Emotional State Detection from Touch-Based Behavior on Touch-Screen Device

This paper presents an attempt to recognize user emotion from finger-stroke pattern through analyzing current situation of emotion recognition applied to the touchscreen devices. There are four emotional selected for our analysis: relaxed, delighted, irritated, boring. First, we propose a set of features for emotion detection based touch-screen behavior. Then, ReliefF is used to analyze the effectiveness of these features and some important features are chosen from them. Finally, ANN and SVM are used to verify the recognizing rate of these chosen features. The experiment results proved that the method based on finger touch-screen behavior can identify the users' emotional state comparatively accurate when they using touch-screen device.