An Increased Robustness of Speech Enhancement on Dual—Microphone System

Abstract

Because of the existence of interferences, non-directional noise, competing speakers, washing machines, vehicle transport, the necessity of speech enhancement, which includes extracting the speech component and removing background noise has attracted attention of researchers. Microphone array (MA) beamforming has commonly installed into several types of speech applications as an optimum alternative solution instead of single-channel approach. Dual-microphone array (DMA2) has the advantage of small compact size, easily implemented into numerous speech applications and applied various types of signal processing algorithm, such as mobile phone, hearing aids, teleconference system, surveillance devices, voice-controlled equipment. Unfortunately, in real-life recording environments, the DMA2’s evaluation often corrupted due to the different microphone sensitivities, the error of sampling frequency, inaccurate estimates of the preferred steering vector. In this research paper, the author presents an improved robustness speech enhancement of DMA2 in a realistic recording situation. The numerical simulation has shown the noise suppression to 10.0 dB, speech distortion was reduced to 11.2 dB and enhancing the speech quality in the term of signal-to-noise (SNR) ratio from 10.3 to 11.4 dB. The author’s proposed method can be applied into a multi-channel system for solving more complicated problems.