Development of surface electromyography-based motion intention recognition for human-machine interface

Exoskeletons are increasingly used to assist elderly and disabled individuals in daily activities. The interaction process between robots and the human body needs to be facilitated through a human-robot interface to assist in daily human activities, such as sit-to-stand movements. The performance of the human-robot interface largely determines the overall performance of the robot. The human intention detection is a key issue for human-machine interface, while surface electromyography (sEMG) is a promising method. In this paper, we developed a sEMG-related motion intention detection in human-machine interface design for sit-standing assistance. We collected twelve channels of sEMG data from 22 elderly individuals during sit-to-stand and walking to demonstrate the usefulness of sEMG in motion intention detection. The data set of sit-to-stand is defined as positive intention, while dataset from walking and resting state is defined as negative. ConvNet deep learning model was utilized as the classifier for sit-standing intension. The results showed that the classifier achieved a accuracy of 99%, with both precision and recall higher than 98%. Our study provides a method of sEMG-related motion intention detection interface for low limb exoskeleton robot control. Our study provides a method of utilizing surface electromyography (sEMG) for motion intention detection, specifically tailored for controlling low limb exoskeleton robots. This approach holds promise for enhancing assistance in daily activities for elderly and disabled individuals.