< Terug naar vorige pagina

Publicatie

Brain-Computer Interfacing based on Time-Domain EEG Response Detection combined with Beamforming

Boek - Dissertatie

Patients suffering from severe motor- and communication disorders are no longer able to interact with their surroundings, and even communication with their families becomes nearly impossible. Due to this disability, they highly depend on caregivers with whom the communication is often troublesome or laborious. Therefore, modern technology that would be able to re-establish a reliable communication channel could considerably improve their quality-of-life and decrease the burden on the caregiver. Brain-computer interfaces (BCIs) are one of such communication channels. They rely on the analysis of brain activity to decode a subject's intention, which can subsequently be performed by a computer. At the time of writing, the BCIs that provide the best results in terms of accuracy, speed and convenience are based on visual stimulation (i.e., flashes, often displayed a computer screen) of a number of selectable targets (e.g., words, letters, symbols) which elicits well-known brain signals that can be used to deduce the intended target. In this thesis, we investigated the feasibility of the spatiotemporal beamforming principle for target selection (decoding) in the context of BCI, and assessed its performance for three widely-adopted visual paradigms. For all three paradigms, the classifier based on spatiotemporal beamforming either outperformed or performed on par with the respective state-of-the-art classifiers. Furthermore, the proposed decoding algorithm is in most cases computationally more efficient, thus requiring less expensive resources (i.e., hardware), making it well suited for retraining during the use of the BCI. In addition to the investigation of the decoding performance of the spatiotemporal beamformer, we investigated the cortical representation of the steady-state visual evoked potential (SSVEP), a widely-adopted paradigm in different fields, including BCI. Using a subdural grid located over the occipital cortex, we showed that the foveal flickering stimulus is represented at the posterior part of the primary visual cortex (V1), with little variation across the stimulation frequencies used (11 to 15 Hz). The peripheral flickering was represented at locations consistent with retinotopic mapping studies. These results give new insights into the possible alternative interpretations of the previous SSVEP-based studies.
Jaar van publicatie:2018
Toegankelijkheid:Open