一种反向串联忆阻突触电路的设计及应用

The memory effect of a memristor is similar to the function of a synapse in a biological nervous system. Its characteristics such as nanoscale size, low power consumption and high integration make the memristive synapse have bionic intelligent information processing capabilities, which is of great significance for constructing neuromorphic systems. In this paper, on the basis of an improved memristor model, a reverse-series memristive switch-type synapse circuit is designed. When the switch is open, the periodic square-wave voltage adjusts the memristance value (weight) to achieve weight update; when the switch is closed, the memristance value in the synapse circuit is used to connect the weight to store information. This synapse circuit has the STDP (spike - time - dependent - plasticity) bionic learning ability and the linear continuous characteristic of resistance value. In this paper, this synapse circuit is applied to the image storage of a crossbar array, the storage scheme is optimized, the influence of noise voltage on image storage is discussed, and numerical analysis and simulation comparison are carried out. The experimental results show that the proposed storage scheme is more reliable and robust than the single memristor crossbar array storage method.

忆阻器的记忆效应类似于生物神经系统中突触的功能,其纳米级尺寸、低功耗和高集成度等特性,使得忆阻突触具有仿生智能的信息处理能力,这对构建神经形态系统具有重要意义.本文在改进忆阻器模型的基础上,设计了一种反向串联忆阻开关型突触电路.当开关断开时,周期方波电压对忆阻值(权值)进行调节,实现权值更新;当开关闭合时,突触电路中的忆阻值被用于连接权值来存储信息.该突触电路具有STDP(spike-time-dependent-lasticity)仿生学习能力和阻值线性连续特性.本文将此突触电路应用于交叉阵列的图像存储中,优化了存储方案,讨论了噪声电压对图像存储的影响,进行了数值分析和仿真比较.实验结果表明:所提出的存储方案比单忆阻器交叉阵列存储方法更具有可靠性和鲁棒性.