改进的HMM-SVM混合模型手写汉字签名认证方法

Improved Hybrid HMM-SVM Method for Handwritten Chinese Signature Verification

提出一种基于改进的HMM-SVM混合模型手写汉字签名认证方法.利用HMM对两类训练签名数据进行有区分性的特征变换及数据压缩.HMM的多维概率输出作为SVM模型的输入矢量.SVM的输出通过Sigmoid函数转化为后验概率以进一步提高认证效果.使用SVC2004数据库中的签名数据对该方法进行验证和分析,结果表明,相对于HMM模型和SVM方法以及HMM-SVM混合模型,该方法可以有效降低等错误率EER,获得了比较好的效果.

An improved hybrid HMM-SVM method for Handwritten Chinese Signature Verification is proposed.First HMM is used as discriminative feature transformation and data compression for two kinds of training data of signature.Then the multidimensional probability output vectors of HMM are treated as the input vectors for SVM.At last in order to improve the performance of verification,the output vectors of SVM are transformed to a posterior probability through sigmoid function.Some experiments with this method are made on the SVC2004 Signature database.The result indicates that the proposed method can get better performance such as lower EER compared to HMM,SVM and HMM-SVM.