Heartbeat Classification Using A Joint EMD- Moment Technique

in this paper, a classifier based on statistical learning theory has been proposed for binary and multi classification of electrocardiogram signals. In the first stage empirical mode decomposition (EMD) is applied on the ECG signals. EMD decomposes a segment of ECG into several components known as intrinsic mode functions (IMF). Then, Hilbert transform is used to calculate power spectrum of each IMF. In the third stage, central moments are computed for each IMF. Finally, estimated statistical moments are used as features to train a support vector machine (SVM). The technique proposed in this paper for feature extraction (i.e. EMD- Hilbert- Statistical moment) provides a good representation of time-frequency characteristics of ECG signals. The proposed method is applied to a set of ECG signals comprising normal and patient groups. The obtained results show a significant improvement in classifying the heartbeats using the proposed method.

Keywords:

Electrocardiogran , Emprical Mode Decomposition , Statistical Moment