ESTIMATING CORRELATION EXPONENTS OF THE HEARTBEAT TIME SERIES

Abstract

The heartbeat time series commonly used in diagnostics is investigated. Since this series is not stationary, the difference series is considered and in the framework of stochastic processes its stationarity and mixing properties are tested. For this kind of data, affected by noise and low precision, a notion of correlation exponent, in some sense related to the correlation dimension, is proposed. The data analysis shows the existence of a profile of the correlation exponents, which evidentiates the amounts of randomness and of determinism in the heartbeat dynamics.