Evaluating the Suitability of Wavelet Transform in Matlab for Application in Wearable Health Monitors

The incidence of neuromuscular disorders, including myopathy and Amyotrophic Lateral Sclerosis (ALS), has sharply risen in areas such as Arequipa, Peru. However, the lack of access to specialized diagnostic mechanisms has made diagnosis perfection impossible, primarily among underdeveloped areas. This study thereby provides the framework for developing and testing an EMG signal analysis system based on machine learning and wavelet transform to enhance diagnosis in wearables. The dataset is from the Nikolic PhD thesis, undertaking EMG recordings from healthy, myopathy, and ALS groups. This offers a strong basis for the discrimination between normal and pathological muscle activity. The EMG signals from the brachial biceps and medial vastus muscles were collected, preprocessed for denoising and artifact removal, and were analyzed using wavelet transform to decompose the EMG signal into its relevant frequency components. Features extracted from wavelet decomposition were then applied to train a Support Vector Machines (SVM) classifier for further differentiation between the three groups. The SVM classifier performed with a degree of accuracy, sensitivity, and specificity (92%, 90% and 94% ) that suggests this approach could help greatly in the early diagnosis and improved healthcare access for those facing neuromuscular disorders, especially in regions lacking healthcare access.