摘要： 提出了一些新的方法来解决联机英文手写识别问题。基于手写英文的主笔画属性,提出了快速有效的解码算法——主笔画层次建立BeamViterbi算法。由于整个识别过程是一个多阶段决策的过程,将多阶段的分类信息融合到决策过程中,较大幅度提高了系统的性能。在Unipen训练集和实验室数据集上测试取得了良好的效果。

关键词: 自由手写识别, 层次建立, 多阶段识别, 解码算法

Abstract: Some new methods are proposed for on-line cursive handwriting recognition.Based on the special primitive attribute of handwriting cursive English,the paper represents a fast valid decoding algorithm,that is,Primitive Level Building Beam Viterbi(PLBBV).Since the decoding process is a multi-phase decision-making process,the paper represents a combining classifiers algorithm that uses all the classifiers’ measurement to make decision.The tests on the Unipen training set and the laboratory set reveal the feasibility of the proposed method.

中图分类号: 

• TP391.43