Tailored List Decoding of Polar Codes
Résumé
In this paper, we provide some experimental evidence that a significant number of metric updates and sorting can be avoided during list decoding of polar codes. Starting from the observation that a non-negligible part of the reliable bit channels are located at the tail of the decoding sequence, it is possible to decode the end of the sequence without updating the metrics of list decoding. This tailored CRC-Aided successive cancellation list (T-CA-SCL) decoder can be seen as a simplification of SCL decoding. Moreover, a heuristic is proposed to determine the parameters required by the tailored-CA-SCL decoder. Experimental results show that the proposed tailored list decoding approach reduces the number of metric updates by 45% without significantly altering the error performance. Finally some leads are provided for further research to improve and extend the proposed approach.