An efficient wavelet-based automated R-peaks detection method using Hilbert transform

Abstract Machine-aided detection of R -peaks is becoming a vital task to automate the diagnosis of critical cardiovascular ailments. R -peaks in Electrocardiogram (ECG) is one of the key segments for diagnosis of the cardiac disorder. By recognizing R -peaks, heart rate of the patient can be computed and from that point onwards heart rate variability (HRV), tachycardia, and bradycardia can also be determined. Most of the R -peaks detectors suffer due to non-stationary behaviors of the ECG signal. In this work, a wavelet transform based automated R -peaks detection method has been proposed. A wavelet-based multiresolution approach along with Shannon energy envelope estimator is utilized to eliminate the noises in ECG signal and enhance the QRS complexes. Then a Hilbert transform based peak finding logic is used to detect the R -peaks without employing any amplitude threshold. The efficiency of the proposed work is validated using all the ECG signals of MIT-BIH arrhythmia database, and it attains an average accuracy of 99.83%, sensitivity of 99.93%, positive predictivity of 99.91%, error rate of 0.17% and an average F -score of 0.9992. A close observation of the simulation and validation indicates that the suggested technique achieves superior performance indices compared to the existing methods for real ECG signal.