一种新的手写体数字识别方法

应用多分支二阶伪二维隐马尔可夫模型进行了手写体数字的识别，引入了多分支二阶伪二维维特比算法对模型进行训练，该训练算法比传统的前后向算有下述优点：１，避免了下溢出问题。２．运算精度大大提高．３。运算速度加快。采用多分支模型可以更充分描述模型特征，其识别率高于普通伪二维隐马尔可夫模型，试验结果训练集识别率９６．１％、测试集９４．９５％。

A New Method for Handwriting Digital Recognition

A new method of two order Multi Branch two Dimensional Pseudo Hidden Markov Models(MB 2D PHMM) been used for handwriting digital recognition. Two order multi branch 2D Pseudo viterbi algorithm used for training model.This algorithm compared to traditional forward backward algorithm has some advantage as follow. 1) It can get rig of down overflow. 2) The accurate of the calculation is higher. 3) The speed of calculation is much faster. As two order multi branch 2D Pseudo Hidden Markov Model can represent the characteristic of the pattern much properly. The recognition rate is higher than conventional 2D Pseudo Hidden Markov Model. The result of the test shows, the recognition rate of training set is 96.1%. The recognition rate of testing set is 94.95%.