超长离散信号聚类方法研究及其在潮型分类中的应用  

A Method for Clustering Analysis of Super-long Discrete Signals and Its Application in Classification of Tidal Types

作　　者：张立振[1] 玄春艳[1] 曹露洁[2] 

机构地区：[1]中国海洋大学数学科学学院,山东青岛266100 [2]中国海洋大学工程学院,山东青岛266100

出　　处：《海洋科学进展》2013年第3期313-318,共6页Advances in Marine Science

基　　金：国家自然科学基金--海洋环境下结构焊接点多轴疲劳损伤演化机理及演化率(51175485);山东省研究生创新计划--侧重海洋应用的数学系研究生培养的实践与认识(SDYY12152);国家海洋环境预报中心重点实验室开放基金--台风影响下极值波高;极值水位联合概率分析新模式(LOMF1101)

摘　　要：通过建立超长离散信号空间到以主频谱和主频率为特征的低维空间上的映射,诱导出了超长离散信号空间上的广义距离,在此基础上,建立了一种针对超长离散信号进行聚类的方法。与传统聚类方法相比,新方法可使计算量大为减少。同时,还可根据具体问题灵活调整广义距离中的参数,实现具有针对性的聚类。文中还将新方法应用于世界各验潮站海面波动信号的聚类,绘出了全球潮型分布图。By establishing the mapping from a super-long discrete signal space U to a low dimensional space V which is characterized by main spectrum and frequency, a generalized distance in the super-long discrete signal space U is defined. Based on this, a new method has been established for clustering the super-long discrete signal space U. Comparing with traditional clustering methods, the new method can reduce the computation significantly. In addition, the parameters of the generalized distance can flexibly be adjusted according to concrete issues, hence realizing the pertinence of clustering. This new method is applied to the clustering of sea level data sets from the tide stations all over the world, and a global distribution of tidal types is plotted.

关 键 词：超长离散信号 傅里叶变换 广义距离 聚类分析 

分 类 号：O212.4[理学—概率论与数理统计]