ABSTRACT
Electromyogram (EMG) is a biomedical galvanic signal recorded from the activity of muscles, which aids in identifying the Neuromuscular Impairment (NMI). Extraction of features, feature selection (FS) and building efficient classifier contributes to discern the abnormalities that exist in the EMG signals. This work exploits Wigner-Ville transformation technique to extract the time-frequency domain features from typical and atypical EMG signals –myopathy (muscle disorder) and amyotrophic lateral sclerosis (neuro-disorder). Nature Inspired Feature Selection (NIFS) algorithms, Whale Optimization Algorithm (WOA), Genetic Algorithm (GA), Bat Algorithm (BA), Firefly optimization Algorithm (FA) and Particle Swarm Optimization (PSO) are utilized to determine the relevant features from the constructed features. The approach of FS focuses on dimensionality reduction of the data retrieved from EMG signals, which significantly reduces the computational time. Further, the reduced subset of features using the nature inspired algorithms are used to build a classifier using the Extreme Learning Machine (ELM) classification model. The performance of the ELM classifier using NIFS techniques is analyzed and evaluated. This work plays a notable importance in building a classifier for assisting clinical investigation in diagnosing the NMI.
