CTH08-6: Context Based Decoding of Split-LDPC Codes

We consider channel decoding of redundant data with context based memory with the recently proposed class of nonsystematic split-LDPC codes. Methods based on denoising techniques and context estimation from universal source coding are used to extract redundancy between decoding iterations of split-LDPC codes. Simulation results show that the use of nonsystematic split-LDPC codes gains in most cases on standard systematic LDPC codes even without prior knowledge of the source statistics. Different context extraction methods are considered, including a novel approach that is used to capture different types of context based redundancy.