Title :
Reduced-search BCJR algorithms
Author :
Franz, V. ; Anderson, J.B.
Author_Institution :
Lehrstuhl fur Nachrichtentech., Tech. Univ. Munchen, Germany
fDate :
29 Jun-4 Jul 1997
Abstract :
Summary form only given. There is great interest in coding systems that employ various kinds of code concatenation. In all of these schemes, an important element in the decoder is the MAP decoder, a device that puts out the probability of trellis states or data bits, rather than simply the most likely state or bit. For trellis encoding and Markov data, the MAP decoder is a special scheme, the BCJR algorithm. Unfortunately, the BCJR algorithm is computationally intensive. The purpose of this paper is to present a strong simplification of it that does not sacrifice the decoder error performance. Our algorithms exploit the fact that most working probabilities in the BCJR algorithm are very small, and with a little care can be ignored without losing performance. We find that the most successful strategy is to ignore working probabilities that fall below a certain threshold
Keywords :
Markov processes; concatenated codes; decoding; maximum likelihood estimation; probability; search problems; trellis codes; BCJR algorithm; MAP decoder; Markov data; code concatenation; coding systems; data bits; error performance; reduced-search BCJR algorithms; simplification; trellis encoding; trellis states; working probabilities; Concatenated codes; Encoding; Error probability; Iterative decoding; Noise reduction; USA Councils;
Conference_Titel :
Information Theory. 1997. Proceedings., 1997 IEEE International Symposium on
Conference_Location :
Ulm
Print_ISBN :
0-7803-3956-8
DOI :
10.1109/ISIT.1997.613145