A Joint Speech Enhancement Algorithm Based On The Tri-Microphone

This paper proposes a novel speech enhancement algorithm combining the coherence-based algorithm and the differential beamforming. The coherence-based algorithm is applied to the broadside dual-microphone, which filters the side noise better than the previous method benefiting from the more accurate signal-to-noise ratio (SNR) estimation. However, the broadside array cannot eliminate the rear noise. Therefore, we apply the differential beamforming to the endfire dual-microphone to form a null at 180 degrees in the beambattern. The improvement in speech quality evaluation suggests that for various noise scenarios and with different input SNRs from - 10dB to 10dB, the proposed algorithm is more effective than the previous coherence-based algorithm and the ordinary differential beamformer with the tri-microphone.