Investigation of vision impairments using pattern reversal VEPs and extreme learning machine

Abstract

Analysis of vision impairment using pattern reversal visually evoked potentials (VEP) is gaining interest from researchers. The VEPs are collected using non-invasive EEG electrodes from the scalp overlaying the occipital lobe while the subject is presented with a pattern reversal checkerboard stimulus. The ophthalmologist is able to determine the abnormalities in visual pathways of a person by investigating these responses. The existing method however, is centered on the detection of amplitude and latency values of the signals. The ensemble averaging technique used to extract these information‟s would result in the variation between the single trials to be lost. The long period of testing and averaging could also result in patients fatigue and compromise the accuracy of the diagnosis. Hence, the analysis of single trial VEPs using wavelet transform is investigated in this paper. Biortogonal spline wavelets are applied to the VEP signals, and statistical features are extracted from the reconstructed signal for the classification. Extreme learning machine is used to distinguish the vision impairments and 92.87% accuracy is achieved using the proposed method.