Adaptive subband Wiener filtering for speech enhancement using critical-band gammatone filterbank

This paper presents a subband denoising technique, called the adaptive subband Wiener filtering (ASWF), based on a critical-band gammatone filterbank. The ASWF is derived from the generalized subband Wiener filtering (SWF) equation and reduces noises according to the estimated signal-to-noise ratio (SNR) in each auditory channel and in each time frame. Design of the subband noise estimator will be presented. This critical-band denoising technique would be of interest to auditory-based speech and audio applications. Objective and subjective tests demonstrated the proposed technique can improve the perceptual quality of speeches.