WAVELET TRANSFORM-BASED FEATURE EXTRACTION FOR THE DETECTION OF MYOPATHY IN SURFACE EMG SIGNALS

Muscle movement and function are compromised by myopathy, a neuromuscular disorder that needs to be identified early for effective treatment.This study presents a deep learning approach to myopathy detection based on surface electromyography (sEMG) signals.The system employs Artificial Neural Networks (ANN), Recurrent Neural Networks (RNN), and Long Short-Term Memory (LSTM) models to classify myopathic and normal signals.Wavelet transform feature extraction is utilized to enhance time-frequency component representation following pre-processing of raw sEMG signals by band-pass filtering to remove noise and artifacts.These features are then used to input models, and regularization techniques such as dropout and hyperparameter adjustment are applied to further refine them.Accuracy, precision, recall, and F1-score are performance factors that indicate the best accuracy of RNN (92%), then ANN (72%), and LSTM (90%).In addition, real-time data visualization and monitoring enhance clinical usefulness.This research demonstrates the ability of deep learning, specifically temporal models, to capture intricate signal patterns for the early myopathy diagnosis.To enhance the diagnostic precision and availability, future work will focus on growing the dataset, developing the model, and deploying in real time.