基于FPGA的手写数字识别系统设计  

Design of handwritten digit recognition system based on FPGA

作　　者：张家炜 易艺 游骏 邝科兴 朱渊 王硕 ZHANG Jiawei;YI Yi;YOU Jun;KUANG Kexing;ZHU Yuan;WANG Shuo(College of Electronic Engineering,Guilin Institute of Information Technology,Guilin 541004,China)

机构地区：[1]桂林信息科技学院电子工程学院,广西桂林541004

出　　处：《电子设计工程》2024年第19期27-32,37,共7页Electronic Design Engineering

基　　金：国家级大学生创新训练项目(202113644014)。

摘　　要：为了解决手写数字识别准确率不高的问题,设计了一种基于FPGA的手写数字识别系统。系统以Xilinx公司的Zynq-7020 FPGA为处理器,利用OV5640摄像头获取卷积神经网络训练完成的手写数字信息,并进行分析和处理,通过HDMI接口发送至液晶显示器进行显示。实验结果表明,该系统能够对卷积神经网络训练完成的不同形状的手写数字0~9进行准确的识别,手写数字识别准确率大于95%,为手写数字识别提供了一种参考。In order to solve the problem of low accuracy of handwritten numeral recognition,a handwritten numeral recognition system based on FPGA is designed.The system uses Zynq-7020 FPGA of Xilinx company as the processor,uses OV5640 camera to obtain the handwritten digital information trained by convolutional neural network,and analyzes and processes it.It sends it to LCD display through HDMI interface.Experimental results show that the system can accurately recognize handwritten digits with different shapes from 0 to 9 trained by convolutional neural network,and the accuracy rate of handwritten digit recognition is more than 95%,which provides a reference for handwritten digit recognition.

关 键 词：深度学习 图像识别 FPGA 图像处理 Zynq 

分 类 号：TN79[电子电信—电路与系统]