一种不确定性二叉树SVM的脱机手写体汉字识别研究

针对脱机手写体汉字特征复杂和类别多样的特点,基于SVM数学模型,采用了一种不确定性二叉树与SVM相结合的分类识别方法设计了一种多类分类器,该设计方法在保证识别准确率的情况下大大减少了支持向量机的数量,简化了二叉树模型,能快速辨识并删除多余的枝节,并具有一定的容错率,加快了辨识速度。实验结果表明,采用不确定性二叉树SVM设计的多类分类器有效地降低了拒识率和漏识率,保证了识别的准确率,提高了识别速度。

Studying on the Off-line Handwritten Chinese Characters Recognition Based on Uncertainty Binary Tree SVM

By adopting the classification and recognition method,a multi-class classifier based on SVM model is established due to the characteristics of handwritten Chinese character,such as the complexity of characteristics and diversity of category.The design method with fault-tolerant rate can reduce number of support vector machine,improve the recognition speed,simplify binary tree model,quickly identify and remove the extra minor under the condition of the accuracy of the identification.The experimental results show that the multi-class classifier based on uncertainty binary tree SVM can reduce the rejection rate and leakage identification rate effectively to ensure the recognition accuracy and improve the recognition speed.