An EMG-based muscle force evaluation method using approximate entropy

This paper proposed a novel muscle force evaluation method to evaluate the patient's rehabilitation condition in the process of rehabilitation training. According to the related literature research, the complexity of the electromyography (EMG) signals were different under different muscle force. The physiological features and the state of muscle can be indirectly speculated by detecting the change of the dynamic complexity of EMG signals. The novel ideal of our research is to propose an EMG-based muscle force evaluation method using approximate entropy. The evaluation system consists of two main parts: the EMG acquisition (BIOFORCEN) and the measurement of muscle force (FingerTPS). Experiments were conducted with a healthy male. 15 groups EMG signals of biceps and triceps were acquired under different muscle force. Raw EMG data were recorded for off-line analysis. The approximate entropy (ApEn) and the power of EMG signals aiming at intercept relatively stable 10 section signal in each group were calculated to compose the training set. In addition, the additional 150 groups feature vectors were obtained to compose a sample set. 15 groups muscle force were divided into 6 levels and two classification method (linear discriminate analysis (LDA) and quadratic discriminate analysis (QDA)) were used to classify the feature vector. Experimental results have shown that the discrimination between ApEn and the power were obvious and 65% classification accuracy was got with the QDA method. The research of this paper can be a promising approach for further research in the field of rehabilitation evaluation.