A combination of semisoft and μ-law thresholding functions for enhancing noisy speech in wavelet packet domain

Estimating an exact threshold value as well as designing an appropriate thresholding function are important criterion for achieving the desired performance of a noisy speech enhancement method in the Wavelet packet (WP) domain. This paper presents a new speech enhancement method, where the threshold in a speech or silent subband is statistically determined using the Teager energy operated WP coefficients of the noisy speech. The threshold thus obtained is applied on the WP coefficients of the noisy speech based on a combination of semisoft and μ-law thresholding functions, which are chosen according to the approximate and detail WP coefficients in a speech or silent subband. Such a thresholding function is designed to minimize signal distortion as well as to prevent time-frequency discontinuity. Standard objective measures as well as subjective observations show that the proposed method outperforms the existing state-of-the-art thresholding based approaches of noisy speech enhancement from high to low levels of SNR thus resulting in an enhanced speech with better quality and intelligibility.