Optimum Filters for Estimating Evoked Potential Waveforms

The design of time-invariant and time-varying minimum mean square error filters for processing event related brain potentials is considered. A model for taking into account the randomness associated with the signal as well as that of the noise is described. Results using simulated data show the perfonnance of the time-varying filter to be greatly superior to that of the time-invariant filter. Results of processing measured visual evoked potentials are presented.