0032554 20240111140637.620060524235959.9 4 DvD-rom cav_un_epca*00765041-4244-0469-X873-876BryanIEEE2006 independent component analysis blind source separation cav_un_auth*0108100 Koldovský Zbyněk UTIA-B Department of Stochastic Informatics Ústav teorie informace a automatizace AV ČR, v. v. i. cav_un_auth*0101212 Tichavský Petr UTIA-B Department of Stochastic Informatics Ústav teorie informace a automatizace AV ČR, v. v. i. 231 kB 12B 1M0572 GA MŠk cav_un_auth*0001814 GA102/05/0278 GA ČR cav_un_auth*0051169 CEZ:AV0Z10750506 In this paper, a number of important differences is discussed tht exist between independent component analysis with and without presence of an additive noise. It is shown that estimation of the mixing/demiging matrix should not be the main goal in the noisy case. Instead, usage of the signal-to-intarference ples noise criterion is suggested. Second, it is proposed to normalize norms of rows of the demixing matrix in the Monte Carlo experiments. V tomto clanku je diskutovana rade existujicich rozdilu mezi slepou separaci linearni smesi signal bez pritomnosti a za pritomnosti aditivniho sumu. Je ukazano, ze odhad mixovaci nebo separujici matice by nemel byt hlavnim cilem separace, ale ze by jim melo byt dosazeny pomer uzitecneho signalu vuci interferenci a sumu. Dale je v clanku zduvodneno proc je vhodne v Monte Carlo simulacich normovat radky de-mixovaci (separacni) matice. cav_un_auth*0215014 IEEE International Conference on Acoustics, Speech and Signal Processing Toulouse 14.05.2006-19.05.2006 FR 2007 BB http://hdl.handle.net/11104/0133017 2006 231 kB cav_un_epca*0076504 Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing 1-4244-0469-X 873 876 Bryan IEEE 2006