Nonstationarity and 1/f noise characteristics in heart rate

The power spectrum of human heart rate (HR) measured over 24 h exhibits "power-law" 1/f alpha-type spectral behavior with alpha approximately 1. We investigate possible nonstationarity in time of the exponent alpha using maximum likelihood estimation, which allows relatively short data segments to be used. Examination of 24-h HR records from ambulatory normal and congestive heart failure (CHF) subjects indicates that the power-law structure of HR is nonstationary. In addition, alpha varies with time scale and is different for normal (alpha approximately 1) and CHF (alpha approximately 1.5) subjects. Simulations suggest that a possible mechanism underlying the observed power-law spectrum may be a switching between values of alpha near zero (white noise) and near two (Brownian motion). This mechanism generates power-law forms quantitatively similar to CHF subjects when the switching occurs very rapidly and similar to normal subjects when the switching is less rapid.