Feature detection from illustration of time-series data

被引：0

作者：

Takezawa, Tetsuya ^{[1
]}

Watanabe, Toyohide ^{[1
]}

机构：

[1] Nagoya Univ, Grad Sch Informat Sci, Dept Syst & Social Informat, Chikusa Ku, Nagoya, Aichi 4648603, Japan

来源：

GRAPHICS RECOGNITION: TEN YEARS REVIEW AND FUTURE PERSPECTIVES | 2006年 / 3926卷

关键词：

geometric feature; detection of reference sequence; time-series data;

D O I：

暂无

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

We propose a method for extracting the geometric feature and the comprehensive fluctuation from time-series data and also a method for detecting a reference sequence effectively on the basis of the distance graph. The prevalent methods such as one based on the frequency characteristics do not deal with time-series data in the time dimention. Therefore, our method for extracting the features is temporally sensitive to fluctuations of time-series data. We experimented using the time-series data whose frequency bands were changed variously in order to make clear the availability of the proposal procedures such as smoothing and encoding.

引用

页码：323 / 333

页数：11