Kalman smoother based time-varying spectrum estimation of EEG during single agent propofol anesthesia

A time-varying parametric spectrum estimation method for analyzing EEG dynamics is presented. EEG signals are first modeled as a time-varying auto-regressive stochastic process and the model parameters are estimated recursively with a Kalman smoother algorithm. Time-varying spectrum estimates are then obtained from the estimated parameters. The proposed method was applied to measurements collected during low dose propofol anesthesia. The method was able to detect changes of event related (de)synchronization type elicited by verbal command.