Blindly Estimating and Localizing Multiple Signals from the Matrix Channel Impulse Response
October 2004
Robert M. Taylor, Jr., The MITRE Corporation
Garry M. Jacyna, The MITRE Corporation
Lamine Mili, ECE Department, Virginia Tech
Amir I. Zaghloul, ECE Department, Virginia Tech
ABSTRACT
We consider the problem of estimating and localizing a set
of unknown real-valued signals simultaneously arriving at
a sensor array composed of elements spaced far enough
apart to induce measurable baseband time difference of arrivals.
The delayed and mixed line-of-sight signals form a
convolutive mixture model. We recover the source signals
through multichannel blind deconvolution (MBD) in which
the channel impulse response estimate provides the direction
of arrival (DOA) information. We consider the two
cases in which either the source signal probability density
is known or the attractor space for which the source density
resides is known. The new MBD algorithm we present
works with known source densities to minimize the symmetric
Kullback-Liebler distance to the standardized equalized
output. Simulation results show successful performance
on acoustic data.
Additional Search Keywords
n/a
|
