Classification of ECG Arrhythmias Based on Statistical and Time-Frequency Features

In this paper a new approach to accurately classify ECG arrhythmias through a combination of the wavelet transform and artificial neural network is presented. Three kinds of features in a very computationally efficient manner are computed as follows: 1-joint Time-Frequency features (discrete wavelet transform coefficients). 2-time domain features (R-R intervals). 3-Statistical feature (Form Factor). Using these features, the limitations of other methods in classifying multiple kinds of arrhythmia with high accuracy for all of them at once are overcome. Finally, a cascade classifier including two ANNs has been designed. Considering the whole MIT-BIH arrhythmia database, 10kinds of a rrhythmia were classified. The overall accuracy of classification of the proposed approach is above 90%.