一种基于新型忆阻器的识别电路

Recognition circuit based on new memristors

文中提出了一种基于新型忆阻器的人工神经网络的硬件实现方法,通过上位机卷积神经网络对输入的图像作特征压缩,输出有限的待检测数据,采用Altera的EP4CE10F17C8N器件和多路数模转换器将数据转换为模拟信号输入到忆阻电路中,再用MCU(Micro Control Unit)对忆阻电路的输出信号进行检测处理,并通过具体的手势识别实验证实了方法的可行性。在改进忆阻器阵列模型的基础上,设计了一种神经形态识别电路,并对该电路做出探究和应用。研究表明,所搭建的忆阻器为核心的神经网络电路成功地实现了目标手势的识别,验证了忆阻器作为新型类脑器件在图像处理方面的应用能力。

A hardware implementation method for artificial neural networks based on a new-type memristor is proposed.The upper computer convolutional neural network is used to compress the input image features and output finite data to be detected.Then,a EP4 CE10 F17 C8 N device from Altera and a multi-channel digital-analog converter are selected to converter the signals into the memristor circuit by the way of converting the data into analog signals.After that,the output signals are detected and processed by a micro-control unit(MCU).The experiments of specific gesture recognition proves the feasibility of the method.The paper has achieved recognition of the specific gesture by the neural network circuit with the core of memristor,thus verifying the application ability using memristor as a new type of brain-inspired device in image processing.