Underdetermined source separation for colored sources
Stefan Winter, Walter Kellermann, Hiroshi Sawada, Shoji Makino · 2006
This contribution focuses on the source separation stage as im-portant part of underdetermined blind source separation (BSS). So far nearly all approaches for underdetermined BSS assume inde-pendently, identically distributed (i.i.d.) sources. They completely ignore the redundancy that is in the temporal structure of col-ored sources like speech signals. Instead, we propose multivariate models based on the multivariate Student’s t or multivariate Gaus-sian distribution and investigate their potential for underdetermined BSS. We provide a simple yet effective filter based on the sources’ autocorrelations for recovering the sources as basis for further ad-vances in underdetermined BSS. The challenge is estimating the filter coefficients blindly. The experimental results support the idea that source separation for underdetermined BSS can be reduced to the separation of their autocorrelations. 1.