基于自适应快速S变换和XGBoost的心电信号精确快速分类方法袁莉芬李松尹柏强*李兵佐磊(合肥工业大学电气与自动化工程学院合肥230009)摘要：针对心电信号(ECG)传统分类方法效率较低的问题，该文提出一种基于自适应快速S变换(AFST)和XG-Boost的心电信号精确快速分类方法。该方法首先通过快速定位算法确定心电信号特征频率点，再根据特征频率点自适应调节S变换窗宽因子，增强S变换的时频分辨率的同时避免迭代计算，大大减少运行时间。其次，基于自适应快速S变换的时频矩阵提取12个特征量来表征5种心电信号的特征信息，特征向量维数低，识别能力强。最后，利用XGBoost算法对特征向量进行识别。MIT-BIH心律失常数据库和患者实测数据验证表明，该方法显著地缩短了分类时间，对5种心电信号的分类准确率分别为99.59%和97.32%，适用于实际检测系统中心律失常疾病的快速诊断。关键词：心电信号；心律失常；S变换；自适应快速S变换；XGBoost算法中图分类号：TN911.7;R540.41文献标识码：A文章编号：1009-5896(2023)04-1464-11DOI:10.11999/JEIT220217AccurateandFastElectroCardioGramClassificationMethodBasedonAdaptiveFastS-TransformandXGBoostYUANLifenLISongYINBaiqiangLIBingZUOLei(SchoolofElectricalandAutomationEngineering,HefeiUniversityofTechnology,Hefei230009,China)Abstract:ConsideringthelowefficiencyoftraditionalElectroCardioGram(ECG)classificationmethods,anaccurateandfastElectroCardioGramclassificationmethodbasedonAdaptiveFastS-Transform(AFST)andXGBoostisproposed.Firstly,themainfeaturepointsoftheECGsignalsaredeterminedthroughafastpositioningalgorithm,andthentheS-Transformwindowwidthfactorisadjustedadaptivelyaccordingtothemainfeaturepointstoenhancethetime-frequencyresolutionoftheS-transformwhileavoidingiterativecalculationandreducingtherunningtimegreatly;Secondly,basedonthetime-frequencymatrixofAFST,12eigenvaluesareextractedtorepresentthecharacteristicinformationof5kindsofECGsignals,withloweigenvectordimensionandstrongrecognitionability.Finally,XGBoostisusedtoidentifytheeigenvectors.TheexperimentalstudiesbasedontheMIT-BIHarrhythmiadatabaseandtheverificationofpatientmeasurementdatashowthat,withtheproposedmethod,theclassificationtimeofECGsignalsissignific...
