Performance Evaluation of Average Methods in the Time Domain Using Quality Measures for Automatic Detection of Evoked Potentials

Abstract

For the reduction of noise and extraction of the evoked potentials several methods have been reported in the literature based on the average of signals, which try to counteract the limitations of the classical technique used for the signal enhancement, which is the coherent average. The quality of the resulting signal can be evaluated using measurements in the time domain or the frequency domain. The objective of this work is to evaluate the performance of different average methods using quality measures in the time domain to determine which combination of an average method with quality measurement is better adjusted to the detection of transient auditory evoked potentials. To fulfill this objective, the classic coherent average, the median, the weighted average and four other versions of the robust average based on the median are tested. The quality measures used to compare the different average methods are the correlation coefficient, the standard deviation ratio and the variance of the single point. It was obtained as a result that the robust method Tanh mean, together with the measure of the variance of single point offers the best results.