Wavelet-Transformed k-NN Pipeline for EEG-Based Eye Blink Classification with Time Wrapping

Abstract

This paper presents a new pipeline to classify eye blinks based on the recording of electroencephalograms (EEG). It combines wavelet transform, k nearest neighbours (kNN) and a time wrapping method. The raw EEG signals are first preprocessed in a careful process by this methodology to remove artifacts and noise. This is done to make the consequent steps reliable. In order to record the complex and interested dynamic of eye blinks through EEG records, the wavelet transformation is used. This greatly removes the time as well as the frequency domain characteristics. The properties extracted are then entered into a kNN classifier with considerations of the spatial relation- ship so as to improve the accuracy of the classification of the model. Another approach is a novel method of time warping to be obligatory to the temporal changes in blink dynamics. This allows the model to capture the differences in the temporal shits of the patterns of blink. It is applied to the proposed framework on a tailored EEG data and contrasts it with the existing methods. It proved to be more accurate and adaptable to implement into the real life applications like in neurological diagnostics, fatigue and human computer interaction systems.