A radial basis function neural network for single sweep detection of somatosensory evoked potentials

The aim of this research is to evaluate the effectiveness of the employment of a Gaussian radial basis function neural network (RBFNN) for the on-line detection of single sweep somatosensory evoked potentials (SswSEPs), paying particular attention to the capability of tracking trial-to-trial variabilities. On the basis of simulations the parameters of the network have been set and the results have then been compared with those obtained from other methods, in particular with the ensemble averaging, the moving window averaging and the ARX filtering. This research shows a better performance of the RBFNN, because it is able to follow changes of the underlying signal even in noisy conditions and does not require prior assumptions.