TIME-VARYING CARDIOVASCULAR OSCILLATIONS

摘要

Signals derived from the human cardiovascular system (CVS) are exceptionally complex, being time-varying, noisy, and of necessarily limited duration. Yet an appropriate analysis of them may be expected to yield detailed information about the dynamics of the underlying physiological processes. A new approach to the analysis and modelling of CVS signals is proposed. It combines decomposition of the signals into principal modes and a novel method of pa- rameter identification in nonlinear stochastic systems based on Bayesian inference. The scheme is tested on a noisy Van der Pol oscillator, for which it yields rapid convergence and correct inference of the known parameters. Preliminary applications to CVS data are discussed.