Universal Context Based Decoding with Non-systematic Low-Density Parity-Check Codes

Universal estimation strategies with the recently developed non-systematic split-LDPC codes are proposed to improve channel decoding of sequences that contain context based redundancy. The new methods combine techniques from universal compression, such as the burrows-wheeler transform (BWT), and segmentation of piecewise stationary memoryless sources (PSMS´s) with recently proposed methods of discrete denoising. Simulation results with non-systematic low density parity check codes show significant improvements of the proposed methods on systematic LDPC codes, even when the actual sequence context model is unknown in advance. The combined method with the non-systematic split-LDPC codes inherits the advantages of each of the separate methods.