Digital Eye Strain Detection System Based on SVM

Usage of digital devices especially smartphones significantly increased in the previous decade. Moreover, COVID pandemic has further shifted much of the work towards digital device assisted applications. In today's era, people across all ages are spending a lot of time in front of these devices. This also implies a surge in Digital Eye Strain cases, which is one of the emerging health issues. Researchers have linked this problem with symptoms such as dry eyes, altered blinking pattern, visual fatigue etc. Although the previous studies on facial features have already focused on blinking patterns, yawn detection and head movement, the proposed research work has concluded that other facial gestures comprising droopy eyes and decrease in glabellar length are also relevant features for this study to increase the accuracy. This paper tries to effectively detect when a user is under strain so that he or she can take timely precautions. A supervised method based on statistical features linked to suggested symptoms is proposed for classifying videos recorded in real time as user under strain using SVM. The main finding is an explicit feature set comprising of two newly proposed features along with four other apposite features derived from previous theoretical studies. The proposed system shows considerable increase in accuracy when tested on YawDD, the best possible dataset available for our use case.