Graphical Parameter Evaluation for Affective Image Classification using Dimensional Affective Groups of International Affective Picture System

Input parameter evaluation for affective image classification using dimensional affective groups is proposed against the existing researches that use discrete affective groups which are not accurate enough due to not covering all three emotional vectors Valence, Arousal and Dominance. We obtain dimensional affective groups by clustering the known emotional vectors of International Affective Picture System set for all subjects, evaluate the affective group contents by using photo description. The histogram based and texture based fundamental graphical features of photos without borders are tested as input data sets for classification. Evaluation is by training neural network to identify the affective groups where the inputs originated. The test results shows the potentials like that the gray scale is suitable for high Valence groups or high Arousal groups, edges and corners are suitable for high Valence groups, blue histogram is accurate for high