Modeling transient correlations in heartbeat dynamics during sleep

We propose a model to generate stochastic signals with transient correlations, i.e. correlations of different strength and different typical duration within finite segments of the signal. The exponents and crossovers characterizing the correlations in the signal and in its variance can be tuned independently and allow us to generate model time series which are in agreement with data of heartbeat dynamics observed during wake and during different sleep stages. We also propose a model dynamics that reproduces the changes in the heartbeat fluctuations during the entire night, including transitions between different sleep stages.