Reverberation robust multi-channel post-filtering using modified signal presence probability

Xiaofei Wang, Xu Li, Yanmeng Guo, Qiang Fu, Yonghong Yan

Abstract: A multi-channel post-filtering algorithm in reverberant environment based on detection and estimation scheme is proposed. The key estimator of signal presence probability (SP-P), which is in consideration of reverberation, is modified with direct-to-reverberate ratio (DRR). A new desired signal detection scheme is proposed to instruct the updating of transient noise and residual interference estimation, which takes advantage of SPP in the time-frequency domain. The proposed multi-channel post-filtering is tested in different non-stationary noisy and reverberant environments. Experimental results show that it outperforms the comparative algorithms on signal preservation of the desired speech with more noise reduction.