Automated recognition of auditory evoked potentials

Auditory evoked potentials are brain electrical potentials recorded from the scalp in response to stimulation of the auditory sensory mechanism. These potentials are very small, typically in the microvolt region, and are largely obscured by the normal brain EEG activity. Measurement demands a process of coherent averaging of the responses to a large number of stimuli in order to extract the required data. Two artificial neural networks (the backpropagation network and the radial basis function network) were chosen to test the data. A series of Moody-Darkin radial basis function (MDRBF) networks were constructed with a range of numbers of processor elements in order to determine optimum overall network architecture. Classification rate was used as a measure of the networks performance and represents the percentage of required responses achieved for each classification outcome. MDRBF networks showed themselves to be measurably more effective then backpropagation networks.