一种短时延的Turbo码并行译码算法

由于迭代译码是Turbo码译码的主要特点,因而在译码的过程中会带来很大的时延.为了减小译码延时,本文将整块译码器分成w个子块,并且运用计算复杂度低的T-BCJR算法,在相邻的子块译码器之间相互运用边界分配值作为下一次迭代的初始值,而不是采用各相邻的子块之间重叠部分进行译码,故使译码延时下降为原来的1/w。

A Parallel Decoding Algorithm for Turbo Code with Low Delay

Since iterative decoding is an iterative algorithm, the decoding delay may not be acceptable. To reduce the delay, in this paper, we divided trellis stages into w sub-blocks and used the T-BCJR algorithm for each sub-block to complete the decoding for Turbo code. The method, different from the one based on adjacent sub-block overlapping decoding, uses the forward and the backward variables obtained in the previous iteration as boundary distributions of each sub-block decoder in the next iteration. And thus, the decoding delay is reduced by 1/w times of the original one.