Active Rehabilitation Gloves Based on Brain-Computer Interfaces and Deep Learning

Abstract

Cerebral stroke is the second leading cause of death and the third leading cause of death and disability in the world, and more than half of these patients have hand dysfunction, making hand rehabilitation an urgent challenge. In this study, a system for hand rehabilitation therapy for stroke patients was designed using novel human-computer interaction technology. The system combines a brain-computer interface, a deep learning algorithm and a rehabilitation glove, and designs an electroencephalogram (EEG) signal acquisition card and a rehabilitation glove to realise the application of motor imagery therapy to the active rehabilitation of patients' hands. On the brain-computer interface-based motor imagery experiments, the Long Short Term Memory (LSTM) recurrent neural network algorithm designed in this study achieves an average accuracy of 95.78% for the classification accuracy of mental tasks in seven motor imagery modes, which is important for the active rehabilitation of patients with hand function based on motor imagery-driven rehabilitation.