Reverberation robust two-microphone Target Signal Detection algorithm with coherent interference

Abstract

In this paper, a reverberation robust Target Signal Detection (TSD) algorithm using two microphones is proposed. Most of traditional TSD algorithms are based on the assumption of free sound field and close-talking scene incorporate with multichannel system. They lack in achieving robustness in reverberant and noisy environment. The proposed TSD algorithm is based on Beam-to-Reference Ratio (BRR), and a novel estimator, Direct-to-Reverberate Ratio (DRR), is introduced to enlarge the basic assumption to reverberant and distant-talking scene. Spatial correlation information between microphones is used to estimate the DRR to revise threshold on each Time-Frequency (T-F) block and to form full-band likelihood using soft-decision information. Experimental results show that the proposed method performs robust in different reverberant environments with coherent interferences when target signal is from priori known direction-of-arrivals (DOA) in distant-talking scene.