EMD method for automatic ECG fiducial points detection

Abstract

The automatic detection of electrocardiogram waves presents an important step for cardiac disease diagnosis. In this work, we developed an algorithm for locating the waveform boundaries by using the empirical mode decomposition which has interesting properties concerning the pseudo periodic signals. The introduced method allows identifying an appropriate and optimum set of intrinsic mode functions (IMFs) to reconstruct a wave from ECG signal. Firstly, the R wave is detected from each heartbeat using the first tree IMFs of empirical mode decomposition (EMD) analysis. Then, the Q wave and T wave of each QRS complex are delineated. Next, the determination of P wave and T wave was performed employing a set of higher order IMFs. We tested the reliability of our algorithm on the QT Database, the manually annotated database. A comparison is also made with other delineation algorithms for the assessment of system performance. The obtained results showed a better performance and accurate detection of ECG fiducial points.