基于单导联的心电信号身份识别技术  

ECG Identification Technology Based on Single Lead

作　　者：王雪婷 张烨菲 张显飞[1] 赵治栋[1] WANG Xueting;ZHANG Yefei;ZHANG Xianfei;ZHAO Zhidong(Hangzhou Dianzi University,Hangzhou Zhejiang 310018,China)

机构地区：[1]杭州电子科技大学,浙江杭州310018

出　　处：《通信技术》2022年第5期625-633,共9页Communications Technology

基　　金：浙江省基础公益研究项目(LGG18F010012)。

摘　　要：随着国家、企业和个人对信息安全领域的愈发重视,基于心电(Electrocardiogram,ECG)信号的身份识别技术因具备活体识别的高防伪性引起了人们的关注。为了实现快捷、高效的个体识别率,构建了基于卷积神经网络的深度迁移识别模型。首先对原始心电信号进行质量等级评估,剔除影响识别结果的较差信号;其次将质量合格的一维心电信号转化为二维ECG轨迹图作为网络层的输入;最后利用GoogleNet网络模型实现对ECG轨迹图的迁移学习,并在MIT-BIH Arrhythmia Database数据库上获得了97.87%的识别准确度。结果表明,提出的识别算法对单导联心电信号具有较高准确率的识别效果。As state,enterprises and individuals attach more importance to the field of information security,the identification technology based on ECG signal has attracted people’s attention because of its high anti-counterfeiting performance of living body recognition.In order to achieve fast and efficient individual recognition rate,a depth migration recognition model based on convolutional neural network is built.First,the quality grade of the original ECG signal is evaluated to eliminate the poor signals that affect the recognition results.Then,the qualified one-dimensional ECG signal is transformed into a two-dimensional ECG trajectory as the input of the network layer.Finally,the transfer learning of ECG trajectory is realized by using GoogleNet network model,and the recognition accuracy of 97.87%is obtained on MIT-BIH Arrhythmia Database.The results indicate that the proposed recognition algorithm has high accuracy for single lead ECG signals.

关 键 词：心电信号 质量评估 卷积神经网络 迁移学习 身份识别 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]