【作者】 王国飞；

【导师】 杭国强；

【作者基本信息】 浙江大学 ， 电子与通信工程， 2012， 硕士

【摘要】 本文主要介绍了神经元MOS管电路和人工神经网络的研究现状,结合人工神经网络的发展方向,探讨了神经元MOS在神经元网络中的应用可能性及其一些特有的优点。分析了神经元MOS管的原理、结构以及Hspice仿真模型,并介绍和仿真了神经元MOS反相器和源极跟随器两种基本电路,它们是构成其它复杂电路的基础。介绍了人工神经网络的发展进程,比较了用软件实现和硬件实现人工神经网络两种方法,分析了用硬件实现的必要性。重点介绍了感知器、Hopfield神经网络以及Hamming神经网络。并将神经元MOS应用到了感知器电路中,利用单层感知器原理设计了与门和或门电路,利用多层感知器原理设计了异或门电路。将神经元MOS管应用到Hopfield神经网络A/D转换器电路中和联想记忆网络中,所设计的电路,结构简单,仿真正确。并根据Hopfield神经网络A/D转换器和感知器的原理,用神经元MOS管设计了二四值编译码器电路。将神经元MOS应用在Hamming神经网络电路中,设计了模板匹配电路和竞争全胜电路,并将系统应用在了数字字符识别中。本文的研究工作,探讨了神经元MOS管在神经网络中的应用前景。从本文的研究内容和仿真结果看,神经元MOS管的结构与神经网络有着很多相似之处。而且具有结构简单,功耗低等特点,非常适合在神经网络电路中应用。更多还原

【Abstract】 This paper analyzes the status and trends of the development of neuron-MOS circuits and Artificial Neural Network. The possibility and some unique advantages of the neuron-MOS in the network application are discussed,which meet the Artificial Neural Network’s development.The principle、structure and the Hspice simulation model are also analyzed.The paper also introduces the neuron-MOS inverter and source follower circuit.They are the basic circuits.which the other complex circuits based on.And a simulation on Hspice is made.The paper also introduces the development of Artificial Neural Network,compares the ways of realizing the neural network,one is by software,and the other is by hardware. And analyzes the necessity for hardware implementation.The paper mainly introduces the Perceptron、Hopfield Neural Network and Hamming Neural Network.The neuron-MOS is applied to the Perceptron circuit. And the AND gate circuit and OR gate circuit is designed by the principle of the single layer Perceptron,which the Exclusive OR gate is designed by the principle of multilayer Perceptron.The neuron-MOS is applied to the Hopfield Neural Network A/D converter circuit and Associative Memory Network circuit.The structure of the designed circuits is simple,and the simulation is right.And by the principle above,four-two value decoder circuit and two-four value encoder circuit are designed with neuron-MOS.In the last,the neuron-MOS is applied in the Hamming Neural Network circuit.The Template Matching circuit and WTA circuit are designed,which are applied in the digital character recognition.This paper discusses the application prospect of neuron-MOS in the Artificial Neural Network.From the research content and the result of simulation,the neuron-MOS and Neural Network have many similarities.But also has the advantages of simple structure,low power consumption,is very suitable for using in the Artificial Neural Network circuit.更多还原