Non-binary Distributed Arithmetic Coding

• Main Author: Ziyang, Wang
• Other Authors: Mao, Yongyi
• Language: en
• Published: Université d'Ottawa / University of Ottawa 2015
• Subjects: DAC
• Online Access: http://hdl.handle.net/10393/32318; http://dx.doi.org/10.20381/ruor-4323

Distributed source coding (DSC) is a fundamental concept in information theory. It refers to distributed compression of correlated but geographically separated sources. With the development of wireless sensor networks, DSC has attracted great research interest in the recent years [26]. Although many channel code based DSC schemes have been developed (e.g., those based on turbo codes [11]and LDPC codes [20]), this thesis focuses on the arithmetic coding based approaches, namely, Distributed Arithmetic Coding (DAC) due to its simplicity in encoding [8]. To date, most of the DAC approaches that have been proposed deal with binary sources and can not handle non-binary cases. Little research has been done to extend DAC for non-binary sources. This work aims at developing efficient DAC techniques for the compression of non-binary sources. The key idea of DAC is representing the source symbols by overlapping intervals, as opposed to the case of conventional arithmetic coding where the intervals representing the symbols do not overlap. However the design of the overlapping intervals has been completely of heuristic nature to date. As such, the first part of this work is a thorough study of various interval-overlapping rules in binary DAC so as to understand how these rules impact the performance of DAC. The insight acquired in this study is used in the second part of this work, where two DAC algorithms are proposed to compress non-binary non-uniform sources. The first algorithm applies a designed overlap structure in DAC process, while the second converts a non-binary sequence into a binary sequence by Huffman Coding and encoding the result in binary DAC. Simulation studies are performed to demonstrate the efficiencies of the two proposed algorithms in a variety of source parameter settings.