Attention-Based Noise Reduction for Surface-Electromyography: A Novel Method for Enhanced Signal Quality in Clinical Diagnostics

Surface electromyography (sEMG) signals are an important tool for monitoring and measuring muscle activity, rehabilitation, Human-Computer Interaction (HCI) systems, and diagnosis of neurological disorders. However, these signals are often affected by various sources of noise and disturbance during recording, which reduces the integrity, quality of the signal and increases the error of diagnostic applications. Traditional denoising techniques, such as filters and decomposition methods, often fail to handle the non-stationary nature of sEMG, resulting in a loss of essential information. This study introduces a novel denoising technique, Generalized Successive Variable Mode Decomposition (GSVMD), which integrates Successive Variational Mode Decomposition (SVMD), Soft Interval Thresholding (SIT), and attention mechanisms to enhance signal clarity. The proposed method was evaluated using data from twelve healthy subjects and twenty-four stroke patients, demonstrating a higher Signal-to-Noise Ratio (SNR) and lower R-squared (R²) values compared to conventional denoising techniques. Moreover, statistical tests, including paired t-tests and Analysis of Variance (ANOVA), confirmed the significant enhancements achieved by the method, with p-values less than ۰.۰۰۱ and p < ۰.۰۵, thereby validating its effectiveness and robustness. GSVMD utilizes data mining to dynamically adjust signal components, ensuring robust denoising without losing critical information. Its reduced dependency on hyperparameters and high computational efficiency make it suitable for real-time clinical applications, providing enhanced accuracy and reliability for neuromuscular assessments.

کلیدواژه ها:

Attention Mechanism ، Clinical Diagnostics ، Data Mining ، Surface Electromyography ، Successive Variational Mode Decomposition