摘要
基于忆阻器的神经形态电路是当前备受关注的研究课题,旨在采用忆阻器为耦合突触模拟生物体神经形态行为.该文基于斯特鲁普效应提出了一种新的神经形态电路用来模拟生物体的认知和纠错功能.首先提出了一种改进忆阻器模型用来模拟神经元间的耦合突触,然后在简化斯特鲁普模型的基础上,设计了一个三输入三输出的神经形态电路.提出的电路结构简单,仅包括感知模块、注意力引导模块、纠错模块和反应模块.电路测试结果表明该电路对优势反应(单词识别)能快速识别,而对非优势反应(颜色识别)则响应较慢.特别是,当单词颜色与单词词义不匹配时,电路会首先产生错误的颜色识别结果,在纠错模块的作用下该电路最终实现正确识别.该电路成功地模拟了生物体的认知和纠错功能,在智能机器人和脑机接口系统中具有一定的应用价值.
Implementing memristor-based neural circuits to mimic the neuromorphic behavior of living organismsis is an interesting subject.In this work,a memristor-based neural circuit with cognitive and error correcting functions based on Stroop effect is proposed.Firstly,an improved memristor model,which is considered as the coupling synapse,is proposed.Then,a simplified Stroop model is induced.Next,based on the simplified Stroop model,a complete neural circuit with three inputs and three outputs is developed in this study,which always responds quickly to the dominant reaction,i.e.,word recognition.Due to the interference from dominant reaction,the circuit responses slowly to the non-dominant reaction(color recognition).Specially,when the ink does not match the word,a wrong color recognition occurs.In this case,the error correction module produces a compensate signal and makes the correct recognition display in the reaction modules.Thus,the neural circuit can successfully simulate the cognitive and error correction processes of Stroop effect.Finally,PSpice simulation is carried out to verify the correctness and effectiveness of the neural circuit.
作者
刘天健
李志军
LIU Tianjian;LI Zhijun(School of Automation and Electronic Information,Xiangtan University,Xiangtan 411105,China)
出处
《湘潭大学学报（自然科学版）》
CAS
2024年第5期69-78,共10页
Journal of Xiangtan University(Natural Science Edition)
基金
国家自然科学基金(62171401,62071411)。
