Iterative decoding of turbo product codes based on the sum-product algorithm for magnetic recording channels

Abstract

This project considers the iterative decoding of turbo product codes (TPC) with eBCH codes (TPC/eBCH) based on the belief propagation (BP) over magnetic recording channels.

The existence of many short cycles in the code graph of eBCH codes seriously degrades the performance of iterative decoding. Hence, the conventional sum-product algorithm (SPA) does not perform well for eBCH codes; while the optimal a posteriori probability (APP) decoding has a high computational complexity. In order to overcome the issue caused by cycles, several approaches have been developed in this project to improve the bit-error-rate (BER) performance. For example, a procedure is introduced to perform the SPA over a binary parity check matrix that adapts based on the Gaussian Elimination (GE). Hence, the procedure is termed the Adaptive Belief Propagation (ABP) algorithm.