Dynamic features of the normal heart rate variability

A set of new features focussing primarily on different dynamic properties of heart rate variability was defined and investigated within the scope of the CinC 2002 challenge. The objective of the challenge was to find differences in heart rate variability between real and artificial heart beat signals. Among a larger number of features, several turned out to have high discriminative value: a) inter/intra signal power spectrum similarity; b) adjacent RR-interval changes; c) short range deceleration-acceleration ratio; and d) complexity of dynamical states and state transitions. Combined in a voting scheme, this set of features showed large correlation (agreement) in attribution of signals, resulting in very high classification accuracy on the challenge set.