摘要
税务、金融等经济领域的手写体数字信息通过计算机进行自动识别处理,可以节省人力、物力和财力,具有较高的实用价值。介绍概率神经网络的基本原理,并将概率神经网络应用于手写体数字识别中,在一定的训练样本和网络扩散速度情况下,实现基于概率神经网络的手写体数字识别。通过MATLAB对MNIST手写体数据库数据进行仿真实验验证,结果表明概率神经网络在手写体数字识别中能够取得较高的识别率,使用的算法可行有效。
Handwritten numeral recognition deals with the information of taxation,finance and other fields through computer or other machines for processing,makes it possible to save manpower and financial resources,with higher practical value.Although the type of identification number is not much,the required accuracy is very strict.Introduces the basic principle of probabilistic neural network,applies probabilistic neural network to handwritten digit recognition to select the best network diffusion speed and the number of training samples,and realizes the digital identification based on probabilistic neural network.MNIST handwritten database through MATLAB simulation experiment,the results show that the algorithm has high recognition rate,which is feasible and effective.
基金
广东省科技计划工业高新技术领域攻关项目(No.2013B010401032)
关键词
概率神经网络
手写体数字识别
贝叶斯决策理论
图像识别
Handwritten Digit Recognition
Probabilistic Neural Networks
Bayesian Decision Theory
Image Recognition
