Blind Separation of Convolutive Mixtures Using Oblique Projections

PENG Chunyi,ZHU Xiaolong,ZHANG Xianda
DOI: https://doi.org/10.3321/j.issn:1000-0054.2005.04.024
2005-01-01
Abstract:A subspace-type algorithm is developed for the blind separation of convolutive mixtures. The model includes three data matrices labeled as “the past”, “the present”, and “the future” to reformulate the convolutive mixing model into a static model via oblique projections. A linear blind separation algorithm is then used to reconstruct the original sources. The algorithm makes full use of the structure information inherent in the problem to obtain a deterministic solution to the first stage, which reduces the computational costs. Computer simulations validate the algorithm’s effectiveness.
What problem does this paper attempt to address?