A Dual-microphone Sub-band Post-filter Using Simplified TBRR for Speech Enhancement

Post-filtering is an effective method for further reducing noise components at a beamformer output. Existing techniques either are inefficient at suppressing highly non-stationary noise, or possess complex calculations. This paper proposes a novel dual-microphone sub-band post-filtering algorithm using a simplified transient beam to reference ratio (TBRR), applicable to adaptive beamformer, which is named SS-TBRR (simplified sub-band-TBRR). The sub-band signal processing approach is extended to the post-filtering for reducing computational complexity and smoothing the spectrum of the processed signal to eliminate musical noise. The relation between the observed signal, beamformer primary output, and the reference noise signal is exploited to differentiate non-stationary noise components from speech components. Based on speech presence probability and combined with an appropriate spectral enhancement approach, non-stationary noise is reduced significantly without canceling desired signal. Experimental results verify the effectiveness and robustness of the proposed algorithm.