基于嵌套GA-BP算法的离线中文签名鉴定

Off-line Chinese signature verification based on GA-BP-Nesting algorithm

将BP算法嵌入GA中形成循环套,采用GA对BP算法搜索得到的极值点进行最优解选择。使用小波变换和伪Zernike矩相结合的特征提取方法进行签名鉴定。实验表明,与传统BP、分离GA-BP算法相比,嵌套GA-BP算法在提高鉴定系统准确率和减少网络训练步数两方面更具有优越性。

Back Propagation（BP） algorithm is embedded in Genetic Algorithm（GA） to form a nested loop,so that the optimum solution can be obtained from the extrema searched by GA.Features Extraction,a method in wavelet transformation is combined with Pseudo-Zernike,is used for signature verification.The results show that the GA-BP-Nesting algorithm can improve the accuracy of signature verification and reduce the training steps of the neural networks,being compared with the traditional BP,GA-BP-Aparting algorithm.