Threshold Parameters Selection for Empirical Mode Decomposition-Based EMG Signal Denoising

Empirical Mode Decomposition (EMD) is a data-driven and fully adaptive signal decomposition technique to decompose into its Intrinsic Functions (IMF). EMD has attained great attention due capabilities process in the frequency-time domain without altering frequency domain. EMD-based denoising techniques have shown potential denoise nonlinear nonstationary signals compromising signal’s characteristics. The procedure comprises three steps, i.e., decomposition, IMF thresholding, reconstruction. Thresholding performed assess which IMFs contain noise. In this study, Interval (IT), Iterative (IIT), Clear (CIIT) been explored for of electromyography (EMG) signals. effect different thresholding operators, SOFT, HARD, Smoothly Clipped Absolute Deviation (SCAD), on performance EMG also investigated. 15 signals, recorded from upper limb 5 healthy subjects, were used identify best possible combination operator assessed by calculating Signal Noise (SNR) ratio results are further evaluated using two-way Analysis Variance (ANOVA) statistical test. demonstrated that mean SNR values yielded IIT outperform IT (P-value < 0.05), but there no significant difference CIIT = 0.9951). For HARD SOFT 0.0968), whereas outperforms SCAD 0.05). It with threshold value equal half universal while preserving original IIT-based yields highest SNR, irrespective level noise embedded signal. Whereas provides comparable successfully preserves shape identified can eliminate various noises