A Study on Finger Grasping Action Classification Based on Multimodal Fusion

The gradual increase in the number of stroke patients is due to the gradual aging of our population and unhealthy lifestyles. Due to limited medical resources, many patients cannot receive rehabilitation training in a timely and effective manner. This paper proposes to recognize hand movements based on myoelectric sensors and curvature sensors, utilizing radial basis function networks and convolutional neural to classify finger movements with good results. The application of multi-sensor fusion improves human-machine interaction, enhances the ability to perceive the environment, better realizes human-centered machine control, and improves the assistive ability of rehabilitation robots.