Record-breaking error correction using Low-Density Parity-Check Codes

Matthew C Davey

Gallager's low-density parity-check codes have near Shannon limit performance when decoded using a Viterbi-like decoding algorithm. We report two advances that further improve the error-correction performance of these codes. First, defining the codes over non-binary fields we can obtain a 0.6dB improvement in signal to noise ratio for a given bit error rate. Second, using parity-check matrices with a variety of row and column weights we obtain gains of up to 0.5dB. When combined, these changes result in codes whose empirical error-correction performance is unbeaten for the additive white Gaussian noise channel.

Status: Essay available in the University Library, University of Cambridge.

Download Compressed postscript.