Adaptive Fast Successive-Cancellation Decoding of Polar Codes Using RBIR

Successive-cancellation (SC) decoding algorithm is the most commonly used method for decoding the polar codes. It can be viewed as a depth-first binary tree search, traversing from root to all the leaf nodes in the direction of left to right, where each leaf node corresponds to a codeword bit to be decoded. It decodes the bits sequentially one after the other, increasing the decoding latency with the increase in codeword length. In this paper, we propose an adaptive fast SC decoder that uses a link abstraction metric called received bit information rate (RBIR), in identifying the depth level, at which all the intermediate nodes and the attached information bits can be decoded directly without traversing till the leaf node by using the recursive structure of the SC decoder. We show that for a target block error rate (BLER) of 10%, the proposed RBIR-based adaptive fast SC (RBIR-AFSC) decoder has around 48% ~ 54% reduction in decoding latency compared to that of conventional SC decoder, without any loss in BLER performance, for 3GPP defined 5G NR TDL-A channel model.