Classifying simulated and physiological heart rate variability signals

The main intention of this contribution is to sketch our way of analysing the 50 time series from the 2002 Computers in Cardiology challenge. The task to cope is to discriminate simulated and physiological heart rate variability signals. Our approach for doing this is rather simple: We exclude time series which show non-physiological behaviour. The methods applied serve to quantify the distribution of the RR-intervals, the circadian beat-to-beat variability as well as the beat-to-beat dynamics. Using cut-offs for these parameters, both time series groups can be discriminated clearly. Thus, the intricate interdependencies of variations in heart rate variability data on different scales are still difficult to simulate, such that even an experienced observer may be misled easily. To demonstrate the suitability of our methods not only for characterising simulated and physiological data, an outline of further applications shall be given.