Detecting Auditory Attention with Neural Networks
Author Information
Author(s): Masoud Geravanchizadeh, Shaygan Asl Amir, Danishvar Sebelan
Primary Institution: University of Tabriz, Iran
Hypothesis
Can a combined transformer and graph convolutional neural network effectively detect auditory attention from EEG signals?
Conclusion
The proposed TraGCNN model outperforms existing methods in detecting auditory attention from EEG signals.
Supporting Evidence
- The TraGCNN model achieved an accuracy of 80.12%, outperforming previous models.
- The model processes EEG data without needing speech stimuli, unlike traditional methods.
- Using a hybrid approach of transformers and graph neural networks improves detection accuracy.
Takeaway
This study created a new way to tell who you're listening to in a noisy room by looking at brain signals, making it easier for devices like hearing aids to help people hear better.
Methodology
The study used EEG data from participants to train a model that combines transformer and graph convolutional neural networks to detect auditory attention.
Potential Biases
Potential bias due to the exclusion of participants with hearing impairments.
Limitations
The study only included normal hearing participants and excluded those with hearing impairments.
Participant Demographics
21 normal hearing participants aged 51–76, with 22 subjects having hearing impairments excluded.
Statistical Information
P-Value
0.0001
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website