Efficient single-lead ECG beat classification using Matching Pursuit based features and an Artificial Neural Network

In this paper we employ the Matching Pursuit algorithm in order to obtain compact time-frequency representations of ECG data, which are then utilized from an ANN to achieve beat classification. To obtain optimum performance, the effect of the following attributes on the classification performance is examined: number of atoms, type of wavelet and number of ECG samples around the R peak. Our goal is to derive an accurate, efficient and real-time beat classification scheme, which could then be implemented on a resource-constrained portable device such as a cell phone. The proposed scheme is based on an existing beat classification method, but has the following favorable attributes: it utilizes less features, a single ECG lead and also only a single MLP in order to be able to discriminate between various abnormal beats. The performance of our approach is evaluated using the MIT-BIH Arrhythmia database. Provided results illustrate the accuracy of the proposed method (98.7%), which together with its simplicity (a single linear transform is required for feature extraction) justify its use for real-time classification of abnormal heartbeats on a portable heart monitoring system.