Theoretical Research On Speech Enhancement Based On Microphone Arrays

With the rapid development of speech processing technology and with maturity of the telecommunication, people hope to acquire higher speech quality, so it is necessary to do some deep research on how to improve speech quality. In fact, to improve speech quality is to enhance speech signal. There are a variety of methods of speech enhancement. In terms of the number of microphones, methods can be divided into single microphone speech enhancement and multiple microphones (microphone arrays) speech enhancement. This paper mainly discusses speech enhancement based on microphone arrays.A microphone array is a system that consists of multiple microphones or loudspeakers placed at the different location. Speech enhancement based on microphone arrays mainly relies on beamforming. Generally, arrays connect with beamformer that provides multiple styles of special filtering. This paper mainly discusses the several typical algorithms in detail, such as, delay-sum beamforming, constant directivity beamforming, generalized sidelobe canceller, and Wiener post-filtering beamforming. By simulating these beamformings and analyzing their processing effect in the different noise field, finally this paper makes conclusions as follows:Delay-sum beamforming can cancel inherent noise perfectly, so that it can improve SNR of speech signals. By theoretical calculating, after the signal is processed by delay-sun beamforming, the SNR of array is N times of single microphone SNR. N is the number of microphones in the array.Constant directivity beamforming is a useful technique for special filtering in broadband signal environment, but the size of the array is related to the lowest frequency of the operation. It does not fit for processing signals with low frequency.Generalized sidelobe canceller fits to cancel the coherent noise signal. On the contrary, Wiener post-filtering beamforming can cancel the inherent noise completely.