神经网络在模式识别中的简单分析及应用

论文编号:TX269  论文字数:16622,页数:47

摘  要

 模式识别就是机器识别、计算机识别或者机器自动化识别，目的在于让机器自动识别事物，使机器能做以前只能由人类才能做的事，具备人所具有的对各种事物与现象进行分析、描述与判断的部分能力。它研究的目的就是利用计算机对物理对象进行分类，在错误概率最小的条件下，使识别的结果尽量与客观事物相符合。
 随着人们对人工神经网络的不断地认识，神经网络是指用大量的简单计算单元构成的非线性系统,它在一定程度和层次上模仿了人脑神经系统的信息处理、存储及检索功能，因而具有学习、记忆和计算等智能处理功能。这样人们利用人工神经网络具有高度的并行性，高度的非线性全局作用以及良好的容错性与联想记忆功能，并且具有良好的自适应、自学习功能等突出特点，可运用MATLAB神经网络工具箱中的神经网络模型，对经过训练的神经网络可以有效地提取信号、语音、图像等感知模式的特征，并能解决现有启发式模式识别系统不能很好解决的不变量探测、抽象和概括等问题。这样神经网络可应用于模式识别的特征提取、聚类分析、边缘检测、信号增强以及噪声抑制、数据压缩等各个环节。使用机器来进行模式的识别是一项非常有用的工作，能够辨别符号等系列的机器是很有价值的。目前，模式识别技术可以应用指纹识别、IC卡技术应用、字符识别等实例。模式识别成为人工神经网络特别适宜求解的一类问题。因此，神经网络技术在模式识别中也得到广泛应用与发展。
 
关键词：模式识别；人工神经网络；神经网络模型；神经网络技术

Abstract

 Pattern Recognition is the machine identification, computer identification or identification of machine automation, machine aimed at automatic identification of things to do before the machine can only be made by man can do, with people with all kinds of things and on an analysis of the phenomenon, described with the ability to determine the part. It is the purpose of the study of the physical object to use the computer for classification, the probability of the smallest in the wrong conditions, so that the results of recognition as far as possible in line with objective things.
 As artificial neural network to recognize the continuing, neural network refers to a large number of simple calculation unit consisting of non-linear system, which to some extent and level system to imitate the human brain's information processing, storage and retrieval functions, which has learning, memory and computing functions such as intelligent processing. Such people to use artificial neural network with a high degree of parallelism, the overall role of a high degree of non-linear and good fault tolerance and associative memory function, and have good self-adaptive, self-learning function, such as prominent features, the availability of MATLAB neural network toolbox The neural network model trained neural network can effectively extract the signal, voice, video and other features of perceptual patterns and heuristics to solve the existing pattern recognition systems are not well resolved invariant detection, such as abstract and summary issues. This neural network pattern recognition can be applied to feature extraction, clustering analysis, edge detection, signal enhancement and noise suppression, data compression, such as various links. The use of machines for pattern recognition is a very useful work, such as series of symbols to identify the machines are of great value. At present, the pattern recognition technology can be applied to fingerprint identification, IC card technology applications, such as examples of character recognition. Artificial neural network pattern recognition has become especially suitable for solving a class of problem. Therefore, the neural network pattern recognition technology is also widely used and development.
 
Key words：pattern recognition；artificial neural network；neural network model；neural network technology
目  录
引言 - 1 -
1  模式识别概述 - 1 -
1.1 模式识别基本概念 - 1 -
1.2 模式识别系统 - 2 -
1.3 模式识别的主要方法 - 2 -
1.4 模式识别应用 - 3 -
2  人工神经网络概述 - 4 -
2.1 人工神经元模型 - 4 -
2.1.1阀值函数、Sigmoid函数和分线段性函数 - 5 -
2.2 人工神经网络模型 - 5 -
2.3 神经网络学习特点 - 6 -
2.4 人工神经网络在模式识别问题中应用优势 - 6 -
3  神经网络模式识别 - 7 -
3.1 基于BP神经网络的模式识别 - 7 -
3.1.1BP神经网络模型简述 - 7 -
3.1.2BP学习算法 - 8 -
3.1.3BP神经网络应用于字符识别 - 9 -
3.2 基于径向基函数神经网络的模式识别 - 16 -
3.2.1径向基函数神经网络模型简述 - 17 -
3.2.2径向基函数神经网络应用特征模式的分类 - 18 -
3.3 基于自组织竞争神经网络的模式识别 - 19 -
3.3.1自组织竞争神经网络基本思想 - 20 -
3.3.2自组织竞争神经网络设计模式分类器的应用 - 20 -
3.4 基于反馈型神经网络的模式识别 - 24 -
3.4.1反馈型神经网络模型简述 - 24 -
3.4.2反馈型神经网络在工程领域的应用 - 24 -
4  实验分析与总结 - 31 -
参考文献 - 32 -
附  录 - 33 -

相关论文