Detection of Reading Movement from EOG Signals
Fatma Lati̇foğlu, Ramis İleri, Esra Demi̇rci̇, Çiğdem Gülüzar Altıntop
Abstract
In this paper, it is aimed to analysis of Electrooculography (EOG) signals recorded during the back to eye movement (retrieving words/re-reading) and skipping lines while reading. Two situations are characterized by large amplitude fluctuations in EOG signals. For this aim, EOG signals were recorded simultaneously while reading a text from 10 volunteers and changes in EOG signals caused by jumping a bottom line and back movements as reading were analyzed. The classification of these signals may allow the development of a new method for early and rapid diagnosis of various reading disorders (for example dyslexia). This study consists of two main processes; feature extraction and classification. Firstly, two features were determined from the recorded EOG signals for determination of retrieving words/re-reading from EOG signal. Then these signals were applied as input to various classifiers. The classifier performances were evaluated by calculating accuracy, sensitivity, specificity, precision and F measure. Overall classification results were obtained with high performance from all classifiers, and the highest accuracy of the classifiers used was 98% for each of the Random Forest and k-NN classifiers. The results show that this proposed method has an important performance for classification of eye movements from EOG signals.