Evoked potentials in the brain: physiological and abstract optimal models for single sweeps processing

Two stochastic models are presented for modeling the evoked and event-related potential elicited in the brain in such a way as to perform an optimal processing of the single electrical response to a stimulus. An autoregressive with exogeneous input model is reported to improve the S/N (signal-to-noise) ratio by some 30 dB, provided that an average template for the potential under analysis is previously known. A time-varying just-autoregressive model is shown to have comparable performance, modeling the EP (evoked potential) as a time-varying EEG: this requires a more complex and time-consuming procedure, such as Kalman filtering instead of a straight least-squares identification, but provides a deeper description of the intrasweep variability of the single EP, besides an alternative description of the intersweep variability