Low-latency polar codes via hybrid decoding

Abstract

In this paper, we propose a family of hybrid decoders for polar codes. By decomposing the overall polar code into an inner code and an outer code, a hybrid decoder in the family uses successive cancellation (SC) to decode the inner code and maximum-likelihood (ML) to decode the outer code. At one extreme in the family is the ML decoder, when the entire polar code is viewed as the outer code; at the other extreme is the SC decoder, when the entire polar code is viewed as the inner code. Since ML decoding has lower latency than SC decoding, a hybrid decoder can achieve lower latency than the conventional SC decoder, at the expense of higher complexity due to the ML decoding of the outer code. We propose a reduction in the complexity of ML decoding by exploiting the structure of polar codes.