Classification of premature ventricular contraction based on Discrete Wavelet Transform for real time applications

Develop of wearable cardiac monitors is becoming an important field of research because Cardiovascular disease is the leading cause of morbidity and mortality in the world. Real time arrhythmias detection algorithms are necessary to improve this kind of devices. This article presents a premature ventricular contraction detection method based on Discrete Wavelet Transform for preprocessing, segmentation and feature extraction. Discrete Wavelet Transform (DWT) is used to perform baseline wander and powerline noise reduction algorithm. Three different feature spaces based on wavelet coefficients are tested. Principal Component Analysis (PCA) is applied to reduce dimension into a lower feature space. K Nearest Neighbor (KNN) and Support Vector Machine (SVM) are developed and compared in terms of both accuracy and computational cost. Specificity of 97.18% and sensitivity of 96.47% with a prediction time of 0.47ms are accomplished. Computational burden is measured and compared with other methods to ensure that the developed method can be implemented in real time.