Author :
Ferréol, Anne ; Albera, Laurent ; Chevalier, Pascal
Abstract :
Most of the current second order (SO)(Belouchrani, A. et al., IEEE Trans. Sig. Process., vol.45, no.2, p.434-44, 1997) and fourth order (FO) (Cardoso, J.F. and Souloumiac, A., IEE Proc.-F, vol.140, no.6, p.362-70, 1993; Comon, P., Sig. Process., vol.36, no.3, 1994) blind source identification (BSI) methods aim at blindly identifying the steering vectors of statistically independent sources, provided the number of sources is not greater than the number of sensors. However in practical situations, the probability of receiving more sources than sensors increases with the reception bandwidth. In this context, the paper proposes a new attractive FO BSI method, able to identify the steering vector of more sources than sensors, jointly with a new pertinent performance criterion for the quality evaluation of the BSI process. The new method implements a FO pre-whitening step and exploits the tri-spectrum diversities of the sources.
Keywords :
array signal processing; blind source separation; identification; blind source identification; fourth order blind identification; performance criterion; pre-whitening step; sensor arrays; statistically independent sources; steering vectors; tri-spectrum diversities; underdetermined mixtures; Bandwidth; Bismuth; Gaussian noise; Narrowband; Niobium; Probability; Robustness; Sensor arrays; Statistics;
Conference_Titel :
Acoustics, Speech, and Signal Processing, 2003. Proceedings. (ICASSP '03). 2003 IEEE International Conference on