Home // BRAININFO 2023, The Eighth International Conference on Neuroscience and Cognitive Brain Information // View article

Recruiting Neural Field Theory for Motor Imagery Data Augmentation

Authors:
Daniel Polyakov
Oren Shriki

Keywords: brain-computer interface, neural field theory, data augmentation, motor imagery, EEG.

Abstract:
Brain-computer interfaces accuracy is often limited due to a lack of diverse training data. In this study, we face this problem by using a computational model of neural dynamics, specifically Neural Field Theory, to generate artificial EEG time series as additional training data. We fitted this model to common spatial patterns of each motor imagery class, jittered the fitted parameters, and augmented the training data by generating time series from the model. We then applied a linear discriminant analysis to classify motor imagery states based on total-power features and tested the accuracy improvement on the ‘2a’ data set from BCI competition IV. Our findings show that data augmentation using Neural Field Theory can significantly improve the accuracy of brain-computer interface classifiers when the number of training samples is limited, providing a biophysically meaningful signal.

Pages: 8 to 9

Copyright: Copyright (c) IARIA, 2023

Publication date: March 13, 2023

Published in: conference

ISSN: 2519-8653

ISBN: 978-1-68558-067-4

Location: Barcelona, Spain

Dates: from March 13, 2023 to March 17, 2023