基于FPGA和卷积神经网络的实时心梗诊断系统  

Real-time myocardial infarction diagnosis system based on FPGA and convolu‐tional neural network

作　　者：高鑫玮 刘文涵 谢文鑫 黄启俊[1] Gao Xinwei;Liu Wenhan;Xie Wenxin;Huang Qijun(School of Physics and Technology,Wuhan University,Wuhan 430072,China)

机构地区：[1]武汉大学物理科学技术学院,湖北武汉430072

出　　处：《电子技术应用》2023年第11期135-139,共5页Application of Electronic Technique

基　　金：国家自然科学基金(81971702)。

摘　　要：针对小型化日常心电监护系统的需求,设计了一套基于FPGA和卷积神经网络算法的心肌梗死疾病实时诊断系统。系统包含形态学滤波器、最小均方算法自适应陷波器、卷积神经网络硬件加速模块三大部分,通过在FPGA中并行化和加速处理,实现对心血管疾病的实时监护和诊断。经过上板验证,系统的相对准确率达到99.91%,片上功耗为2.39 W,处理时间为3.81 ms,可满足各项设计需求。Aiming at the demand of miniaturized daily ECG monitoring system,a real-time diagnosis system of myocardial in‐farction disease based on FPGA and convolutional neural network algorithm is designed.The system consists of morphological fil‐ter,least mean square adaptive notch filter and neural network hardware acceleration module.By parallel and accelerated process‐ing in FPGA,real-time monitoring and diagnosis of cardiovascular diseases are realized.After verification on the board,the sys‐tem achieves 99.91%relative accuracy,the on-chip power consumption is 2.39 W,and the processing time is 3.81 ms,which is suitable for various design requirements.

关 键 词：心肌梗死 实时系统 数字滤波器 神经网络加速 

分 类 号：TN911.72[电子电信—通信与信息系统]