Emotion Classification from EEG Signals Using Machine Learning
Author Information
Author(s): Mendivil Sauceda Jesus Arturo, Marquez Bogart Yail, Esqueda Elizondo José Jaime
Primary Institution: Tecnológico Nacional de México, Campus Tijuana
Hypothesis
This study aimed to evaluate the performance of three neural network architectures for emotion classification using the SEED-V dataset.
Conclusion
The study highlights the challenges of generalizing across emotional states using raw EEG signals and emphasizes the need for advanced preprocessing and feature-extraction techniques.
Supporting Evidence
- ShallowFBCSPNet achieved the highest accuracy at 39.13%.
- Deep4Net followed with an accuracy of 38.26%.
- EEGNetv4 had the lowest accuracy at 25.22%.
Takeaway
The researchers tried to teach computers to recognize emotions by looking at brain signals, but they found it hard to tell different feelings apart.
Methodology
The study used the SEED-V dataset, which includes EEG recordings from 16 individuals exposed to 15 emotion-eliciting video clips, and evaluated three neural network architectures.
Potential Biases
EEGNetv4 showed severe bias by predicting all instances as a single emotion in some configurations.
Limitations
The models struggled to generalize across emotional states and exhibited significant misclassification issues.
Participant Demographics
Participants were 16 individuals exposed to various emotional stimuli.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website