Low Complexity MAP Algorithm for Turbo Decoder

As a promising decoding algorithm for turbo codes in terms of relatively low BER, the maximum a posteriori MAP algorithm is most widely used. However, the conventional MAP algorithm requires a large number of computations. A modified MAP algorithm is therefore proposed for reduction of the associated memory size and ultimately power saving. A newly introduced block combing is performed for the memory efficiency such that two branch metrics BMs are merged into one branch metric. When calculating FSM Forward State Metric of the associated state transition, BM is included in the subsequent FSM, and thus when calculating APP A Posteriori Probability, the BM is exempted and the number of computations for LLR Log Likelihood Ratio is reduced. Simulation results demonstrate reduced memory size in use and equivalent performance, compared to the conventional MAP algorithm.