Task-state EEG feature extraction for spatial cognition analysis: a power spectral density and permutation conditional mutual information approach
The permutation conditional mutual information (PCMI) method which combined time and spatial domain information was found effective in analyzing EEG signals of neuron groups. Considering the absence of frequency domain information in the PCMI algorithm, a feature extraction method based on Power Spe...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
Elsevier
2025
|
| Online Access: | http://psasir.upm.edu.my/id/eprint/120746/1/120746.pdf http://psasir.upm.edu.my/id/eprint/120746/ https://linkinghub.elsevier.com/retrieve/pii/S0306452225007183 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The permutation conditional mutual information (PCMI) method which combined time and spatial domain information was found effective in analyzing EEG signals of neuron groups. Considering the absence of frequency domain information in the PCMI algorithm, a feature extraction method based on Power Spectral Density Permutation Conditional Mutual Information (PSDPCMI) was first proposed to analyze task-state EEG signals. In this study, the performance of the PSDPCMI algorithm was employed for a VR spatial cognitive training experiment based on a Virtual Community training game and a Virtual City Walking testing game as carriers for subjects’ training and evaluation. The PSD, PCMI, and PSDPLV feature extraction methods were compared with the proposed PSDPCMI algorithm. According to the results, the highest classification accuracy was achieved across all frequency bands, with notable performance in the Theta, Beta2, and Gamma frequency bands. Besides, the new algorithm proposed also had higher precision, recall, F1-score, and AUC-score than the PSD, PCMI, and PSDPLV algorithms in most frequency bands and performed well in the classification accuracy of different classification models and another spatial cognitive dataset. The experimental outcomes demonstrated that the proposed PSDPCMI method in feature extraction showed a significant performance for the EEG signal analysis. |
|---|