样条权函数神经网络克服了很多传统神经网络(如BP、RBF)的缺点：比如局部极小、收敛速度慢等。样条权函数神经网络的拓扑结构简单，训练后的神经网络的权值是输入样本的函数，能够精确记忆训练过的样本，可以很好地反映样本的信息特征，亦可以求得全局最小值。为了克服传统网络在指纹识别中的弊端，文中利用了样条权函数神经网络的优点，介绍了其在指纹识别中的应用。首先通过主成分分析方法对指纹图像进行特征提取，然后利用样条权函数神经网络进行指纹识别，最后通过Matlab仿真与其他传统的神经网络进行比较，验证了样条权函数在指纹识别方面的可行性且比传统神经网络效率更高。%Spline weight function neural network overcomes many defects of traditional neural networks (like BP,RBF),such as local minima,slow convergence. The topology structure of Spline weight function neural network is very simple, the trained neural network weights are the function of input samples,so it can remember trained samples and accurately reflect the characteristics of the sample infor-mation,and also can be obtained global minimum. In order to overcome the traditional networks' shortcomings in fingerprint identifica-tion,introduce the application in fingerprint recognition with the advantages of the spline weight function neural networks. Firstly extract the feature of the fingerprint images through principal component analysis,and then use the spline weight function neural network to do the fingerprint recognition,finally compare the spline weight function neural network and other traditional neural networks through Matlab simulation to verify the feasibility of spline weight function in fingerprint recognition and it is more efficient than the traditional neural networks.
展开▼
