| Paper: | MLSP-L3.1 | ||
| Session: | Learning Theory II | ||
| Time: | Thursday, May 18, 10:00 - 10:20 | ||
| Presentation: | Lecture | ||
| Topic: | Machine Learning for Signal Processing: Information-theoretic learning | ||
| Title: | NORMALIZED INFORMATION THEORETIC CRITERIA FOR BLIND SIGNAL EXTRACTION | ||
| Authors: | Sergio Cruces, Ivan Duran-Diaz, University of Seville, Spain | ||
| Abstract: | In this paper we present new normalized criteria for the extraction of the scaled sources whose density have the minimum support measure or the minimum entropy. Both criteria are part of a more general entropy minimization principle based on Renyi's entropies. However, the proposed approach (based on Renyi's entropies or orders zero and one) have some special advantages, which allow to relax the assumption of having identically distributed source signals. | ||
©2018 Conference Management Services, Inc. -||- email: webmaster@icassp2006.org -||- Last updated Friday, August 17, 2012