Sequential Estimation Of The Power Spectrum For The Analysis Of Variability Of Non Stationary Cardiovascular Signals

The typical approach to the study of the variability of physiopathological conditions, is based on the analysis of the oscillatory signal components associated to appropriate time series, through various power spectral density (PSD) estimators [l]. Most methods of PSD estimation require the limiting assumption of stationarity of the time series which forces the use of short duration data segments, and is to be contrasted with the requirements of an adequate frequency resolution of the PSD estimation. This situation occurs typically in patient monitoring and exercise testing, producing long sequences of data. In order to identify the proper method for PSD estimation and to be aware of the relative advantages and limitations of the different solutions, a comprehensive system for power spectrum analysis has been realized [2]. The results have documented the superiority of the power spectrum estimation based on Wigner-Ville distribution. One limitation of such an approach is the presence of artifactual spectral contributions, at intermediate frequencies, making the estimation difficult to be interpreted in presence of multiple component signals. A recently published alternative to WV PSD estimation is presented in this paper [3].