Cardiac Arrhythmia Classification Using Neural Networks and a Fuzzy Combination of Wavelet Transforms mid Autoregressive Modeling

This paper presents a new approach of feature extraction for reliable heart arrhythmia detection. This classification method is comprised of three components including data collection, feature extraction and classification of Electrocardiogram (ECG) signals. The new proposed feature extraction method is a fuzzy combination of three types of wavelet transforms (WT) together with the 4th order Autoregressive (AR) model coefficients to obtain the feature vector of ECG data. Then the multilayer perceptron neural network (MLP) is used to classify different ECG signals with different kinds of arrhythmias. In this paper, four types of heartbeats are classified: Normal beats, AF beat, VT beats and PSVT beats. Computer simulations are provided to verify the performance of the proposed method. The accuracy of ECG signals classification by using WT coefficients, WT coefficients together with 4th order AR model coefficients, and fuzzy combination of three kinds of WT coefficients together with 4th order AR model coefficients was obtained 86.7%, 91.3% and 94.2%, respectively.

Keywords:

Fuzzy , Wavelet Transform , AR Model Coefficients , ECG Classification , MLP Neural Network