CRC-aided Iterative Decoding of Turbo-hadamard Codes

Turbo-hadamard codes is an excellent low-rate turbo-type codes, but the decoding cost increases rapidly when the number of iterations increases with large parallel branches. This paper proposes a CRC-aided iterative decoding method to reduce the number of iterations while suffering no performance degradation. At each iteration, we perform Cyclic Redundancy Check (CRC) on the resulting bit sequence, and stop once it passes CRC to avoid unnecessary iterations. Otherwise, we generate multiple candidate paths based on the smallest corresponding LLRs. Then we perform CRC again until a first one passes CRC or reaches the maximum number of paths. Simulations show that this scheme can effectively reduce the number of iterations. Compared to the 2-state turbo-hadamard codes, the frame error rate increases by 1.1 dB at 10
^-3
and the bit error rate increases by 0.9 dB at 10
^-5
with 240-bit pseudo-random interleaving on the additive white Gaussian noise channel.