Multi-channel speech enhancement (MC-SE) could solve the drawbacks of single-channel speech enhancement (SC-SE) algorithms. Postfiltering is often needed after the spatial filtering for the MCSE algorithms. Among the MC-SE algorithms, the two-channel adaptive wiener post-filtering (TC-PF) is a simple and effective way for both noise reduction and dereverberation. The theoretical limits of the TCPF is based on the assumption that all the parameters could be estimated accurately. However, it is well-known that both the speech and the noise are stochastic, only the approximate parameters could be obtained due to the limited number of available data.

ZHENG Chengshi, ZHOU Yi, HU Xiaohu, and LI Xiaodong of Institute of Acoustics, Chinese Academy of Sciences carried out a series of studies and studies the statistical properties of the gain functions, which are often used for two-channel post-filtering (TC-PF) algorithms.

They reveal that the smoothing factor has a significant impact on both noise reduction and musical noise. They proposes an adaptive smoothing scheme by detecting the sudden change of the system. Moreover, the residual noise floor is adaptively chosen based on the structure of the noise power spectral density (NPSD) to further suppress the tonal noise components. Experimental results show the better performance of the proposed algorithm in terms of the segmental signal to- noise-ratio (SNR) and the PESQ improvements.

This research result was published on the recently issued IEEE International Conference on Acoustics, Speech and Signal Processing (2011, 1745-1748).