Active Movements Intention Recognition for Upper Limb Rehabilitation Robots Based on EMG Signals

The recognition of upper limb movements based on Electromyographic(EMG) signals is an important method for the implementation of active assisted training modes for rehabilitation robots. In this paper, EMG signals are used for the recognition of six common upper limb movements. Specifically, the acquired EMG signals are first processed by noise reduction, active segment detection, and windowing. Then, fusing deep learning algorithms and attention mechanisms, two deep learning networks are designed to recognise the pre-processed segmented time-domain signals and spectrograms respectively. The experimental results show that the accuracy of upper limb movement recognition based on segmented EMG signals is 93.7%, which is better than the recognition accuracy based on spectrograms and meets the requirements of active-assisted training mode for upper limb rehabilitation robots.