概率神经网络在识别带有噪声字母中的应用

字符识别是模式识别中的一个应用，通过训练网络可以教会计算机如何识别字符，这在票据处理方面可以大大地提高效率。该文中所建立的神经网络为具有局部响应的高斯函数的三层概率神经网络，它以牢固的插值理论为基础，具有学习速度快，不易陷入局部极小等优点。本文介绍了概率神经网络的学习算法和一个三层概率神经网络对带有噪声的26个英文大写字母的识别。其中利用MATLAB编写仿真程序对概率神经网络进行训练，仿真结果表明，训练的概率神经网络可以对给定的带有噪声的字母作出正确的识别。

Application of Probabilistic Neural Network in Identification of English Letters with Noise

Character identification is an application in pattern identification. Computer could learn how to identify character by training network to greatly improve efficiency in processing coupon. The neural network is a tri -level probabilistic neural network with Gauss function of partial response on the basis of firm interpolation concept and has the advantage of fast learning and difficult getting into local minimum. This introduces learning ＆ calculative method of probabilistic neural network and identification of tri - level probabilistic neural network to 26 English capital letters with noise. MATLAB is used to compile simulation programme for training probabilistic neural network so that trained probabilistic neural network could correctly identify specific letters with noise as a result of simulation.