Title :
Separation of convolutive mixtures using higher-order statistics
Author :
Chen, Binning ; Petropulu, Athina P.
Author_Institution :
Dept. of Electr. & Comput. Eng., Drexel Univ., Philadelphia, PA, USA
Abstract :
In this paper we propose a frequency domain approach for the separation of convolutive mixtures based on higher order statistics only. The system frequency response is first obtained up to a diagonal phase ambiguity matrix based on generalized eigen-decomposition of two cross-trispectrum matrices. Then the phase ambiguity is removed by exploiting different slices of the cross-trispectrum
Keywords :
MIMO systems; convolution; eigenvalues and eigenfunctions; frequency response; frequency-domain analysis; higher order statistics; matrix decomposition; spectral analysis; HOS; convolutive mixtures; cross-trispectrum matrices eigendecomposition; diagonal phase ambiguity matrix; frequency domain approach; generalized eigendecomposition; higher-order statistics; phase ambiguity removal; received signals; system frequency response; Biomedical engineering; Digital signal processing; Fourier transforms; Frequency domain analysis; Frequency estimation; Frequency response; Higher order statistics; MIMO; Matrix decomposition; Speech enhancement;
Conference_Titel :
Wireless Communications and Networking Confernce, 2000. WCNC. 2000 IEEE
Conference_Location :
Chicago, IL
Print_ISBN :
0-7803-6596-8
DOI :
10.1109/WCNC.2000.904797
