Analyses of coding and compression strategies for data storage and transmission

Citation

Sayano, Masahiro (1992) Analyses of coding and compression strategies for data storage and transmission. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/xxh5-v758. https://resolver.caltech.edu/CaltechETD:etd-08132007-084323

Abstract

Selected topics in error correction coding and data compression for data storage and transmission will be analyzed here. In particular, a model for the mean time to failure for computer memories protected by error correction coding, characteristics and applications of phased burst error correcting array codes, and locally adaptive vector quantization for image and data compression will be examined.

A model of the mean time to failure (MTTF) of semiconductor random access memories protected by single error correcting-double error detecting (SEC-DED) codes on the chip and with soft error scrubbing and multiple types of hard failures will be presented. Only a few assumptions and approximations will be made. This model will provide a more complete picture of the expected failure modes, reliability, and the mean time to failure of memory systems protected by on-chip error correction coding. Special cases will also be addressed, such as slow or fast scrubbing and dominance of hard or soft errors.

Characteristics of a family of phased burst error correcting array codes will be addressed. In particular, allowable and optimal code sizes will be examined. When used in non-binary applications, these codes retain their characteristics and can correct "approximate" errors with even higher rate: If the amount any q-ary symbol can be in error is bounded by some value, these codes can be designed to address this type of error with even fewer check symbols.

Improvements to a locally adaptive vector quantization compression strategy will be discussed. The basic strategy involves reorganization of the code book after each use so that the most recent codewords are moved to the front. With the various improvements covered in this work, the algorithm is capable of matching the performance of other more computationally intensive algorithms at a fraction of the computational complexity.

Thesis Files