Automatic ECG interpretation via morphological feature extraction and SVM inference nets

This paper presents a novel approach to the intelligent heart rhythm recognition, via integration of Hermite based orthogonal polynomial decomposition (OPD) and support vector machines (SVMs) classification. In regard to feature characterization, the orthogonal transformation based on Hermite basis polynomials is proposed to characterize the morphological features of ECG data. For the goal of multi-class ECG classification, the one-against-all (OAA) strategy is applied to reduce the multi-class SVMs into several binary SVMs. In this study, most of the heart rhythm type in MIT-BIH arrhythmia database is concerned. The numerical result shows out the good performance of proposed automatic interpreter in reliability and accuracy.