Detecting Neuromuscular Disorders Using EMG Signals Based on TQWT Features

Abstract

Neuromuscular disorders are characterized by abnormal functioning of muscles and nerves that communicate with the brain, resulting in muscle weakness and ultimately damage to nervous control, for instance amyotrophic lateral sclerosis (ALS) and myopathy (MYO). Diagnosis of these disorders is frequently done by examining ALS, MYO and normal electromyography (EMG) signals. In the present work, an efficient technique that involves wavelet transform using tunable-Q dynamics (TQWT) is proposed in order to identify disorders related to the neuromuscular domain of EMG signals. The EMG signal is decomposed by the TQWT technique into sub-bands, and these sub-bands are used to determine spectral features including spectral flatness, spectral stretch and spectral decrease, and statistical features including kurtosis, mean absolute deviation, and interquartile range. The extracted features are used as inputs into extreme learning machine classifiers in order to identify and analyze EMG signals associated with neuromuscular dysfunction. The results achieved with this technique illustrate a much better classification with regard to neuromuscular disturbance in electromyogram signals when compared with previous methods.