A Study on Modified Graph Decoding for Low‐Density Parity‐Check Convolutional Codes

• Main Authors: Lu, Zu-Han, 呂祖漢
• Other Authors: Wang, Chung-Hsuan
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/8s4jcb

碩士 === 國立交通大學 === 電信工程研究所 === 104 === If there exist cycle in tanner graph of error-correcting code, the messages for computation in the sum-product algorithm (SPA) are observed to be statistically dependent, thereby degrading the decoding performance from the optimal one. Such performance degradation is even severer for low‐density parity‐check convolutional codes due to the repetitive structure among check and variable nodes in their tanner graphs. In this thesis, we propose two modified methods of SPA for mitigation of the cycle effect. Both the methods are based on a simple idea that the harmful cycles are replaced by proper trellis-decoding modules such that the dependence among messages in the modified tanner graph can be alleviated. By the simulation results, the proposed methods are verified to achieve satisfactory performance enhancements compared with SPA.