Detection and Removal of Impulsive White Noise from Noisy Speech

Clear speech sometimes will be polluted by noise. Speech enhancement aims at denoising noisy speech signal and extracting the clean speech. In this work, the noise regarded is impulsive white noise. Based on the difference of the power spectrum between noisy speech and white noise, the method of detecting impulsive white noise is developed. In detection, analyze the detail coefficients of noisy signal from discrete wavelet transform to find the information of impulsive white noise in time domain. The denoising method which is developed in wavelet domain is used in the removal. In denoising, the energy distribution of impulsive white noise in wavelet domain can be used to find thresholds for wavelet thresholding techniques. The efficiency of the algorithm is evaluated.