A Method for Clustering Analysis of Super-long Discrete Signals and Its Application in Classification of Tidal Types
Lizhe Zhang · Haiyang kexue jinzhan · 2013
By establishing the mapping from a super-long discrete signal space Uto a low dimensional space V which is characterized by main spectrum and frequency,ageneralized distance in the super-long discrete signal space Uis 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.