On LDPC decoding with natural redundancy

Big-data storage has substantial challenges due to accumulative noise in storage media. To ensure its long-term reliability, new techniques for error correction are being explored. This paper studies how to discover natural redundancy in data and use it for error correction. It explores the combination of natural redundancy decoding with low-density parity-check (LDPC) codes for enhanced error-correction performance. It derives analytical equations for the density evolution of LDPC decoding given information from natural-redundancy decoders. It proposes a theoretical model for compressed languages, and studies the performance of iterative decoding between the LDPC decoder and the natural-redundancy decoder. It also presents an upper bound to the code sizes of error-correcting codes given the assistance from natural-redundancy decoders.