A genetic algorithm-based support vector machine model for detection of hearing thresholds

Abstract

ABSTRACT Auditory evoked potentials (AEPs), which are detected on the EEG auditory cortex area, are very small signals in response to a sound stimulus (or electric) from the inner ear to the cerebral cortex. These signals are recorded from electrodes attached to the scalp and are used for measuring the bioelectric function of the auditory pathway. In order to characterise their dynamic behaviour and due to the complex behaviour of nonlinear dynamic properties of EEG signals, several nonlinear analyses have been performed. In this work, Detrented Fluctuation Analysis (DFA) is applied to estimate the Fractal Dimension (FD) from the recorded AEP signals of the normal and the impaired hearing subjects. This aims at detecting their hearing threshold level. With the aim of classifying both groups, the normal and hearing impaired subjects, a hybrid approach based on the Support Vector machines (SVM) and genetic algorithms (GA) is taken: GA is applied to simultaneously optimise both SVM kernel parameters and feature subset selection. Our results indicate that the hybrid GA-SVM is promising; it is able to determine well the SVM kernel parameters along with feature subset selection, which will result in a high classification accuracy.