并行提取多个次成分的改进型Moller算法  

Modified Moller algorithm for multiple minor components extraction

作　　者：高迎彬 孔祥玉 胡昌华[1] 侯立安[1] 

机构地区：[1]第二炮兵工程大学控制工程系,西安710025 [2]第二炮兵驻石家庄地区军事代表室,石家庄050081

出　　处：《控制与决策》2017年第3期493-497,共5页Control and Decision

基　　金：国家杰出青年科学基金项目(61025014);国家自然科学基金项目(61673387,61374120,61074072)

摘　　要：次成分分析是信号处理领域一门重要的工具.然而,到目前为止能够进行多个次成分提取的算法并不多见,一些现存算法还存在很多限制条件.针对这些问题,采用加权矩阵的方法将M?ller算法扩展为多个次成分提取算法.该算法对于输入信号的特征值没有要求,而且在不需要模值限制措施的情况下,仍然具有很好的收敛性.仿真结果表明,该算法可并行提取多个次成分,而且收敛速度优于一些现有算法.Minor component analysis（MCA） is a powerful tool in the signal processing field.Up to now, there are few algorithms, which can extract multiple minor components from input signals, and many limitation conditions exist before using some existing algorithms.In order to solve these problems, the M？ller algorithm, which can only extract one minor component, is modified into a multiple minor components extraction algorithm by using the weighted matrix method.The proposed algorithm has no limitation on the smallest eigenvalue and has a good convergence property, while no norm operation is needed.Simulation results show that the proposed algorithm can efficiently extract the multiple minor components of an input signal and has a faster convergence speed than some existing algorithms.

关 键 词：多个次成分 Moller算法 加权矩阵 神经网络 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]