Denoising abnormal and normal biomedical signal using Optimal Wavelet Transform with Adaptive Thresholding

Electromyography (EMG) is widely used for diagnosing neuromuscular disorders by analyzing muscle electrical activity. However, EMG recordings are often corrupted by motion artifacts, power-line interference, and background noise, which significantly degrades signal interpretation, particularly in pathological cases such as myopathy. To address this challenge, this paper proposes an Optimal Wavelet Transform with Adaptive Thresholding (OWAT) for effective EMG signal denoising while preserving essential physiological features. The proposed approach automatically selects the most suitable wavelet basis from multiple wavelet families and determines the optimal threshold among four classical strategies— Sqtwolog, Rigrsure, Heursure, and Minimax—based on signal-to-noise ratio (SNR) improvement. The method is evaluated on both normal and myopathic EMG signals contaminated with additive Gaussian noise at various SNR levels. Experimental results show that the proposed OWAT method consistently outperforms EMD and conventional DWT in terms of MSE, RMSE, PRD, and SNR improvement, particularly under low-SNR conditions. These results confirm the robustness and effectiveness of the proposed approach for reliable EMG denoising in biomedical and clinical applications.