Effect of Action Units, Viewpoint and Immersion on Emotion Recognition Using Dynamic Virtual Faces

Area of Science:

• Psychology
• Computer Science
• Human-Computer Interaction

Background:

• Facial affect recognition is crucial for social interaction but often impaired in psychiatric conditions.
• Virtual human (VH) and virtual reality (VR) technologies offer novel methods for assessing and training emotion recognition skills.

Purpose of the Study:

• To investigate the impact of specific factors on emotion perception using virtual humans.
• To examine the influence of action units (AUs), virtual human positioning, and VR immersion levels on recognizing emotions.

Main Methods:

• Thirty-six healthy participants evaluated dynamic virtual faces (DVFs) displaying basic emotions and neutral expressions.
• Conditions included varying levels of virtual human facial animation (AUs), viewing angles (frontal vs. mid-profile), and VR immersion (desktop vs. fully immersive).

Main Results:

• Accurate implementation of action units (AUs) in virtual faces significantly improved emotion recognition.
• Frontal views of virtual humans were recognized more effectively than mid-profile views.
• Immersive VR environments provided a slight enhancement in emotion recognition accuracy compared to desktop displays.

Conclusions:

• The precise animation of facial action units (AUs) is critical for effective virtual human-based emotion recognition.
• Optimizing virtual human presentation (frontal view) and VR immersion can enhance training efficacy.
• These findings contribute to the development of advanced virtual reality tools for emotion recognition training.