Spectral Processing Methods for Degraded Speech Signal

Assessment of clean speech from a noisy speech signal has been a research topic for a long time. This research finds its variety of applications, which includes the present mobile communication also. The most important outcome of this research is the improved quality and reduced listening effort in the presence of an interfering noise signal. In this paper the performance of various noise reduction techniques namely spectral subtraction, wavelet transforms, iterative subtraction, MMSE and Wiener filtering is done. This paper proposes a time-frequency estimator for enhancement of noisy speech signals in the discrete frequency transform domain. In the proposed method the estimation is based on modeling and filtering frequency components of noisy speech signal using Kalman filters. Experimental outcome show that the proposed method provides the better performance as compared to the other Spectral processing approaches.