Automatic prediction of trait anxiety degree using recognition rates of facial emotions

The trait anxiety degree is a significant standard to measure the psychological status, and the measurement (score) of trait anxiety degree is generally obtained by a very complex text questionnaire, which usually takes large amount of time and is subjectively various according to environmental condition. On the other hand, the researches in psychological field have proven that personality recognition of different facial emotions is strongly related to the degree of trait anxiety. In this work, we propose to automatically predict the trait anxiety score using the recognition rates of different facial emotions. In order to select compact and discriminant features, we investigate a correlation-based feature selection strategy in both raw data and PCA transformed space. Experimental results show that our proposed strategy can achieve reasonable trait anxiety score, which, also can validates the reliable relation between recognition rates of facial emotion and trait anxiety score.