Adaptive estimation and detection of evoked potentials

A multichannel adaptive signal enhancer designed for estimating and detecting evoked potentials is described. Studies involving mean-square-error measurements show that learning can be significantly improved by using a gating algorithm that updates the weights only when the signal is present. A symmetric noncausal format for the weights was found to extract the initial peak of the signal with the most consistency. Tests were performed using human visual evoked potential data. An investigation of several filter configurations showed that averaging performed as well as adaptive filtering for estimating the evoked potential. When detection of the signal is most crucial, the adaptive filter offers more advantages than averaging