(An) efficient pruning method and its hardware architecture for successive-cancellation decoding of polar codes

Abstract

Polar codes are a class of error-correcting codes that provably achieve the channel capacity with the low-complexity encoding and decoding algorithms. Traditionally, successive-cancellation decoding is widely to decode polar codes. However, the serial nature of SC decoding leads to a long latency for long polar codes. This thesis proposes an improved method for the successive-cancellation decoding of polar codes to reduce the long latency. To avoid computations associated with redundant tree-traversals and syndrome calculations, recursive properties of polar codes are newly exploited in the proposed algorithm. Instead of computing a syndrome vector at every node, some syndrome vectors are directly obtained by recursively decomposing the syndrome vector computed previously. Furthermore, a modified syndrome check rule is proposed to prune unnecessary sub-trees efficiently. Compared to the latest pruning method, the proposed pruning method reduces the latency by 23% for a (2048, 1024) polar code without sacrificing the error-correcting performance. Moreover, the thesis proposes an efficient hardware architecture implementing the proposed SC decoding algorithm. A depth-constraint and a hardware sharing scheme are adopted to increase the hardware efficiency. As a result, the throughput of the polar decoder increases significantly at the cost of small amount of the additional hardware.