Emotion recognition from posed and spontaneous dynamic expressions: Human observers versus machine analysis

Area of Science:

• Psychology, Cognitive Science, Computer Science, Artificial Intelligence

Background:

• Most emotion recognition research uses posed facial expressions from limited datasets.
• Understanding emotion recognition from both posed and spontaneous expressions is crucial for real-world applications.

Purpose of the Study:

• To compare human and machine performance in recognizing emotions from posed versus spontaneous facial expressions.
• To investigate cross-corpora emotion classification accuracy for automated systems.

Main Methods:

• Dynamic facial stimuli of six basic emotions were sourced from diverse databases.
• Human observers and a machine classifier evaluated the facial stimuli.
• Cross-corpora investigation methodology was employed.

Main Results:

• Machine classifiers outperformed humans on posed expressions with prototypical features.
• Machine performance was comparable to humans for spontaneous facial expressions.
• Both humans and machines achieved higher accuracy rates with posed stimuli.

Conclusions:

• Automated emotion recognition systems leverage expression prototypicality.
• Machine classifiers demonstrate human-level performance in cross-corpora spontaneous emotion recognition.
• Prototypicality is key for machine accuracy in emotion classification.