Evidence for dimensional representations and anticipatory dynamics in facial expression perception

Area of Science:

• Neuroscience
• Cognitive Science
• Computer Vision

Background:

• Facial expression recognition is crucial for social interaction.
• Understanding the neural basis of dynamic expression perception is challenging.

Purpose of the Study:

• To investigate the neural representation of dynamic facial expressions using electroencephalography (EEG).
• To decode a wide range of emotional and conversational expressions from EEG data.
• To reconstruct dynamic visual representations from neural signals.

Main Methods:

• Decoding of 24 distinct facial expressions (14 emotional, 10 conversational) from human adult EEG data.
• EEG-based video reconstruction to visualize dynamic neural representations.
• Time-resolved decoding analysis to identify anticipatory neural dynamics.
• Validation of neural reconstructions against behavioral data.

Main Results:

• A broad spectrum of facial expressions, including subtle variations, were successfully decoded from EEG signals.
• The representational structure of decoded expressions aligned with valence and arousal dimensions.
• EEG-based reconstructions captured dynamic and fine-grained differences between similar expressions.
• Anticipatory neural dynamics were identified, predicting expression onset with enhanced accuracy.

Conclusions:

• Neural signals encode rich, dynamic information about facial expressions.
• EEG-based reconstruction offers a method to visualize and understand neural representations of visual stimuli.
• This study provides insights into the neural mechanisms underlying expression recognition and visual perception.