Method and apparatus for blind source separation improving interference estimation in binaural Wiener filtering

摘要

The invention claims a method and an appropriate acoustic signal processing system for noise reduction of a binaural microphone signal (x₁, x₂) with one target point source (s) and M interfering point sources (n₁, n₂, ...,n_M) as input sources to a left and a right microphone (2) of a binaural microphone system, comprising the step of:- filtering a left and a right microphone signal (x₁, x₂) by a Wiener filter (14) to obtain binaural output signals (ŝ_L,ŝ_R) of the target point source (s), where said Wiener filter (14) is calculated as $\[ H_W=1-\frac{{\mathrm{\Phi }}_{\left(x_{1,n}+x_{2,n}\right)\left(x_{1,n}+x_{2,n}\right)}}{{\mathrm{\Phi }}_{\left(x_1+x_2\right)\left(x_1+x_2\right)}}, \]$where H_w is said Wiener filter (14), Φ(x_1,n+x_2,n)(x_1,n+x_2,n) is the auto power spectral density of the sum of all the M interfering point sources components (x1,n, x2,n) contained in the left and right microphone signal (x1, x2) and Φ(x₁+x₂)(x₁+x₂) is the auto power spectral density ofthe sum of the left and right microphone signal (x1, x2).;Owing to the linear-phase property of the calculated Wiener filter (14), original binaural cues are perfectly preserved not only for the target source (s) but also for the residual interfering sources (n1, n2, ... nM) .