Analysis of EOG Signal Deviations and CNN-Based Classification of Ocular Fatigue Due to Prolonged Consumption of Digital Screen

This study focuses on analyzing and classifying deviation of the EOG signal in eyes with ocular symptoms in contrast to normal eyes while using a digital screen (DS). Continuous Wavelet Transformation (CWT) is performed to decompose the signal, and the spectral power of the EOG signal is computed. The obtained $\mathbf{p}$-values and effect sizes ($\mathbf{d}$) from statistical tests, indicate significant changes in the EOG signal at different time segments due to prolonged exposure to DS. For symptomatic eyes, lower $\mathbf{p}$-values are compared to the assumed significance level of 0.05 and medium to large level effect sizes are observed, indicating irregular and uncontrolled eye movement. A Convolutional Neural Network (CNN) has been used to classify the EOG deviation, and a classification accuracy of $\mathbf{9 1. 6 5} \pm \mathbf{3. 8 7 \%}$ is achieved by utilizing EOG spectrogram images. This study will be helpful for physicians to make critical decisions regarding the purpose of their diagnosis based on the EOG signal.