论文部分内容阅读
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法有下述优点：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％．
The multi-branch second-order pseudo-two-dimensional hidden Markov model is used to recognize handwritten digits. The multi-branch second-order pseudo-two-dimensional Viterbi algorithm is introduced to train the model. The training algorithm has the following advantages over the traditional forward-backward algorithm: 1. Avoid underflow problems. 2. Operation accuracy is greatly improved. 3. Operation speed. Multi-branch model can be more fully described model features, the recognition rate is higher than the ordinary pseudo-two-dimensional hidden Markov model. Test results Training set recognition rate of 96.1%, 94.95% of the test set.