BLIND ESTIMATION OF CORRELATED SOURCE SIGNALS
Lang Tong, Rueywen Liu · 2005
Blind estimation of source signals is a signal extraction problem that assumes no information of the channel characteristics. In this paper, the problem of blind estimation of possibly correlated multiple source signals transmitted through a memoryless channel is investigated. In particular, our objective of blind signal estimation is to obtain fhe estimates that preserve the waveforms of the original source signals (waveform-preserving estimates). A necessary condition and two sufficient con dit ions for a c h ie v ing w a v e f o rm -p r e s e r v i ng estimation are proved. The AMUSE (Algorithm for Multiple Signal Extraction) [8] is extended so that the signal extraction can be achieved for possibly correlated sources.