基于ReS2/h-BN/石墨烯/h-BN浮栅结构的感存算一体化器件OA
A sensing-memory-computing integrated device based on ReS2/h-BN/Graphene/h-BN floating-gate structure
传统机器视觉系统普遍采用图像传感、存储与处理相互分离的架构,导致系统结构臃肿、能效低下等问题,严重限制了其在低功耗场景中的应用.针对这一挑战,构造了一种具有栅压可调的正负双极性光响应行为的ReS2/h-BN/石墨烯/h-BN浮栅结构器件,该器件能够将光学感知、信息处理与存储功能集成于一体,为发展新型低功耗、结构简单的机器视觉系统带来了机遇.此外,还构建了手写希腊字母(小写)的数据集,利用基于ReS2/h-BN/石墨烯/h-BN浮栅结构器件双极性光响应特性,采用物理权重8 bit量化的AlexNet网络模型进行识别测试.结果表明,系统字符识别准确率高达99.95％.
Traditional machine vision systems generally adopt an architecture in which image sensing,storage,and processing are separated from each other,resulting in a bloated system structure and low energy efficiency,which seriously limits their application in low-power scenarios.To address this challenge,a floating-gate device,ReS2/h-BN/Graphene/h-BN,with gate-voltage-adj ustable positive and negative bipolar photoresponse behavior was constructed.This device integrates optical sensing,information processing,and storage functions,creating opportunities for the development of new low-power consumption,simple-structure machine vision systems.Furthermore,this work constructed a dataset of handwritten Greek letters(lowercase),conducted pattern recognition testing using a physically weighted 8-bit quantized AlexNet network based on the ambipolar photoresponse characteristics of ReS2/h-BN/Graphene/h-BN devices,and ultimately achieved a system recognition accuracy as high as 99.95％.
方伟杰;王所富;龙明生
安徽大学物质科学与信息技术研究院,信息材料与智能感知安徽省实验室,安徽 合肥 230601安徽大学物质科学与信息技术研究院,信息材料与智能感知安徽省实验室,安徽 合肥 230601安徽大学物质科学与信息技术研究院,信息材料与智能感知安徽省实验室,安徽 合肥 230601
信息技术与安全科学
感存算一体化机器视觉系统浮栅结构双极性光响应
integrated sensing-memory-computationmachine vision systemsfloating-gate structurebipolar photoresponse
《安徽大学学报（自然科学版）》 2026 (3)
36-42,7
信息材料与智能感知安徽省实验室开放课题(IMIS202207)
评论