CTH14-6: Error Correction Using a Message-Passing Decoder to Process Cyclic Redundancy Checks

This paper proposes a method for correcting errors in messages encoded using cyclic redundancy checks (CRCs), which are typically only used for error detection. This is accomplished for messages of a known length by deriving a parity-check matrix representation of the CRC. The parity-check matrix can then be used to correct errors using the message-passing decoder (MPD) commonly used to decode LDPC codes. The CRC parity-check matrix is not sparse, which inhibits performance of the MPD, but this effect can be reduced by modifying the matrix using recently proposed sparsification techniques to eliminate short cycles. The technique is practical mainly for codes with short block sizes, and can also be applied to coding schemes that concatenate a CRC with an outer error correcting code. Simulation results demonstrate some performance gains, although less than what can be gained using custom-designed LDPC codes for the same rate and message length.