DocumentCode :
791420
Title :
Analysis and optimization of irregular LDPC codes for joint source-channel decoding
Author :
Poulliat, Charly ; Declercq, David ; Lamy-Bergot, Catherine ; Fijalkow, Inbar
Author_Institution :
ETIS, Cergy Pontoise, France
Volume :
9
Issue :
12
fYear :
2005
Firstpage :
1064
Lastpage :
1066
Abstract :
In this paper, we present a characterization, through its convergence analysis, and an optimisation of a joint source-channel receiver composed of a LDPC decoder and a Soft Input Soft Output (SISO) source decoder. Under Gaussian approximation, assuming the knowledge of the extrinsic mutual information transfer function (EXIT chart) of the source decoder, we derive the Mutual Information evolution equations, that semianalytically describe the convergence of the iterative system behavior and, to complete the study, the stability condition at the convergence fixed point is derived for the joint receiver. From this analysis, a general optimisation method of the irregularity of the LDPC codes is proposed, which can be reduced to a linear programming optimisation problem. Simulation results show improved performance when compared to an AWGN optimized LDPC code.
Keywords :
Gaussian channels; approximation theory; combined source-channel coding; convergence of numerical methods; iterative decoding; linear programming; parity check codes; transfer functions; EXIT chart; Gaussian approximation; JSCD; SISO decoder; convergence analysis; extrinsic mutual information transfer function; general optimisation method; irregular LDPC code; iterative system; joint source-channel decoding; linear programming; mutual information evolution equation; optimisation problem; soft input soft output source; Convergence; Equations; Gaussian approximation; Iterative decoding; Linear programming; Mutual information; Optimization methods; Parity check codes; Stability; Transfer functions;
fLanguage :
English
Journal_Title :
Communications Letters, IEEE
Publisher :
ieee
ISSN :
1089-7798
Type :
jour
DOI :
10.1109/LCOMM.2005.1576589
Filename :
1576589
Link To Document :
بازگشت