中国科学院自动化研究所机构知识库

Rehabilitation training combined with human psychological and physiological information can enhance patients' neural engagement. For this purpose, a racial expression and muscle fatigue based rehabilitation training method is proposed in this paper. Signals from major zygomaticus and corrugator supercilii muscles are used for racial expression recognition, and signals from rectus femoris and biceps femoris muscles are used for fatigue level analysis. Facial expressions (positive, neutral, and negative) are recognized by a classifier which is constructed by wavelet packet features and neural network, and median frequency (MF) is applied to analyze fatigue level. A passive training mode and five-level active training modes are included. Different training modes have different damping levels. When the patient is with positive expression and without fatigue, the damping will be raised automatically in order to increase exercise difficulties and enhance the patient's engagement; when with negative expression and mild fatigue, damping will be decreased properly to reduce exercise difficulties and ease user's burden to obtain more efficient training. Moreover, when patient is severe fatigue, passive training is selected to avoid overfatigue and muscle injury. Feasibility of the proposed method is validated by the experiment conducted on the platform of a damping adjustable treadwheel.