Interaural coherence preservation in MWF-based binaural noise reduction algorithms using partial noise estimation

摘要

Besides noise reduction an important objective of binaural speech enhancement algorithms is the preservation of the binaural cues of both desired and undesired sound sources. Recently an extension of the binaural Multi-channel Wiener filter (MWF), namely the MWF-IC, has been presented which aims to preserve the Interaural Coherence (IC) of the noise component. Since for the MWF-IC no closed-form solution exists, in this paper we propose to preserve the IC using the binaural MWF with partial noise estimation (MWF-N), for which a closed-form solution exists. Furthermore, we derive a closed-form expression for the trade-off parameter in the MWF-N yielding a predefined IC at the filter output. Experimental results in a diffuse noise scenario show that both the MWF-IC and the MWF-N preserve the IC of the output noise component. However, the MWF-IC yields a better noise reduction performance whereas the MWF-N introduces less speech distortion.