Fuzzy Entropy Spectrum Analysis for Biomedical Signals De-Noising

Abstract

Singular spectrum analysis (SSA) is widely applied to de-noise noisy biomedical signals in a broad range of applications. The major idea behind SSA de-noising is to divide the noisy signal into signal and noise subspaces by truncation of singular spectrum at a certain order. However, there is no clear 'noise floor' for many real-world biomedical signals. In addition, such de-noised signal contains the highest possible residual noise level. In this study, fuzzy entropy (FuzzyEn), a robust measure to quantify the signal complexity in chaos theory, is introduced to provide a genuine noise floor, which indicates the noise level of each SSA components relative to white noise and original noisy signal. Based on the FuzzyEn spectrum and filter bank characteristics of SSA, an iterative soft threshold SSA approach (SSA-IST) is then proposed to remove the noise in each component. The experimental results of de-noising speech and electromyographic (EMG) signals using the proposed approach are presented and compared with the results obtained using existing truncated SSA, wavelet transform (WT), and empirical mode decomposition (EMD) techniques.