论文部分内容阅读
提出一种用签名的分段差异值作为隐马尔可夫模型(HMM)观测值的在线签名认证应用方法.首先,采用双向后向合并DTW算法确定签名中关键点之间的对应关系.然后,采用经典DTW度量签名中各种细微的差异,用这些DTW差异值作为观测值训练HMM模型.将模型状态的意义定义为相似程度,将状态转移结构设定为全概率转移.在SVC2004签名数据库上,验证了该方法的有效性.
This paper proposes a method of signature verification using signature subsection difference as hidden Markov model (HMM) observations.Firstly, the two-way backward combination DTW algorithm is used to determine the correspondence between the key points in the signature.And then, Using the subtle discrepancies in the classic DTW metric signatures, these DTW difference values are used as the training value to train the HMM model.The meaning of the model state is defined as the degree of similarity, and the state transition structure is set as the full probability transfer.On the SVC2004 signature database , Verify the effectiveness of the method.