基于HMM的离线签名识别

Recognizing the Pen Trajectories of Static Signatures Using Hidden Markov Model

给出一种以隐马尔可夫(HMM)模型为基础的离线签名的识别方法,此方法把离线签名图像中的所有汉字作为一个整体,首先利用图像处理技术,把整个字体区域分割出来,再统计每一行字体部分的像素点数。利用隐马尔可夫模型来对这个整体进行建模;然后利用Baum-Welch算法对模型进行训练;最后,利用已经训练好的HMM模型对一些签名图片进行识别。试验表明,识别率可达95.7%,为离线签名识别系统的进一步应用奠定了基础。

The writer gives an advance recognition algorithm about static signature based on the hidden Markov model. This algorithm treated all word in a static signature image as one,then to extract the region of the word by using the technology of image processing, model this whole using HMM. Then train this model by Baum-Welch algorithm. Finally, recognize some static signature using trained model. This simulation results show that the recognition rate can reach 95.7%.establish the basic of the system of recognizing static signature.