Number of the records: 1
Fast Approximate Joint Diagonalization Incorporating Weight Matrices
- 1.
SYSNO ASEP 0321670 Document Type J - Journal Article R&D Document Type Journal Article Subsidiary J Článek ve WOS Title Fast Approximate Joint Diagonalization Incorporating Weight Matrices Title Rychlá přibližná diagonalizace s váhovými maticemi Author(s) Tichavský, Petr (UTIA-B) RID, ORCID
Yeredor, A. (IL)Source Title IEEE Transactions on Signal Processing - ISSN 1053-587X
Roč. 57, č. 3 (2009), s. 878-891Number of pages 14 s. Publication form www - www Language eng - English Country US - United States Keywords autoregressive processes ; blind source separation ; nonstationary random processes Subject RIV BB - Applied Statistics, Operational Research R&D Projects 1M0572 GA MŠMT - Ministry of Education, Youth and Sports (MEYS) CEZ AV0Z10750506 - UTIA-B (2005-2011) UT WOS 000263431900006 DOI 10.1109/TSP.2008.2009271 Annotation We propose a new low complexity Approximate Joint Diagonalization (AJD) algorithm, which incorporates nontrivial block-diagonal weight matrices into a Weighted Least-Squares (WLS) AJD criterion. We show how the new algorithm can be utilized in an iteratively-reweighted separation scheme, thereby giving rise to fast implementation of asymptotically optimal BSS algorithms in various scenarios. In particular, we consider three specific (yet common) scenarios, involving stationary or block-stationary Gaussian sources, for which the optimal weight matrices can be readily estimated from the sample covariance matrices (which are also the target-matrices for the AJD). Comparative simulation results demonstrate the advantages in both speed and accuracy, as well as compliance with the theoretically predicted asymptotic optimality of the resulting BSS algorithms based on the weighted AJD, both on large scale problems with matrices of the size 100 x 100. Workplace Institute of Information Theory and Automation Contact Markéta Votavová, votavova@utia.cas.cz, Tel.: 266 052 201. Year of Publishing 2009
Number of the records: 1