forearm muscles were considerably reduced when the AWO device was enabled.