Analysis of Recorded Surface Electromyography Signals Under Varied Muscle Fiber Proportions Using Fractal Dimension

Surface electromyography (sEMG) signals are highly comp- lex and nonlinear in nature. Fractal dimension (FD) is used to study the complex properties of these signals. The objective of the study is to identify the variation of FD based on varied muscle fiber-type proportions using sEMG signals. Using standard protocol, signals are recorded from gastrocnemius lateralis (GL) and soleus (SOL) muscles. The initial 1-s segment is considered nonfatigue (NF) and the final 1-s as fatigue (FT) segment. Methods, such as box counting (BC), revised Katz's (RK), and detrended fluctuation analysis (DFA), are used to compute the FD from the NF and FT segments and their variation during FT is analyzed. From the results, it is observed that FD decreases for GL and increases for SOL during the transition from NF to FT in BC and RK methods. The distinct muscle fiber-type proportions in GL and SOL might have contributed to the result. However, the DFA shows a different trend with an increase in both muscles. Popular machine learning techniques, namely, LibLINEAR, SimpleLogistic, and ripple-down rule learner (Ridor), are used to classify between the FT states NF and FT of both GL and SOL muscles, with combined FD of BC, RK, and DFA. Ridor exhibits the highest classification performance, while LibLINEAR exhibits the lowest. From the results, it is observed that FD is related to fiber-type proportion. The proposed study can be used for analyzing the hidden dynamics of neuromuscular system.