Speech Denoising Based on Wavelet Transform of Correlation Function
Xiaofan Li, Sheng Yin · 2021 International Conference on Electronic Information Technology and Smart Agriculture (ICEITSA) · 2021
Speech signal is a non-stationary short-time transient signal. The useful signal is in the same frequency range as the noise. The traditional de-noising method in the form of filter can not separate the noise effectively. Wavelet transform has the characteristics of time-frequency local analysis. By decomposing the noisy signal, separating the noisy signal and reconstructing the useful signal, the noise can be effectively removed. White noise is a stationary random signal, and the wavelet transform on different scales is irrelevant. Combined with the basic idea of wavelet de-noising, a method of speech de-noising based on wavelet transform of correlation function is proposed. Through MATLAB simulation, the de-noising method determined by correlation function can effectively remove the white noise of speech signal.