Decoding Working Memory Load from EEG and Behavioral Dynamics

Xiaoyu Ye

doi:10.54097/y2c4fp94

Authors

Xiaoyu Ye

DOI:

https://doi.org/10.54097/y2c4fp94

Keywords:

Working Memory, Support Vector Machine, CSP-SVM Framework, EEG, Common Spatial Pattern

Abstract

Working memory (WM) is critical for cognitive tasks such as reasoning and problem-solving, and its performance is heavily influenced by cognitive load. This study explores the use of EEG, combined with Common Spatial Pattern (CSP) and Support Vector Machine (SVM) classification, to predict cog- nitive states under different memory load conditions (easy, medium, and hard). Behavioral performance, including accuracy and reaction time (RT), was assessed across memory load levels, with RT increasing and accuracy decreasing as memory load became more difficult. EEG data were preprocessed to isolate probe-related activity, and CSP was employed for spatial feature extraction, followed by SVM classifica- tion to decode cognitive states. Our results demonstrated robust binary and multi-class classification per- formance, with average test accuracies consistently exceeding 80% across subjects. Notably, the model achieved high accuracy in distinguishing between memory load conditions, with minimal misclassifi- cations and strong consistency between test and cross-validation results. These findings highlight the potential of the CSP-SVM framework for real-time cognitive state monitoring, offering a non-invasive method for tracking cognitive load in both healthy and clinical populations. Future work will explore further applications in neuroadaptive technologies and cognitive impairment monitoring.

Downloads

Download data is not yet available.

References

[1] P. L. Ackerman. Cognitive ability and working memory as predictors of cognitive performance. Jour- nal of Cognitive Psychology, 26(2):107–128, 2014.

[2] Alan Baddeley. Working memory: Theories, models, and controversies. Annual Review of Psychology, 63:1–29, 2012.

[3] B. Blankertz, S. Lemm, M. Treder, S. Haufe, and K. W. Mueller. Generalizing common spatial patterns across brain-computer interface tasks. IEEE Transactions on Biomedical Engineering, 55(6):1–9, 2008.

[4] Vince D. Calhoun, Tamsyn Adali, and Godfrey D. Pearlson. Method for estimating the number of independent components in multichannel data: An example using fmri data. NeuroImage, 22(2):1200– 1217, 2004.

[5] Richard J. Davidson. Neural bases of emotion regulation: Affective neuroscience perspective. Biolog- ical Psychiatry, 73(4):293–295, 2013.

[6] A. Delorme and S. Makeig. Eeglab: an open source toolbox for analysis of single-trial eeg dynamics including independent component analysis. Journal of Neuroscience Methods, 134(1):9–21, 2004.

[7] A. Delorme and S. Makeig. Automated rejection of artifactual eeg segments using ica. NeuroImage, 34(3):1446–1455, 2007.

[8] Randall W. Engle. Working memory capacity as executive attention. Current Directions in Psycho- logical Science, 11(1):19–23, 2002.

[9] H. He, D. Wu, S. Zhang, and B. L. Lu. Transfer learning for brain-computer interfaces: A euclidean space data alignment approach. IEEE Transactions on Biomedical Engineering, 62(7):1517–1527, 2015.

[10] R. N. A. Henson and M. D. Rugg. Analysis of event-related potentials in cognitive neuroscience. Brain Research Reviews, 41(3):227–242, 2002.

[11] R. Johnson, J. M. Olichney, and S. A. Raskin. Event-related potential indices of executive control in working memory and attentional control tasks. Biological Psychology, 62:231–246, 2003.

[12] T. P. Jung, S. Makeig, C. Humphries, T. W. Lee, M. McKeown, V. Iragui, and T. J. Sejnowski. Remov- ing electroencephalographic artifacts by blind source separation. Psychophysiology, 37(2):163–178, 2000.

[13] Tzyy-Ping Jung, Scott Makeig, Marita Stensmo, and Terrence J Sejnowski. Brain dynamics of mental arithmetic: Evidence from eeg coherence and neural networks. NeuroImage, 22(3):1412–1421, 2004.

[14] John G Keilp, Harold A Sackeim, Beth S Brodsky, Maria A Oquendo, Kevin M Malone, and J John Mann. Neuropsychological dysfunction in depressed suicide attempters. American Journal of Psychi- atry, 158(5):735–741, 2001.

[15] M. D. Low and R. McLoughlin. High-pass filtering in eeg analysis. Brain Topography, 12(4):297–305, 1999.

[16] B. Luna, A. Padmanabhan, and K. O'Hearn. Neural regions involved in the maintenance of information in working memory. Neuropsychologia, 36(11):1159–1172, 1998.

[17] Mohammad Moradi, Mohammad Ali Abtahi, and Sahand Gharibzadeh. Transient artifact removal using wavelet transform and ica in eeg data analysis. Signal Processing, 98:160–172, 2014.

[18] P. L. Nunez and R. Srinivasan. Electric fields of the brain: The neurophysics of EEG. Oxford University Press, Oxford, 2nd edition, 2011.

[19] J. Onton, A. Delorme, and S. Makeig. Frontal midline eeg dynamics during working memory. Neu- roImage, 27(2):341–356, 2005.

[20] J. Onton, A. Delorme, and S. Makeig. Frontal midline eeg dynamics during working memory. Neu- roImage, 27(2):341–356, 2005.

[21] R. D. Pascual-Marqui, C. M. Michel, and D. Lehmann. Eeg microstate maps are related to sleep spindles. NeuroImage, 91:155–165, 2014.

[22] Tamara A. Victor, Holly A. Marusak, Kelsey Wilkins, Jacob Berman, Deborah Malaspina, and Lynn E. DeLisi. Cognitive impairments in patients with major depressive disorder at risk for suicide. Cognitive Therapy and Research, 35(2):144–156, 2011.

[23] X. Wan, F. Liang, and X. Yin. Fast and accurate spatial interpolation for eeg/meg data. IEEE Transac- tions on Biomedical Engineering, 63(7):1349–1355, 2016.

[24] Z. Wang, L. Xie, X. Wang, and H. He. Eeg-based classification of mental workload during n-back tasks using wavelet transform and common spatial patterns. Journal of Neural Engineering, 15(2):026018, 2018.

[25] Huijuan Yang, Siavash Sakhavi, Kai Keng Ang, and Cuntai Guan. On the use of convolutional neural networks and augmented csp features for multi-class motor imagery of eeg signals classification. In 2015 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pages 2620–2623. IEEE, 2015.