Soft Exoskeleton Glove for Hand Assistance Based on Human-machine Interaction and Machine Learning

This paper proposes a human machine interaction system in the field of stroke rehabilitation, based on the concept of mirror therapy (MT). It aims to improve the hand motor function of stroke patients, enabling a true synchronization between the affected hand and non-affected hand (healthy hand) for the stroke patient. It consists of a soft exoskeleton glove, a surface electromyography (sEMG) signal collecting armband and machine learning (ML) algorithms. The glove is developed by integrating low-power motors to provide force strength for the hand movement. Unlike the rigid exoskeleton devices, the glove is comfortable to wear and lightweight, so it is more suitable for rehabilitation training of stroke patients in daily life. The armband collects the sEMG signals for pattern recognition by the ML algorithms. In the experiment, four subjects perform 10 hand gestures to collect data for model training. A comparison of data preprocessing is conducted to find the optimal data segmentation method and feature vector sets. A series of pattern recognition algorithms are developed and assessed in different aspects, including prediction accuracy, training time and predicting time. All 10 gestures can be recognized in offline mode with an accuracy up to 99.4%. The control of soft exoskeleton glove in real-time manner is also carried out, and the accuracy is 82.2%. The experiment result demonstrates the feasibility of the proposed system. The innovations and limitations of the work are discussed at the end of the paper.