For many signal sources such as speech with distinct, nonwhite power spectr
al densities, second-order statistics of the received signal mixture can be
exploited for signal separation. Without knowledge on noise correlation ma
trix, we propose a simple and yet effective signal extraction method for si
gnal source separation under unknown temporally white noise. This new and u
nbiased signal extractor is derived from the matrix pencil formed between o
utput autocorrelation matrices at different delays, Based on the matrix pen
cil, an ESPRIT-type algorithm is derived to get an optimal solution in leas
t square sense. Our method is well suited for systems with colored sensor n
oises and For nonstationary signals.