The factor graph EM Algorithm: applications for LDPC codes

Author

Eckford, Andrew W.

Author_Institution

Dept. of Electr. & Comput. Eng., Toronto Univ., Ont., Canada

fYear

2005

fDate

5-8 June 2005

Firstpage

910

Lastpage

914

Abstract

The factor graph EM (FGEM) algorithm is introduced, which is a way of describing the EM algorithm as a message-passing algorithm over a factor graph. Some interesting properties of this algorithm are discussed, such as its ability to break certain cycles in factor graphs. Systems with LDPC codes are used as a starting point for practical applications of the FGEM algorithm. In particular, in channels with an unknown channel state, FGEM-based estimation-decoding algorithms can be straightforwardly obtained, and specific examples are given using the block fading channel. Applications for LDPC-based source coding, especially the Slepian-Wolf coding problem, are also given.

Keywords

decoding; expectation-maximisation algorithm; graph theory; message passing; parity check codes; source coding; FGEM; FGEM-based estimation-decoding algorithm; LDPC-based source coding; Slepian-Wolf coding problem; factor graph EM algorithm; low density parity check code; message-passing algorithm; Channel estimation; Fading; Inference algorithms; Iterative algorithms; Maximum likelihood decoding; Maximum likelihood estimation; Parameter estimation; Parity check codes; Signal processing algorithms; Sum product algorithm;

fLanguage

English

Publisher

ieee

Conference_Titel

Signal Processing Advances in Wireless Communications, 2005 IEEE 6th Workshop on

Print_ISBN

0-7803-8867-4

Type

conf

DOI

10.1109/SPAWC.2005.1506272

Filename

1506272