Research On Soft Decision Decoding

AbstractChannel coding is a fundarnental method to improve the conununication reliabilityTO reach the channel caPacity is the ulthate objechve of the research on channel codingtheory. Extensive prachce has shOwn tha soft decision decoding is one of the mostprospective aPproaches to this goal. Recentiy, the problem of finding computationallyefficient and practically imPlementable softdecision opted (or sub-optimal) decodingis still an open and challenging problem of channel coding theory. This thesis put itsemphasis on this problem to devise efficient soft decision decoding algorithms. Themain resultS are as following.l. Derive a simple form of generalised threshold, and proPOse a new search nde fortree-based A* decoding of block codes, and then obtain a threshold A* decodingalgorithIn. The proPOsed algorithm avoid redUndant search, thus speed up the A*decoding while mai-ng maximum lthelihood decwhng performance.2. Derive a generai form of generalized threshold, and obtain two kinds of thethreshold, i.e., Fano and A* generalized thresholds, and then propose a trellis-basedthreshold sequential decoding Stack algorithIn, which can avoid redundant searchand improve the decoding speed of the Stack algorithIn. The effects of path metricfunction bias item and search direction are corresPOndingly analyzed on bothdecoding Performance and complexity3. Use a computational efficient metric function and a more et1icient searchalgorithm --Dijkstra's algorithIn to search raPidly through the directed tree (graPh)of biM block codes for the maximurn likelihood error pattern. and performoptimality test on ermr pattems to speed up the decoding fiJwher. It is emphasizedthat usage of non-optimal bosmission signal wou1d result in nearly 3dBPerformance loss.4. ProPOse a fast soft decision decoding, which makes use of parallel structure andheuristic searching ability of the genetic algorithm (GA). \\!e fOrmalize thedecoding of block codes as an aPpropriate combinaorial optimization problem, andthen use GA to complete oPtimiZaion computaion raPidl}r, thus obtain a fast soft-decision decoding algoritIun. We indicate the some drawacks of M decodingalgorithIn of convolutional codes and convert the decoding into search process ingenetic space. It utilizes GA's good characteristic in POpulation diversity, widesearch region and global search ability to imProve the decoding performance.5. Point out that Chase algorithIn would generae reduplicate candidate codeworddining test decoding, which can lead to slow decoding speed. So we propose a novel iterative threshold Chase algorithm (ITC). It partitions the set of test patterns into the equivalence classes and generates the equivalence class representatives. Both test decoding and optimality test are performed only on the representatives. Therefore, much more redundant test decoding can be avoided and much more hard decision decoding time also can be saved while its decoding performance keeps same as Chase algorithm.