PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Channel Decoding with a Bayesian Equalizer
Luis Salamanca, Juan Jose Murillo-Fuentes and Fernando Perez-Cruz
In: IEEE ISIT 2010, June 13-18, 2010, Austin, TX.

Abstract

In this paper we show that, in case of uncertainties during the estimation, overconfident posterior probabilities tend to mislead the performance of soft-decoders. Maximum likelihood (ML) estimates of the channel state information (CSI) make the equalizer to provide overconfident posterior probabilities of the equalized symbols half of the time, that can derail the decoder in case of wrong estimated bits. Thus, as a solution we propose and analyze a Bayesian equalizer that produces more accurate probabilities, because it considers the uncertainties in the estimation. This approach is based on an averaged BCJR over the probability density function of the estimated CSI. We exploit the improvement in the posterior probabilities by feeding the channel decoder with these better estimates. The proposed method exhibits a much better performance compared to the ML-BCJR when a LDPC decoder is considered, as illustrated in the experiments.

PDF - Requires Adobe Acrobat Reader or other PDF viewer.
EPrint Type:Conference or Workshop Item (Talk)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Learning/Statistics & Optimisation
Theory & Algorithms
ID Code:7549
Deposited By:Fernando Perez-Cruz
Deposited On:17 March 2011