Khelouia, RomeyssaTouil, SoumeyaDaamouche, Abdelhamid (supervisor)2022-06-222022-06-222019https://dspace.univ-boumerdes.dz/handle/123456789/961959 p.This report proposes a method to detect and classify two types of heart beat, namely Normal (N) beats and Atrial Fibrillation (AF) beats. Each electrocardiogram (ECG) record is band-pass filtered, then segmented into beats to form the original features. After that, statistical features of the Discrete Wavelet Transform (DWT) approximation and detail coefficients of each beat constitute a feature. The RR intervals surrounding the beat are also used as features. Finally, extracted features are classified using a Support Vector Machines (SVM) classifier. The MIT-BIH atrial fibrillation and MIT-BIH arrhythmia databases are used to evaluate the performance of the proposed method. Results from extensive experimentations appear very promising, with an accuracy of 98.2723%, a sensitivity of 98.3709%, and a specificity of 97.2313%.enWaveletsElectrocardiogramAtrial fibrillationThesis