An automatic atrial fibrillation detection scheme based on statistical features from differential electrocardiogram signals

Atrial fibrillation (AF) is a irregular heartbeat which can lead to many heart-related complications. The risk due to this complications can be greatly minimized if AF can be detected early. Therefore, computer aided AF detection technique from Electrocardiogram (ECG) signal is in great demand. This paper proposes an automatic efficient atrial fibrillation detection scheme from Differential Electrocardiogram (dECG) signals. At first, from the given ECG signal, baseline signal is removed with the help of median filter. Then, variation patterns in consecutive samples are analyzed and R-peaks are detected from dECG. After that, statistical features are extracted from the possible P-wave location from each ECG bit. Finally, to classify the Atrial Fibrillation frames with the help of extracted statistical features, the Linear Discriminant Analysis (LDA) classifier is employed. Experimentation on a publicly available database show that the proposed methodology shows noteworthy performance in detection of AF in terms of accuracy, sensitivity and specificity.