Speech enhancement based on noise-compensated phase spectrum

In this paper, a noisy speech enhancement method based on noise compensation performed on short time phase spectrum is presented. Here the noise estimate to be used to modify the noisy speech phase spectrum is proposed to be determined by exploiting the low frequency regions of noisy speech of current frame rather than depending only on the initial silence frames. We argue that this approach of noise estimation offers the capability of the tracking the time variation of the non-stationary noise. By employing the noise estimates thus obtained, a procedure is formulated to compensate the distortion in the phase spectrum, which is kept unchanged in the typical speech enhancement methods. The noise compensated phase spectrum is then recombined with the magnitude spectrum to produce a modified complex spectrum prior to synthesize an enhanced frame. Extensive simulations are carried out using NOIZEUS database in order evaluate the performance of proposed method. It is shown in terms of objective measures, spectrogram analysis and informal subjective listening tests that the proposed method consistently outperforms a state-of-the-art method of speech enhancement from noisy speech corrupted by car or babble noise of very low levels of SNR.