Development of a Novel Feedback System to Guide Skill Acquisition for Brain-Computer Interface Control
Year: 2020
Type: dissertation
Abstract: A brain-computer interface (BCI) enables control of external devices via modulation of neural activity. Currently, BCI use is limited as many users struggle to produce neural modulations that are machine-discernible. Recent evidence suggests ineffective user-training may be partially responsible. Here, we present and evaluate a feedback interface to improve BCI user-training. The system provides feedback derived from recently proposed Riemannian geometry metrics. First, we explore the validity of these metrics for user evaluation by assessing their relation to classifier performance. Regression models showed the metrics were able to account for 45% - 59% of inter-subject variation in classification accuracy for different classifiers. Second, we evaluate whether users improve performance in feedback interpretation tasks using the proposed system over classifier feedback. Using signal detection theory, we find the proposed system significantly outperforms the classifier system in all interpretation tasks. The results suggest the system should be explored further in real BCI-training scenarios.
Source: TSpace (University of Toronto)
Author: Nicolas Ivanov
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W3139923987