Neuromuscular disease detection based on feature extraction from time–frequency images of EMG signals employing robust hyperbolic Stockwell transform

In this paper, a novel technique for detection of healthy (H), myopathy, (M) and amyotrophic lateral sclerosis (ALS) electromyography (EMG) signals is proposed employing robust hyperbolic Stockwell transform (HST). HST is an efficient signal processing technique to analyze any nonstationary signal in joint time–frequency (T–F) plane. However, a major issue with HST is the optimum selection of Gaussian window parameters since the resolution in the T–F plane depends on the shape of the window. Considering the aforesaid fact, in this article, a genetic algorithm (GA) based optimized HST is proposed for improved EMG signal analysis in T–F plane. Several novel features were extracted from HST spectrum and features with high statistical significance were selected for classification using several benchmark classifiers. It was observed that optimized HST resulted in better classification accuracy of EMG signals which indicates its potential for clinical applications.