基于短时付里叶变换的目标识别方法  

A TARGET RECOGNITION METHOD BASED ON SHORT-TIME FOURIER TRANSFORM

作　　者：王蕴红[1] 刘国岁[1] 王一丁 

机构地区：[1]南京理工大学电子工程系,210094

出　　处：《模式识别与人工智能》1998年第2期206-210,共5页Pattern Recognition and Artificial Intelligence

基　　金：国家博士点基金

摘　　要：本文研究了短时付里叶变换后再提取主元作为特征矢量的特征提取方法,并对提取出的特征利用神经网络进行了识别.从理论上证明了该方法的合理性,说明该方法是一种良好的特征提取方法,通过计算基于散布矩阵可分性测度结果得出该方法提取的特征比一般付里叶变换法提取的特征具有更好的可分性.利用此方法提取出的目标特征及付里叶交换法提取出的目标特征分别进行识别实验,结果表明短时付里叶交换后再提取主元的特征抽取方法可以获得更好的识别结果.A novel feature extraction method based on STFT(short-time fourier transform) and PCA(principal component analysis) is proposed in this paper. We use a evaluating function to evaluate the features and compare this method to Fourier Transform. A BP network is used as classifier to classify the features extracted by the two methods. The theoretical and experimental results have proved that the features extracted by the way proposed in this paper has better distinctive ability. The better classification results can be achieved. This extraction method is promising in the application of pattern recognition.

关 键 词：短时付里叶变换 特征提取 目标识别 模式识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]