Research on Dependable Techniques for Compressed Text Data

这项研究提出了可靠的无损压缩数据技术，保护解压后的数据不受压缩数据中错误的影响。(1)霍夫曼编码：提出了一种能够双向解码的霍夫曼编码，该编码不仅可以从压缩数据的顶部解压，而且可以从压缩数据的尾部解压。通过比较前向和后向解压缩得到的解压缩结果，得到错误位置的候选位置。(2)算术编码：将压缩数据分成固定长度的块。解压缩过程中的参数被认为是解压缩器的内部状态，并包含在每个块的开头。这些内部状态用于错误检测和从错误状态恢复。针对ZIV-Lempel编码，提出了一种基于不等差错保护(UEP)的纠错方法。压缩数据包括对错误比其他部分更敏感的部分。已提出对错误敏感部分使用附加检查信息的嵌套式错误控制。对于LZ77编码和霍夫曼编码的两级压缩，提出了分离压缩方法。LZSS编码是LZ77编码的一种变体，其压缩数据被分成符号和整数两个集合，并分别用霍夫曼编码进行压缩。通过比特重排，每个码字的顶部位置固定。(5)基于分组模式匹配的压缩：中间码字的重要部分采用差错控制编码，其余部分采用具有差错恢复能力的熵编码进行压缩。

This research has proposed dependable techniques for lossless compressed data which protect the decompressed data from errors occurring in the compressed data. The following presents the proposed techniques.(1)For Huffman coding:Huffman coding capable of bidirectional decoding, which can be decompressed not only from the top but also from the tail of the compressed data, is proposed. By comparing decompression results obtained by forward and backward decompression, the candidates of error locations are obtained.(2)For arithmetic coding:Compressed data are divided into fixed-length blocks. Parameters in decompression are regarded as internal states of the decompressor and are included into the beginning of each block. These internal states are used in both error detection and recovery from the erroneous state. Recovery method for the erroneous block by additional check information has also been proposed.(3)For Ziv-Lempel coding:The method based on Unequal Error Protection, or UEP has been proposed. Compressed data include a part which is more sensitive to the errors than others. Nested error control which uses additional check information for the error sensitive part has been proposed. The separated error control coding where error sensitive part is encoded-separately has also been proposed.(4)For 2-level compression by LZ77 coding and Huffman coding:Separated compression method has been proposed. The compressed data of LZSS coding, a variation of LZ77 coding, are divided into two sets, sets of symbols and integers, and are compressed by Huffman coding separately. By bit-rearrangement, the position of the top of each codeword is fixed.(5)For compression based on group pattern matching:Important part of the intermediate code word is encoded by error control codes and other part is compressed by entropy coding with error recovery capability.