基于新型忆阻器的多端输入LIF神经元电路的设计  

The design of multi-input LIF neuron circuit based on novel memristor

作　　者：柯善武 金尧耀 蒙嘉豪 吴鑫江 王今朝[2] 叶葱[1] KE Shanwu;JIN Yaoyao;MENG Jiahao;WU Xinjiang;WANG Jinchao;YE Cong(School of Microelectronics,Hubei University,Wuhan 430062,China;School of Materials Science and Engineering,Hubei University,Wuhan 430062,China)

机构地区：[1]湖北大学微电子学院,湖北武汉430062 [2]湖北大学材料科学与工程学院,湖北武汉430062

出　　处：《微电子学与计算机》2025年第2期86-92,共7页Microelectronics & Computer

基　　金：国家自然科学基金(62274058);湖北省重点研发计划(2022BAA020);武汉市重点研发计划(2022012202015055,2023010402010612)。

摘　　要：由于传统的互补金属-氧化物-半导体(Complementary Metal Oxide Semiconductor,CMOS)神经元电路与生物学的契合性较差且电路复杂,提出了一种基于忆阻器的多端口输入的泄露-整合-激发(Leaky-Integrate-Fire,LIF)神经元电路。该电路由运放、逻辑门等器件以及忆阻器构成,主要分为信号叠加模块和神经元信号产生模块。通过施加多个双尖峰脉冲信号并调节输入信号的数量和频率,模拟了生物神经元受到的不同程度刺激。研究发现施加到神经元上信号的数量和频率达到一定的值,神经元电路才会输出电压信号,这与生物体中只有受到一定程度的刺激时才会做出反应的现象是一致的。进一步,调节该电路中神经元信号产生模块的阈值电压大小,研究发现输入相同的信号,只有当电路的阈值电压较低时,神经元电路才能输出电压信号,这与生物中不同部位受到相同的刺激,神经元兴奋程度越高,越容易做出反应的现象一致。由此,该文所提出的LIF神经元电路不仅解决了传统电路输入信号单一、输入信号波形与生物信号波形差异大等问题,而且能模拟生物神经元的兴奋程度,这为人工神经网络的设计提供理论依据。Since the conventional Complementary Metal Oxide Semiconductor(CMOS)neuronal circuits are biologically incompatible and complex,this paper proposes a multi-port input Leaky-Integrate-Fire(LIF)neuronal circuit based on memristors.This thesis proposes a multi-port input LIF neuron circuit based on memristors.It consists of an op-amp,a logic gate and a memristor,and is mainly divided into a signal superposition module and a neuron signal generation module.By applying multiple bispike pulses and adjusting the number and frequency of the input signals,it simulates different levels of stimulation to biological neurons.It was found that the neuronal circuit would output voltage signals only when the number and frequency of signals applied to the neuron reached a certain value,which is consistent with the phenomenon that only a certain level of stimulation is required in living organisms.Further,by adjusting the threshold voltage of the neuronal signal generation module in the circuit,it was found that the neuronal circuit would only output a voltage signal when the threshold voltage of the circuit was lower for the same input signal,which is consistent with the phenomenon that the higher the excitation level of the neuron,the more likely it is to respond to the same stimulus in different parts of the organism.Thus,the proposed LIF neuron circuit not only solves the problems of single input signal and large difference between input signal waveform and biological signal waveform,but also simulates the excitation level of biological neurons,which provides a theoretical basis for the design of artificial neural networks.

关 键 词：忆阻器 LIF神经元电路 多端输入 阈值电压 人工神经网络 

分 类 号：TN389[电子电信—物理电子学]