Blind source separation of nonstationary signals using the singular value decomposition.

摘要

A new algorithm is proposed for the blind source separation problem utilizing the nonstationary statistics of the signals. This approach makes use of the singular value decomposition and Givens rotations to force separation of mixed signals. The algorithm works successfully on both speech and image signals and is ideal for on chip applications that require blind source separation to be done on the fly as data samples become available.