Non Verbal Behaviour Analysis for Distress Detection Using Texture Analysis

Facial expressions and gestures, which forms a part of inter personal communication, can reflect a person's true psychological state and his intent towards the whole communication scenario. During the long journey of technology development, lot of methods have been introduced for detecting the mental state of a person from such nonverbal communication aspects exhibited by the subject. The analysis of such non verbal behaviours can be used for many applications in psychiatry, in clinical, educational and organizational psychology some of which includes social skills training, military psychiatric evaluation, understanding interactions between a patient and therapist, children and parents, couples, civil servants and public etc. In this paper a novel method is presented to determine the distress level of a person by interpreting his facial expressions from salient patches using Segmentation based Fractal Texture Analysis (SFTA). To acquire texture patterns, fractal dimensions of the regions are found. Naïve Bayes classifier is used to classify the detected emotion into a predefined class. Viola Jones algorithm is used for the purpose of detection of face. The changes in the selected facial patches due to a change in emotion are extracted and analyzed for the detection of the affective state of the person. Successful testing of the proposed system is done with Extended Cohn-Kanade (CK+), Japanese female facial expression (JAFFE) database and Affectiva-MIT Facial Expression Dataset (AMFED).