Spectral estimation of nonstationary time series

The spectral estimation of nonstationary time series is a long standing problem of major importance. Examples of such time series can be drawn from many diverse fields such a communications, foetal heart rate and condition monitoring. The time series involved can be nonstationary in mean, covariance or both. The authors examine a time series which is artificially generated and can have either a slowly or a rapidly changing spectrum. Attempts are made to model such time series using two different techniques so that their spectra may be estimated. The trade-offs involved in using the two different modelling techniques are then discussed