Abstract
Compared with thresholding methods based on the traditional wavelet transform (WT), empirical wavelet transform (EWT) has been demonstrated to outperform in terms of noise elimination by constructing an adaptive filter bank. However, as the state-of-the-art version of EWT, enhanced EWT (EEWT) requires that the number of components in the superposed signal as prior knowledge is known, which is impractical in reality and limits the application of this method. In this paper, a novel EWT that can adaptively estimate the number of components in the signal and achieve spectrum segmentation is proposed and is referred to as the spectrum adaptive segmentation empirical wavelet transform (SAS-EWT). Furthermore, a customized SAS-EWT for speech enhancement is proposed. According to the experimental results, our proposed SAS-EWT provides more accurate boundary detection and better denoising performance. The proposed method improves the performance by up to 5% in terms of PESQ, STOI, and SNR in comparison to EEWT.
Highlights
Noise removal, as an important research area of signal processing, focuses on clean signal extraction from noisy and nonstationary signals
According to the experimental results, our proposed customized SAS-empirical wavelet transform (EWT) improves the performance up to 5% on the objective metrics of perceptual evaluation of speech quality (PESQ), short-term objective intelligibility (STOI), and signal-to-noise ratio (SNR) in comparison to enhanced EWT (EEWT)
EVALUATION METRIC We evaluate the spectrum segmentation effectiveness based on three metrics: signal-to-noise ratio (SNR), perceptual evaluation of speech quality (PESQ) and short-term objective intelligibility (STOI)
Summary
As an important research area of signal processing, focuses on clean signal extraction from noisy and nonstationary signals. Some signal processing methods, such as WT and EMD, attempt to decompose the mixed noisy signals into distinct modes and extract the dominant mode as the original signal to remove noise. Different from the improvements in [16]–[18], Yue et al [19] considered the spectrum shape and proposed an enhanced empirical wavelet transform (EEWT) to overcome the poor segmentation of noisy and nonstationary signals and achieved excellent results. A spectrum adaptive segmentation empirical wavelet transform (SAS-EWT) method is proposed. Compared with EEWT in [19], our proposed SAS-EWT does not require the number of components in noisy signals in advance, avoids manual interference and realizes automatic spectrum segmentation. Compared with the DNNbased method, our proposed SAS-EWT does not need to collect a large number of clean-noisy signal pairs and omits the training process. The boundary effect and stop criteria of the EMD method are important and should be optimized [29]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
