一种适用于流数据分析的快速EMD算法  

A new fast EMD algorithm for streaming data analysis

作　　者：胡利萍[1] 宋恩亮[1] 李宝清[1] 袁晓兵[1] 

机构地区：[1]中国科学院上海微系统与信息技术研究所无线传感网实验室,上海200050

出　　处：《振动与冲击》2012年第8期116-120,共5页Journal of Vibration and Shock

基　　金：国家科技重大专项(2009ZX03006-001-02)资助

摘　　要：采用具有局域控制特性的B样条函数直接拟合均值线,采用局域标准差终止准则判断原型模态函数(Proto-mode Function,PMF)的对称性,得到一种适用于流数据分析的快速经验模态分解(Empirical Mode Decomposition,EMD)算法。相比经典EMD算法,该快速算法不仅时间复杂度和空间复杂度均有所下降,而且分解精度基本维持不变。该快速算法适用于战场侦察传感网等流数据分析场合,实时性好,能够得到与序列等长的连贯本征模态函数(IntrinsicMode Function,IMF)。A new fast empirical mode decomposition （EMD） algorithm was developed adopting moving mean averages of extrema with a combination of B splines and the local stopping criterion to estimate the symmetry of proto-mode function, it was suitable for streaming data analysis. Compared to classical EMD, both of time complexity and space complexity of the fast algorithm decreased markedly, but the decomposition precision retained. It was especially appropriate for streaming data analysis, such as, in occasion of war sensor networks, for its real time and continuous intrinsic mode functions acquired.

关 键 词：快速EMD B样条函数 局域终止准则 流数据 

分 类 号：TN911.6[电子电信—通信与信息系统]