首页 | 本学科首页   官方微博 | 高级检索  
     

轨迹网络及其在字母识别中的应用
引用本文:彭汉川,沙利烽.轨迹网络及其在字母识别中的应用[J].电路与系统学报,1997,2(4):40-43.
作者姓名:彭汉川  沙利烽
作者单位:东南大学生物医学工程系,中国科学院自动化所模式识别国家重点实验室
基金项目:国家自然科学基金,江苏省青年科技基金,国家模式识别实验室开放课题
摘    要:本文提出一种新的轨迹神经网络模型,它利用输入模式时间序列形式模式类的轨迹,从而实现模式空域不变性特征提取。

关 键 词:模式序列  轨迹网络  字母识别  模式识别

The Trace Net and Its Applications in Alphabet Recognition
Abstract:A new Trace Neural Net(TNN)model finding spatial invariance through temporal pattern streams is proposed.The trace is recurrently interpreted as the weighted average of the trace at the previous time and the current pattern representation produced by the feature extractor.The optimization function is defined to be the sum of the squared error between the previous trace and the current pattern representation.The learning algorithm is derived to construct the concrete network.An improved TTN is also presented.Applications of these networks to Alphabet recognition imply better performance than the traditional Delta learning scheme.
Keywords:Pattern strearn  Trace  TNN  Alphabet recognition    
本文献已被 CNKI 维普 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号