Spectral Subtraction using Critical Band Wavelet Packet Decomposition with Noise Estimation Based on Spectral Entropy

Abstract

he wavelet packet decomposition, which reflects the human audi- tory system, can be used, in order to improve intelligibility of the enhanced speech. In this paper, we propose the spectral subtraction method using critical band wavelet packet decomposition with noise estimation based on spectral en- tropy. The aim of proposed method is to alleviate drawbacks of speech en- hancement method using wavelet thresholding in low SNR environment. The proposed method is evaluated on various noise conditions . white Gaussian, pink, F-16 cockpit, and speech babble noise. We compare the proposed method with the node dependent thresholding one. The proposed method shows better performance than the node dependent thresholding method