Efficient Pruning for Successive-Cancellation Decoding of Polar Codes

Abstract

This letter presents an efficient pruning method for the successive-cancellation (SC) decoding of polar codes. In the proposed method, unnecessary sub-trees are pruned if the syndrome check of a constituent code is satisfied. Without sacrificing the error-rate performance, the recursive operation associated with such a constituent code is replaced with simple computations. Compared with the traditional SC decoding, the proposed pruning decreases the decoding latency by 85% on the average for the polar (1024, 512) code.