Efficient R-peak Detection in Electrocardiogram Signal Based on Features Extracted Using Hilbert Transform and Burg Method

Electrocardiogram (ECG) is a non-invasive test which is highly adopted as a primary diagnostic tool for cardiovascular diseases. ECG recording appears as a non-stationary and quasi-periodic electrical signal. This electrical signal has important segments: P-wave, QRS-complex, and T-wave. R-peak is an important component of these segments. Computer-aided diagnosis is preferable as manual diagnosis using naked eye may mislead the detection. Therefore, in this paper, features extracted using both Burg method of autoregressive (AR) modeling and Hilbert transform are used for enabling efficient automated R-peak detection in ECG signal. Burg method is considered for extracting features due to its better frequency resolution, flexibility in selecting AR model orders and faster convergence for short-time signals. Next, Hilbert transform is used to find missed information in terms of the spectral components. The proposed technique is validated using Massachusetts Institute of Technology-Beth Israel Hospital Arrhythmia database. K-nearest neighbor (KNN) classifier is used for classification as it requires only few parameters to be tuned (K and the distance metric). In this paper K = 3 is selected to avoid any tie situation and Euclidean distance metric is selected because it does not require any weights for features. Also, KNN is more effective for classifying three classes as compared to those handled by other existing classifiers. The performance of the proposed technique is evaluated on the basis of sensitivity (Se), positive predictivity (PP), accuracy (Acc) and duplicity (D). The proposed work yielded Se of 99.90%, PP of 99.93%, Acc of 99.84%, and D of 0.006361%. These results indicate improvement in heart diagnostic leading to correct treatment of the subject (patient) over other existing state-of-the-art methods.