Averaging of nonlinearly aligned evoked potentials in impulsive noise environment

Abstract

The aim of this study was to adjust the method of averaging of nonlinearly aligned signals to make it work efficiently in impulsive noise environment. The main difficulty was the unfavorable influence of large amplitude impulses on the operation of nonlinear alignment (or time-warping). When the classical definition of the alignment costs was replaced by the Euclidean norm based one, and when additionally the ordered weighted averaging was used to diminish the role of large noise impulses, the method appeared rather effective. Not only was it robust against this type of noise but also allowed for effective suppression of the usual Gaussian noise.