Methods for extracting self-similarity properties from heart rate variability signal in normal and heart failure patients

We propose new methods for the extraction of self-similarity characteristics from 24-hour Heart Rate Variability signals (HRV). The first method is deduced from the “Second Order Difference plot”: the distribution dispersion obtained by a sequence HRV appears to be an estimation of the long-term correlation (H) of the signal. The second method (CYCLES) provides a decomposition of the self-similarity characteristics as a fraction of the time scales. We analyze 10 Normal, 10 Heart Failure (HF) and 2 transplanted subjects. Results show that, in normal subjects, the signal information is only present in very-short and long period (control scales), as the pathological group has different and heterogeneous patterns, with a global effect of reduced correlation in the control scales