Modified PAC Codes

Polarization-adjusted convolutional (PAC) codes can approach the normal approximation (NA) bound using Fano decoding. However, when the received information is unreliable, the decoding may linger over the decoding tree, resulting in both a high decoding complexity and latency. This paper proposes the modified PAC (MPAC) codes and their hybrid Fano-SC (HFSC) decoding that prevents an impractical decoding. For the MPAC codes, only a subset of the information bits undergo the convolutional transform. Its output then concatenates the remaining information bits for the inner polar transform. Consequently, the Fano decoding and the successive cancellation (SC) decoding are deployed to recover the information bits that have undergone the convolutional transform and the remaining information bits, respectively. Both the MPAC code design and the HFSC decoding insight are studied. Our simulation results show that with a limited complexity, HFSC decoding of MPAC codes can yield a better performance-complexity tradeoff than Fano decoding of PAC codes and SC list (SCL) decoding of cyclic redundancy check (CRC)-polar codes.