关于前向最大后验概率(MAP)算法的研究

On Forward-only Maximum a Posteriori (MAP) Algorithms

推导了前向 MAP的一种简化形式—— F- MAP,并给出了它的两种实用算法 :Log- F- MAP和 Max- Log- F- MAP.F- MAP与 SOVA算法 (寄存器交换模式 )间有着密切的联系 ,特别是可将Log- F- MAP看作对 SOVA的进一步改进 .通过定点仿真 ,对 F- MAP和其他流行 SISO算法在Turbo译码中的 BER/FER性能进行了比较与分析 ,得到了译码延迟 D的合理值为 5～ 6倍的约束长度 .

In contrast to the traditional MAP (Maximum α posteriori) algorithm, forward-only MAP can find its applications in continuous turbo decoding/detection. This paper deduced the simplified version of forward-only MAP, called F-MAP, and proposed two practical version of F-MAP:Log-F-MAP and Max-Log-F-MAP. There is a close relationship between F-MAP and SOVA (register-exchange mode); specifically, the Log-F-MAP can be considered as the improved version of SOVA. By the fixed-point simulation, this paper discussed and compared the BER/FER performance with other popular SISO algorithms, while obtaining the judicious decoding delay D in the interval of 5-6 times constraint length.