基于CS算法的数字信号压缩和高效数字系统设计的研究

项目名称： 基于CS算法的数字信号压缩和高效数字系统设计的研究

项目编号： No.61273195

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 李刚

作者单位： 浙江工业大学

项目金额： 80万元

中文摘要： 以压缩感知(Compressed Sensing-CS)理论为指导, 研究数字信号的高效压缩和数字系统的高效实现。信号的稀疏表示是CS的必备条件, 构造高性能字典的根本是使其与信号更好的"匹配"，而这却是现有信号字典无法做到的。以信号建模、系统结构理论和优化算法为工具，提出了信号自适应字典(Signal Adapted Dictionary-SAD)概念和一种基于SAD的多级压缩系统(SAD-m-CS)的理论框架和实现方法; 结合信号稀疏和系统结构稀疏两个概念，提出了基于模块结构的高效数字系统设计的新型准则和实现方案；基于CS重构算法和滤波器算子(字典)概念，提出了一种IIR数字滤波器设计的新方法，以有效地克服现有滤波器设计方法中存在的非凸性和稳定性两大难点。作为应用研究：拟采用SAD-m-CS系统，研发低速率语音编码器；研究移动通信系统中RRC成型-匹配滤波器高效实现并建立设计平台。

中文关键词： 压缩感知；优化；信号/数据压缩；数字滤波器结构；有限字长效应

英文摘要： In this project, digital signal compression with high rate and implementation of digital systems with high performance are investigated based on the theory of compressed sensing (CS). As well known, signal sparse representation is the key to the success of the CS techniques and high performance dictionaries need to adapt the signal characteristics. The latter is hardly met by the current dictionary design strategies. With help of the theories of signal modeling, system structures and optimization techniques, a digital signal compression system based on a proposed concept, referred to as signal adapted dictionary (SAD), is investigated. New methods for designing digital systems as well as their efficient implementation are proposed using the concepts of sparsity and structure-modality. Taking advantages of the existing CS signal construction algorithms, a new strategy for designing IIR digital filters is outlined, which is targetted to overcome the two major difficulties existing in the current design methods: non-convexity and instability. As applications, it is aimed at developing low-bit-rate speech coders using the proposed SAD-m-CS framework and setting up a design platform for efficient implementation of root-raised-cosine(RRC) matched digital filters used in communications systems.

英文关键词： compressed sensing；optimization；signal/data compression；digital filter structure；finite wordlength effect