Audiovisual Feature Analysis for Personality Extraversion Trait Recognition

There are a million of definitions and conceptions about personality. Almost all of these suggest that personality directly influence individual’s behavior. Thus, physiological responses, such as facial expression and speech, are the personality externalized cues or personality features. Human cognition easily can identify relevant temperament features during interactions and associate them through different ranges of personality traits. On the other hand, computer systems have severe limitations toward perceiving personality traits key features which negatively affect human-computer interactions. Motivated by that, the purpose of this work is to introduce an audiovisual feature analysis for extraversion personality trait classification. We intend to identify visual and audio features which contribute most for the extraversion trait classification. Different descriptors were applied on a personality extraversion trait dataset to achieve this. A linear SVM classifier was used to feature influence analysis. Our results show relevant visual and audio features for automatic personality recognition tasks.