Tracking Eye Movements To Predict The Valence of A Scene

Studying human bio signals such as eye movements and tracking them can help in identifying and classifying the emotional essence of a scene. The existing methods employed to evaluate the reaction of the eyes based on exposure to a scene or image often use a classifier to extract features from eye movements. These extracted features are then evaluated to determine the valence of a scene. On the contrary, as much as eye movement has proved to be a reliable source in scene or image detection, factors such as how each feature affects the outcome of the prediction have not been explored. For the determination of the emotional category of images using eye movements, images are categorized into pleasant, neutral and unpleasant images and then these images are shown to the test subjects to record their response. Features of eye movement like fixation count, fixation frequency, saccade count, and saccade frequency among others, along with a machine learning approach was used for scene classification.