Research on premature ventricular contraction real-time detection based support vector machine

This paper proposes a support vector machine (SVM) for real-time detection of premature ventricular contraction (PVC) from normal beats and others. This includes a signal feature extraction module and a statistical pattern recognition module. In feature extraction, time, frequency and morphological features are extracted, here six features are selected and made up a feature vector for input the pattern identifier. After this, an SVM is used to recognize PVC from normal beats and others; this classifier is fit for the requirements of precision and real-time at the same time. Finally, by means of testing electrocardiogram (ECG) data which from MIT-BIH arrhythmia database, the correct rating is more than 97%. Through the comparison with other methods, this achieves favorable results both in real-time and accuracy requirement.