Error-Pattern-Correcting Codes: A New Approach to Error Correction Coding for Interference-Dominant Channels

University Of Minnesota-Twin Cities, Minneapolis MN

Investigators

Abstract

Error-Pattern-Correcting Codes: A New Approach to Error Correction Coding for Interference-Dominant Channels Jaekyun Moon, University of Minnesota Abstract Intersymbol interference (ISI) is a major source of transmission errors in a wide variety of communications channels. This research concerns error correction for such channels. Input-constrained ISI channels are of specific interest in this research, as efficient coding methods for such channels are still largely an open area. This work focuses on 1) developing low-complexity error correction methods for input-constrained ISI channels, with emphasis on high-density data storage applications, and 2) clearing a path to approaching theoretical limits of error correction for such channels with fundamentally improved complexity/performance tradeoffs. One effective way of designing codes for input-constrained ISI channels is to target at dominant error cluster patterns inherent at the channel output, as the structure of the ISI typically give rises to a dominance of a relatively small set of distinct error cluster patterns, regardless of the channel signal-to-noise ratio. This set can easily be found using distance analysis or through empirical observation based on simulation or laboratory measurement. Focusing only on the dominant set, very high rate codes are possible with excellent overall error correction capability. Encoding and decoding are necessarily highly tailored to the few known error patterns. This type of code can be applied either as a stand-alone code or as a critical building block in a larger coding system, such as a turbo equalizer based on iterative soft decision processing. While the research addresses computationally efficient approaches, efforts are also directed to establishing a theoretical framework as well as to understanding theoretical performance limits of the channel-matched codes.