Real time estimation of evoked potentials by adaptive time-dependent filter

An improved consistent estimation of the signal is based on the average response. The proposed filter depends on the estimation of the signal and the noise autocorrelations. The filter coefficients are obtained by minimizing the mean squared error of the optimal filter of a non-stationary process applied on a single sweep. The average and autocorrelation matrixes can be built in real time without saving the measured data. A short off-line processing of the calculated data is needed afterwards.<>