权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Analysis on the codeword generating process of an arithmetic code and its application to source distribution transformer

算术码码字生成过程分析及其在源配变中的应用

基本信息

批准号：
16560326
负责人：
MORITA Hiroyoshi
金额：
$ 1.34万
依托单位：
The University of Electro-Comunicat ions
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
2004
资助国家：
日本
起止时间：
2004 至 2005
项目状态：
已结题

项目摘要

Throughout this project, we have obtained the following results on the codeword generating process of an arithmetic code.2004-We evaluated theoretically, in the average sense, the redundancy included in a codeword generated by an arithmetic encoder. Moreover, we developed an algorithm to transform a sequence of information source symbols into another sequence with a given target distribution. This algorithm dissolve the following two practical problems that result from arithmetic codes:1) when the decoding process halt.2) how to reproduce exactly the original sequence from the transformed one.2005-We considered the encoding/decoding processes for an information source that outputs symbols with a random interval and studied on the number of code symbols, that is, a fragment of a codeword, that an arithmetic encoder outputs at a unit time and the interval that it outputs them. From computer experiments, the interval of fragments can be modeled by geometric distribution while the length of each fragment has a different one. We observed the similar results in the decoder side. That is, the arithmetic decoder outputs a fragment of a codeword with a geometric distribution on its interval. However, our experiments showed that the length of fragment is neither individual nor geometrically distributed. Finally, we investigate the delay of the coding process and evaluated the probability distribution of the delay for a memoryless binary source.

在整个项目中，我们在算术编码的码字生成过程中获得了以下结果。2004-我们从平均意义上从理论上评估了算术编码器生成的码字中包含的冗余度。此外，我们还开发了一种算法，将信息源符号序列转换为具有给定目标分布的另一序列。该算法解决了算术编码产生的两个实际问题：1)当译码过程停止时；2)如何从变换后的序列准确地再现原始序列。2005-我们考虑了以随机间隔输出符号的信源的编解码过程，并研究了算术编码器在单位时间输出的码元的数目，即码字的片段及其输出的间隔。从计算机实验来看，碎片的间隔可以用几何分布来模拟，而每个碎片的长度是不同的。我们在解码端观察到了类似的结果。也就是说，算术解码器输出在其间隔上具有几何分布的码字的片段。然而，我们的实验表明，片段的长度既不是单个的，也不是几何分布的。最后，我们研究了编码过程的时延，并估计了无记忆二进制源的时延的概率分布。