基于稀疏信号的学习字典算法及其应用

发布时间：2018-01-13 16:23

本文关键词：基于稀疏信号的学习字典算法及其应用　出处：《北京工业大学》2014年硕士论文　论文类型：学位论文

【摘要】：本文主要研究用于稀疏表示的字典学习算法。随着压缩感知理论的快速发展，信号的稀疏表示问题成为近些年的研究热点。冗余字典具有很好的表示能力，可以对于复杂信号进行稀疏编码。而通过学习得到的冗余字典可以自适应地对于一类信号进行更稀疏的表示，具有一定的研究价值。本文对得到学习字典的算法进行了研究，发现目前常用的字典学习算法在前期的稀疏表示阶段浪费了计算复杂度，且可能会过编码，使结果无法达到最优。针对上述问题，提出了全局支撑交错追踪K-SVD算法，并对其进行了改进，，得到交错追踪先导K-SVD算法。最后本文对于提出的算法对于图像去噪进行了实验，检验了本文提出算法的优越性。
[Abstract]:This paper focuses on dictionary learning algorithms for sparse representation. With the rapid development of compressed sensing theory, the sparse representation of signals has become a hot topic in recent years. Redundant dictionaries have a good representation ability. It can be used for sparse coding of complex signals, and the redundant dictionaries can be used to represent a class of signals more sparsely adaptively, which has some research value. In this paper, the algorithm of learning dictionary is studied. It is found that the current commonly used dictionary learning algorithms waste computational complexity in the sparse representation stage, and may be overcoded, so that the results can not be optimized. To solve the above problems, a global support staggered tracking K-SVD algorithm is proposed and improved. The K-SVD algorithm with staggered tracking pilot is obtained. Finally, the proposed algorithm is tested and the superiority of the proposed algorithm is tested.
【学位授予单位】：北京工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.7

【参考文献】