Speech Enhancement Based on Wavelet Transform and Improved Subspace Decomposition

This paper considers the enhancement and noise reduction of noisy speech assuming single-channel observation. A system based on the application of two different techniques is put forth. We propose to apply a denoising technique based on the reconstruction of the wavelet transform coefficients. This technique combines the straightforward multivariate generalization of the wavelet denoising method with the principal component analysis technique. In order to detect noisy time-frequency zones, a subspace decomposition-based post-processing technique is implemented. A number of simulations were achieved under various types of noise to evaluate the performance of our approach. Standard objective measures as well as subjective evaluations show that the proposed approach outperforms the compared speech enhancement methods for noise-corrupted speech at low levels of SNR.