A Ternary Brain-computer Interface based on Single-trial Readiness Potentials of Self-initiated Fine Movements
Year: 2016
Type: dissertation
Abstract: This thesis investigates the feasibility of a ternary brain-computer interface (BCI) using the single-trial readiness potential (RP) of self-initiated fine movements. The BCI classified among an idle state and fine motor movements in the left and right hands. Three studies were conducted, each investigating a step towards this ternary BCI. The first study presented a novel technique of feature fusion from an optimal number of electrodes, for the prediction of a self-initiated fine movement from an idle state in single-trials. This technique achieved significantly lower classification error than the best single classifier and some conventional classifier fusion methods. The first study also confirmed the detection of the self-initiated fine movements on a single-trial basis, at above-chance levels, starting 396 ms before their motoric realization. The highest detection accuracy was 82.21% at 250 ms before movement. The second study formulated a novel pipeline of spatio-temporal filtering (PSTF) feature extraction method for laterality classification of self-initiated fine movements. The PSTF achieved significantly higher average classification accuracy (74.99%) than two conventional methods. This second study achieved a significant improvement in single-trial fine movement laterality prediction from RP features alone. The final study investigated a ternary BCI based on a single-trial RP of self-initiated fine movements. The study extended the PSTF for feature extraction in the ternary case, and proposed a diversified classification algorithm for multiclass binary decomposition. The proposed approach achieved significantly higher average classification accuracy (96.24% for idle class; 77.79% for left class; 72.37% for right class) than the conventional multiclass classifier and popular binary decomposition methods. This thesis demonstrates the potential of a ternary BCI based exclusively on RP features. It supports further study of RP-based BCI for intuitive asynchronous control of the environment, such as in augmentative communication or wheelchair navigation.
Source: TSpace (University of Toronto)
Author: Elias Abou Zeid
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W2788585236