Human recognition of basic emotions from posed and animated dynamic facial expressions

摘要

Facial expressions are crucial for social communication, especially because they make it possible to express and perceive unspoken emotional and mental states. For example, neurodevelopmental disorders with social communication deficits, such as Asperger Syndrome (AS), often involve difficulties in interpreting emotional states from the facial expressions of others.Rather little is known of the role of dynamics in recognizing emotions from faces. Better recognition of dynamic rather than static facial expressions of six basic emotions has been reported with animated faces; however, this result hasn't been confirmed reliably with real human faces. This thesis evaluates the role of dynamics in recognizing basic expressions from animated and human faces. With human faces, the further interaction between dynamics and the effect of removing fine details by low-pass filtering (blurring) is studied in adult individuals with and without AS. The results confirmed that dynamics facilitates the recognition of emotional facial expressions. This effect, however, was apparent only with the facial animation stimuli lacking detailed static facial features and other emotional cues and with blurred human faces. Some dynamic emotional animations were recognized drastically better than static ones. With basic expressions posed by human actors, the advantage of dynamic vs. static displays increased as a function of the blur level. Participants with and without AS performed similarly in recognizing basic emotions from original non-filtered and from dynamic vs. static facial expressions, suggesting that AS involves intact recognition of simple emotional states and movement from faces. Participants with AS were affected more by the removal of fine details than participants without AS. This result supports a "weak central coherence" account suggesting that AS and other autistic spectrum disorders are characterized by general perceptual difficulties in processing global vs. local level features.