中文摘要：

在研究基于隐马尔可夫模型的识别器和基于距离分类器的识别器的识别结果基础上，提出两种基于集成神经网络的手写识别系统：比较神经网络识别系统和全排列神经网络识别系统。实验分析表明，该系统对西文手写体的识别率最高可达到99％，比单独使用原始识别器的识别率提高10个百分点，达到了良好的识别效果。

英文摘要：

According to the study result of the recognizers based on Hidden Markova Model and on distance classification respectively,in this paper we introduce two handwriting recognition systems based on neural network ensemble, the comparison neural network recognition system and the permutation neural network recognition system. Experimental analyses express that the recognition accuracy of the system on handwriting of western languages could be up to 99%, i. e. 10 percentage points higher than just using the primary recognizer, which achieves fairly good recognition effect.