Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

Neural Signatures of Motion: A Deep Learning Framework for Classifying EEG Responses to Dynamic Facial Expressions

Poster Session E - Monday, March 9, 2026, 2:30 – 4:30 pm PDT, Fairview/Kitsilano Ballroom

Taylor Hamilton¹ (), Natalie Ceballos¹, Reiko Graham¹; ¹Department of Psychology, Texas State University

Decoding cognitive states from electroencephalographic (EEG) data remains a foundational challenge in neuroscience. This study used a machine learning (ML) framework to classify neural responses to dynamic versus static facial expressions. We analyzed single-trial, artifact-free EEG data from 23 participants while they viewed static and dynamic displays of fear and anger. The data were pre-processed in Scan 4.5 (Compumedics, Neuroscan). Features were extracted by calculating the Differential Entropy (DE), a measure of signal complexity, across four frequency bands (Delta, Theta, Alpha, Beta) using a sliding time window. This process transformed each trial's raw EEG data into a 3D tensor (bands, channels, time-steps), capturing the dynamic evolution of neural activity across space, time, and frequency for input into the model. These DE features were standardized and structured into channel-by-frequency 'images'. A residual neural network (RESNET18), a deep convolutional neural network (CNN), was then adapted to classify the data. This method achieved a mean accuracy of 97% by learning a consistent neural signature for discriminating between dynamic and static stimuli. The model learned that successful classification was driven by a pattern of decreased entropy in the alpha and beta bands, prominent at posterior visual processing regions. The same network was unsuccessful at accurately classifying neural activity to different facial expressions. Overall, preliminary results suggest that a decrease in entropy/randomness in EEG at lateral occipital regions may occur in response to dynamic facial displays. EEG features that distinguish between facial expressions may be manifested in other aspects of the signal.

Topic Area: METHODS: Electrophysiology