A Spectrum Adaptive Segmentation Empirical Wavelet Transform for Noisy and Nonstationary Signal Processing

Compared with thresholding methods based on the traditional wavelet transform (WT), empirical (EWT) has been demonstrated to outperform in terms of noise elimination by constructing an adaptive filter bank. However, as state-of-the-art version EWT, enhanced EWT (EEWT) requires that number components superposed signal prior knowledge is known, which impractical reality and limits application this method. In paper, a novel can adaptively estimate achieve spectrum segmentation proposed referred (SAS-EWT). Furthermore, customized SAS-EWT for speech enhancement proposed. According experimental results, our provides more accurate boundary detection better denoising performance. The method improves performance up 5% PESQ, STOI, SNR comparison EEWT.