奇异值分解去噪中有效秩阶次的自身辅助确定  

Self-assisting determination of effective rank degree in SVD denoising

作　　者：查翔[1] 倪世宏[1] 谢川[1] 张鹏[1] 

机构地区：[1]空军工程大学航空航天工程学院,西安710038

出　　处：《计算机应用研究》2016年第5期1359-1362,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(61372167;61379104)

摘　　要：为解决奇异值分解(singular value decomposition,SVD)去噪中有效秩阶次难以确定的问题,提出一种利用SVD本身进行辅助确定的方法。充分借助SVD在奇异性检测中表现出的优良特性,将原始含噪信号进行SVD处理后获得的奇异值序列视做一个新的信号,并对该信号对象重新进行奇异值分解;通过对各分量信号的奇异性检测,将奇异值序列的奇异点位置作为有效秩阶次确定的依据。对一个仿真实例信号的实验结果表明,该方法可准确地确定出有效秩阶次,从而能够实现信号的有效去噪。To determine the effective order rank in SVD denoising, this paper proposed a SVD self-assisting method. Fully con- sidering the excellent performance in singularity detection, it regarded !he singular value sequence of original signal by SVD processing as a new kind of signal, and decomposed the new signal object using SVD again. The singular points position of new singular value sequence acted as the criterion of determining effective rank degree through singularity detection of each branch signal. Experiment results of a simulated signal show that the proposed method can accurately select effective rank degree and then achieve effective denoising.

关 键 词：奇异值分解 信号去噪 HANKEL矩阵 奇异性检测 有效秩 

分 类 号：TN911.7[电子电信—通信与信息系统]