中文说明：自动识别手写的文字如信件、 手稿，或整本书的几个几十年来一直深入研究的一个重点。它值得提及谷歌和雅虎已经宣布他们打算让手-可通过其搜索引擎访问的书。因此在这种情况下，关键字发现将是一个宝贵的工具，用于浏览这些书籍的内容。手写的字或词是更难的承认，因为不同由多个作家的作品。成功的字词识别手写文档如果数据库的大小用来训练神经网络取决于数据库的大小，神经网络是更大的成功识别概率则更多。但更大数据库放识别过程的速度限制。所以的这 disserta-的目的对工作是实现系统与数据库大小之间的完美匹配和识别的速度。
English Description：
The automatic recognition of handwritten text-such as letters, manuscripts, orentire books-has been a focus of intensive research for several decades.It is worthmentioning that Google and Yahoo have announced their intention to make hand-written books accessible through their search engines. Hence in this context, keywordspotting will be a valuable tool for browsing the contents of these books.Handwritten characters or words are more difficult for recognition due to diversewritings by multiple writers. Success of handwritten document for word recognitionusing neural network depends on size of database, if size of database used to trainneural network is larger then probability of successful recognition is more. But largerdatabase put limitation on speed of recognition process. So purpose of this disserta-tion work is to implement system with perfect match between size of database andspeed of recognition.