结合小波变换(WT)、二维主元分析(2DPCA)和椭球基函数(EBF)特点,提出了一种基于WT、2DPCA和EBF神经网络指纹识别方法。利用小波变换将原始图像分解为高频分量和低频分量,并忽略水平高频与垂直高频分量,获得原始图像的基本特征。再通过2DPCA算法对该图像进行降维,获取降维特征;最后结合椭球基函数神经网络(EBFNN)完成指纹识别。本算法将2DPCA优化的特征提取与EBFNN的自适应性相结合,在FVC2000(国际指纹竞赛数据库)上做了测试,总的正确识别率可达91.4%,具有一定的实用价值。与WT-PNN算法和WT-2DPCA-RBF算法进行比较,结果表明,本文提出的算法在平移、旋转及光照变化的指纹数据库上的识别效果优于WT-PNN算法和WT-2DPCA-RBF算法。

In combination with Wavelet Transform(WT),Two-dimensional Principal Component Analysis(2DPCA) and Ellipsoidal Basis Function(EBF),a fingerprint recognition algorithm based on WT,2DPCA and EBF neural network(EBFNN) is proposed.Original images are decomposed into high-frequency and low-frequency components with WT,and horizontal and vertical high-frequency components are ignored,so the prime features of original images can be obtained;then,the projected features are solved by 2DPCA;finally,fingerprint recognition can be realized by EBFNN.The algorithm combines the optimization of 2DPCA and the adaptability of EBFNN and achieves the accurate recognition rate of 91.4%.The experimental results based on FVC2000 verify that proposed algorithm has higher recognition rate than that of WT-PNN and WT-2DPCA-RBF.