LL Logo



Multichannel Blind
Equalization for
Non-White Signals

Athina P. Petropulu
Drexel University
Department of Electrical and Computer Engineering
Philadelphia, PA 19104
email: athina@artemis.ece.drexel.edu

We propose a novel adaptive scheme for the deconvolution of an unknown non-white signal, which is observed through two or more unknown FIR channels. This situation appears often in communications, where the unknown signal is the information bearing sequence, and the unknown channels model the effect of multipath propagation. The assumption that the unknown signal is non-white allows us to consider the interception and recovery of precoded by some unknown precoding filter communications signals.

Existing multichannel adaptive system identification algorithms estimate the orders and root locations of the channels, a task which in general leads to very slow convergence rates and large estimation errors. The proposed approach directly estimates the channels, instead of their root locations, by minimizing an error that is function higher-order cepstra of the observations. The channel lengths are initially overestimated, leading to the reconstruction of channels that have common zeros. This is again a two-channel blind deconvolution problem, but now the signals involved are the deterministic channels. As such, it can be solved using a non-adaptive estimation procedure.



LL Logo Disclaimer

Direct comments and questions to: webmaster@ll.mit.edu

MIT Lincoln Laboratory. All rights reserved.