A non-stationary model of single evoked EEG activity and its estimation

There are many different techniques of estimating evoked EEG activity. Each construction of an estimator depends on the model of the process generating the evoked activity. In this paper a model is proposed in which the evoked activity is a transient signal described by the so-called uniformly modulated stochastic process. The usual supposition, that the evoked activity does not depend on the background activity is replaced by the conditional independence of the spontaneous prestimulation and evoked activities. A linear filter with a time varying frequency transfer function is used to construct the estimator of the evoked activity. A statistical relationship between the spontaneous prestimulation and the background activities is considered and the single evoked activity estimator is proposed.