Study On Methods For Microphone Array Speech Enhancement Base On Near-Field

To the most of microphone array processing approaches are based on far-field models. However, the far-field assumption may not hold in practical applications such as speech acquisition with microphone arrays in small rooms or hands-free mobile telephone in automobiles. This thesis focuses on microphone array speech enhancement approaches which are based on near-field model.In this paper, firstly, we based on characteristic of acoustic propagation about near-field source, according the method by difference distance to reduce the impact of reverberation and noise which is located in the far-field on the voice. To achieve near-field beamforming and accurate expectations of near-field, using near-field distance and azimuth compensations technique. Secondly, researching on beamforming methods of Near-field based on far-field beamforming algorithm, then propose an improved method of NLMS algorithm, the simulation prove that the new algorithm have obtained better performance. At last, study far-field microphone array speech enhancement method using of oversampled cosine-modulates subband technology. Propose a microphone array speech enhancement method based on near-field beamforming is presented in the paper. Simulation results demonstrate that the proposed method exhibits a better noise reduction performance than other methods.