Wavelet models of event-related potentials

Presents a wavelet approach to the statistical analysis of event-related potentials (ERPs). The recorded data are represented by a linear combination of unknown component waveforms. A unique decomposition is obtained by prior time/frequency constraints imposed in the wavelet representation, and by differential effects of experimental conditions on the components. Even though the wavelets are orthogonal, the modeled component waveforms need not be orthogonal and may overlap in time and frequency. The author demonstrates estimation of components and condition effects in applications to the cat auditory evoked potential recorded at various stimulus rates and to the human auditory P300 recorded before and after administration of clonidine