Analysis of ECG Signal based on Feature Fusion and Two-Fold Classification Approach

Automated analysis of electrocardiogram (ECG) signal for home health care system can ensure early stage detection of cardiac arrhythmia. Most of the published works are based on advanced signal processing algorithms, large number of ECG data, complex classifiers for arrhythmia diagnosis. Hence it is difficult to implement these algorithms in portable health monitoring device. This study introduces two-fold feature selection and classification approach for easier and faster arrhythmia identification. The discrimination of different ECG data is employed by fusion of different non-linear, temporal and statistical parameters of ECG signals. The significances of individual feature in each stage of detection process are evaluated and ranked by introducing Euclidean score. Further, the eligible features are selected to construct of an integrated feature called fused feature. These fused features are efficient to accurately classify ECG beats with lesser computational burden. The proposed method is validated using MIT-BIH arrhythmia database. Experimental result of each stage classification shows better performance of the fused feature compare to the individual feature combinations. Furthermore, proposed work produces encouraging result comparing with existing work.