一种冗余字典下信号稀疏分解新方法.pdfVIP

2008 4 ( ) Apr. 2008 35 2 J OUR NAL OF XIDIAN UNIV ER SITY Vol . 35 No. 2 刘 丹华, 石光明, 周佳社 (西安电子科技大学 电子工程学院, 陕西 西安 7 10071) : 针对目前冗余字典下信号稀疏分解常用算法计算复杂 高的问题, 提出一种分组匹配追踪算法. 该算法首先利用多组正交基构造冗余字典, 然后采用迭代式分组匹配追踪, 每次迭代从字典中选出 一组 和原始信号或残余最匹配的正交基, 采用正交分解快速算法进行正交分解得到少量重要系数, 多次迭代 后逐渐稀疏逼近原始信号. 实验结果表明, 基于小波正交基级联冗余字典进行信号稀疏分解时, 在同等 稀疏条件下, 与匹配追踪( MP) 算法相比, 该算法的计算速 提高了大约30 倍, 而且可避免过匹配现象. : 稀疏分解; 冗余字典; 匹配追踪算法; 信号压缩 : T N9 11. 72 : A : 1001-2400( 2008) 02-0228-05 New method for signal sparse decomposition over a redundant dictionary LI UDan-hua, SHI Guang-ming, ZH O UJia-she ( School of El ctronic Engin ring , Xidian Univ . , X ian 71007 1, China) Abstract: For th xtr m ly high compl xity of usual algorithms for spars d composition, a n w group matching pursuit algorithm is pr s nt d bas d on a r dundant dictionary with s v ral orthonormal bas s. Th algorithm adopts th th id a of it rativ group matching pursuit and s l cts th optimal basis from th dictionary by comparing th matching d gr b tw n th signal or th r sidua and v ry basis. Each op ration of d composing r sults in a f w important co ffici nts by using th fast calculating algorithm of orthogonal d composition. Aft r s v ral such it rations, th original signal is approximat d with a f w co ffici nts v ntually. Simulation r sults show that th calculating sp d of th algorithm in this pap r incr as s by about thirty tim s compar d w ith MPs. Mor ov r , this algorithm can avoid ov r-matching. Key Words: spars d composition;