Adaptive estimation of evoked potentials using wavelets and radial basis functions neural networks

A new signal estimation technique is proposed that uses the wavelet transform to determine the time-frequency information content of human evoked brain potentials. By adaptively weighting the individual time-frequency bins using a neural network, a time-varying estimate of the evoked potential can be obtained. The weighting procedure enhances the time-frequency bins corresponding to the EP while attenuating those corresponding to the EEG.