Modeling the Electrocardiogram as Oscillatory Almost-Cyclostationary Process

A new model for the electrocardiogram (ECG) signal is proposed. Specifically, the ECG signal is modeled as an amplitude-modulated and time-warped version of a cyclostationary process which is the sum of a periodic signal and a zero-mean cyclostationary term. For the proposed model, the second-order characterization is derived in both time and frequency domains. The autocorrelation function is shown to be the superposition of amplitude- and angle-modulated sine waves and the Loeve bifrequency spectrum a spread version of that of the underlying cyclostationary process. The signal model belongs the recently introduced class of the oscillatory almost-cyclostationary processes. A procedure for estimating the second-order statistical functions in both time and frequency domains is outlined. The effectiveness of the proposed model and of the estimation procedure is corroborated by measurements on real ECG signals. These measurements are in full agreement with the theoretical analytical expressions. The proposed model is shown to be effective with observation intervals much larger than those adopted up to now with the classical cyclostationary model and is suitable to be exploited for arrhythmia modeling and characterization, and for diagnosis and biometric purposes.