Comparative Analysis of Filtering Noise from Voice Signal by Different Decomposition Methods

This study compares the denoising performance of Empirical Mode Decomposition (EMD) and Variational Mode Decomposition (VMD) techniques on signals corrupted with varying levels of noise. Results demonstrate that both methods effectively reduce noise and preserve signal characteristics, with higher Signal-to-Noise Ratios (SNRs) leading to better denoising performance. VMD exhibits slightly superior performance at lower SNRs, making it preferable in such scenarios. Both techniques maintain high correlation coefficients, indicating the preservation of signal shape and characteristics. However, in challenging cases with mixed noise types, both EMD and VMD face greater difficulties and exhibit higher error metrics. In conclusion, EMD and VMD are suitable for noise reduction tasks, with the choice between them depending on application specifics and desired trade-offs between computational complexity and denoising performance.