Nonlinear analysis of biceps surface EMG signals for chaotic approaches

Abstract

In the present study, the chaos is detected and quantified in the electromyographic (EMG) signal of the biceps muscle using nonlinear features. Biological systems are complex and nonlinear systems composed of a large number of interconnected elements. A thorough understanding of the behavior of these systems, which are sometimes chaotic, will lead to an understanding of the behavior of diseases, finding effective treatments, and using appropriate rehabilitation equipment. Therefore, designing a novel stimulation equation based on Rossler nonlinear dynamics causes chaos in the biceps EMG signal. This stimulation, the sitting and hand positioning is designed in such a way to prevent muscle fatigue and saturated stimulability and also to ensure that with minimal stimulation of the musculocutaneous nerve, there is a maximum activity in the biceps muscle. Then, each signal was divided into chaotic and non-chaotic parts with the positive largest Lyapunov exponent (LLE), and its nonlinear and chaotic criteria, including fractal dimension (FD) using Petrosian method, correlation dimension (CD), for both parts of the EMG signal were calculated as first time. Then, after applying the values of these criteria as features to the support vector machine (SVM) classifier, the separation was carried out using a linear kernel. The results showed that if the classification features include the two features of FD and CD, they are separated from non-chaotic EMG signals with 98.97 % accuracy. Finally, for the first time the extracted features of CD, FD are a good representation of the chaotic behavior of the biceps EMG signal.