Iterative multistage maximum likelihood decoding algorithm for multilevel codes and its applications.

摘要

The role of channel coding in digital communication system is to provide reliability, that is, a successful information transmission in the presence of noise and interference, with as small an error rate as required.; Half a century of research in Information Theory and Communications resulted in construction of many good codes and classes of codes. In general, longer codes achieve better performance, but the required time, memory, and amount of computation needed for successful decoding of these codes may in practice be infeasible. Thus the search for an efficient decoding algorithm is as important as the search for a good code. A good trade-off between the performance, measured by the low error probability, and efficiency, measured by the low decoding complexity, is set as a criterion.; Multistage decoding is devised for decoding codes with multilevel structure to achieve an efficient trade-off between error performance and decoding complexity. Multilevel code structure is used to simplify decoding. Component codes are decoded level-by-level in series of decoding stages, with the decoded information passed between them. Optimal for the codes of small and medium lengths and number of decoding stages, this technique shows a significant drop in performance when applied to longer codes, thus sacrificing performance to achieve efficiency.; In this dissertation, we develop an efficient soft-decision iterative multistage decoding algorithm for decoding decomposable and multilevel concatenated codes. This algorithm achieves maximum likelihood (ML) performance through iterations with optimality tests at each decoding stage. It is the first proposed multistage decoding algorithm that achieves ML performance, and at the same time has a significant reduction in average decoding complexity compared to other known ML decoding algorithms, such as Viterbi algorithm.; The application of the algorithm to two general classes of multilevel codes, decomposable linear block codes, on the example of Reed-Muller codes, and multilevel block coded modulation codes, is presented. The results show that this new algorithm achieves excellent performance-complexity trade-off.