A Multi-channel Speech Enhancement Method Based on Subband Affine Projection Algorithm in Combination with Proposed Circular Nested Microphone Array

Abstract

In this paper, a novel multi-channel speech enhancement system is introduced based on a proposed circular nested microphone array (C-NMA) in combination with subband affine projection algorithm (SB-APA). The multi-channel speech enhancement methods have better accuracy because of information redundancy in comparison with single-channel methods. Firstly, a novel C-NMA is proposed with low computational complexity in comparison with other speech recording microphones. The C-NMA eliminates the spatial aliasing in microphone signals. Then, a subband step is implemented based on the speech components to increase the frequency resolution. The affine projection algorithm is implemented adaptively on the subband signals by C-NMA. Finally, the subband signals are combined by the synthesize filters and the enhanced signal is produced. The accuracy of the proposed method is compared with least mean square (LMS), traditional APA, recursive least square (RLS), and real-time generalized cross-correlation non-negative matrix factorization (RT-GCC-NMF). The results show the superiority of the proposed method in comparison with other previous works in noisy and reverberant environmental conditions.