HRV adaptive spectral estimation for transient detection

Quantitative monitoring of cardiac autonomic control can be provided, noninvasively, by of heart rate variability (HRV) analysis. The study of the HRV is finding more and more fields of application. HRV long-term and short-term analysis can be performed. Among the latter, some applications have to deal with relatively brief changes of the sympatho-vagal activity, as, for example, the study of the fetal heart rate (correlated with uterine contractions). Parametric AR spectral estimation is commonly used to analyse the HRV signal association with the power spectrum bands related to simpatho-vagal activities. However, the computation is limited by a compromise between time resolution and stability. This study aims to indicate and to evaluate the performances of a multiple, recursive, time-varying AR identification of HRV power spectral density, which adapts to signal characteristic. This solution provides both synthetic information in stationary periods and accurate signal tracking during transients. An additional information, related to signal variation speed, can be used to provide a recognition of transient episodes. Test signals were used to measure the performance of the adaptive method with respect to the standard spectral estimation. Error analysis confirms the enhanced performances of the adaptive method