基于傅立叶变换的多维数据可视化分类  

Visualization Classification Method of Multi-dimensional Data Based on Fourier Transform

作　　者：于家新[1] 刘文远[1] 李芳[1] 王宝文[1] 洪文学[2] 

机构地区：[1]燕山大学信息工程学院,秦皇岛066004 [2]燕山大学电气工程学院,秦皇岛066004

出　　处：《计算机工程》2008年第15期173-175,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(60671025;60474065)

摘　　要：基于雷达图表示多维数据的原理,提出一种利用傅立叶描述子识别雷达图形的可视化数据分类新方法。该方法采用多元统计中的雷达图表示多维数据,不同模式类别的多维数据构成不同形状的雷达图形。在此基础上对雷达图的边界曲线进行傅立叶变换,计算傅立叶描述子作为雷达图的图形特征,并运用改进的概率神经网络进行识别。实验结果表明该方法具有较好的分类效果,分类精度比传统分类方法提高了约8.25%。A novel method for designing the classifier of multi-dimensional data is proposed, which uses radar chart of multi-statistics to show multidimensional data and applies Fourier descriptors to recognize the radar chart. Different multi-dimensional data forms different radar chart and distinguishes different category. And Fourier descriptor takes as chart characteristic, by Fourier transform on the boundary curve of radar chart. Based on the improved Probabilistic Neural network（PNN）, radar chart is recognised automatically. Experimental results show this method has better result of classification, classification precision is out approximately 8,25 percentage, compared with the traditional classifier.

关 键 词：数据可视化 雷达图 傅立叶描述子 形状识别 概率神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]