一种提高喷泉码译码成功率的算法

Algorithm improving the decoding performance of fountain codes

喷泉码采用的信度传播(BP)译码算法与最大似然译码相比具有低复杂度的优势,然而受到编码随机性的影响,其失败概率也较高。该文提出一种改进译码算法,在传统的BP译码失败时,通过将未译出的数据包假设为已知,使得迭代译码的条件重新满足,译码过程得以延续,并通过后续的译码过程求出假设包的真实值,从而译出更多的数据包。研究了该算法中假设包选取策略和假设包数等参数的设置问题。仿真结果表明,对于码长为10000,译码开销为5%的LT码,采用最大度选取策略,假设包数为20时,该算法使译码成功率达到98.2%,比现有BP译码算法提高约18.3%。

The belief propagation （BP） iterative decoding algorithm used in fountain codes has lower decoding complexity than the maximum likelihood algorithm but higher failure possibility due to the encoding randomicity. A decoding algorithm was developed to increase the probability of successful decoding （PSD） for fountain codes,which continues the iterative process of BP decoding by assuming some undecoded data packets as decoded ones when the original BP algorithm fails. By figuring out the assumed packets in the consequent decoding process,more data can be decoded. The simulation with an LT code with a code-length of 10 000 and an overhead of 0.05 shows that the algorithm,having a PSD value of 98.2%,increases the PSD by 18.3% compared with the original BP algorithm,with the maximum degree strategy used and the assumption scale set to be 20.