Study On Microphone Array Speech Enhancement Algorithm

In the speech communication application such as teleconference, hands-free mobile telephone and hearing aids and so on, speech is usually inevitably be interfered by noise, reverberation even other audio signals. These interferences will not only cause the degradation of speech quality, but also lead to a sharp deterioration performance of the speech processing system. Consequently it is becoming more urgent to be solved that choose which speech enhancement algorithms to get as clean as possible speech from noisy speech.This paper describes and compares several commonly used microphone array speech enhancement algorithms. Then, a detailed analysis of the Generalized Sidelobe Canceller structure is given. In the Generalized Sidelobe Canceller structure, the mismatch between the estimated and actual direction of arrival of the desired signal will cause signal leakage which resulting in speech distortion. In view of this shortcoming, two improved algorithms based on Generalized Sidelobe Canceller are proposed. The Multiple Signal Classification algorithm is introduced in the first improved algorithm to estimate the direction of arrival of the desired signal, which reduces the position error and improves the position performance, but it still can't eliminate the degradation performance caused by the position error completely. So, in the second improved algorithm, the adaptive algorithm of the improved Generalized Sidelobe Canceller structure is changed to avoid the signal leakage, a feedback module is introduced to speed up the convergence of the systems and a Wiener filter structure is adopted to suppress the incoherent noise more effectively. The simulation results show that: compared with traditional Generalized Sidelobe Canceller, two proposed improved algorithms achieve better robustness in the presence of mismatch of direction of arrival, and have better output signal-to-noise ratio.