Home // GLOBAL HEALTH 2022, The Eleventh International Conference on Global Health Challenges // View article
Authors:
Thibault Porssut
Alix Gouret
Solène LeBars
Keywords: VEP, SSVEP, BCI, VR, EEGNet, Deep Learning
Abstract:
One of the most commonly used non-invasive Brain- Computer Interface (BCI) paradigm for virtual reality control relies on particular brain signals: Visually Evoked Potentials (VEP). However, the optimization of virtual 3D targets is required in order to conciliate satisfying VEP induction - leading to high classification accuracy - and visual discomfort minimization. This constitutes a real challenge that could unlock new possibilities for rehabilitation, gaming or other applications. In the current experiment, we designed 30 original 3D-stimuli by combining particular visual patterns with various 8Hz-movements. The objectives were (1) to test new associations of stimuli for better BCI-VR ergonomics and (2) to test a new paradigm of VEP-based BCI that discriminates stimuli according to their visual features (e.g., motion type) without exploiting any variation in flickers’ frequency (constant frequency = 8Hz). Offline classification abilities were assessed using an EEGNet deep learning model. The results suggested the possible role of the stimulation patterns on the visual fatigue induced. The EEGNet model successfully classified all the 30 stimuli with a high level of accuracy (97.58%). This development broadens VEP-BCI stimulation possibilities and could allow overcoming the problem of epileptogenic frequencies by exploiting visual properties of targets instead of frequency variations to discriminate VEP-BCI stimuli.
Pages: 22 to 26
Copyright: Copyright (c) IARIA, 2022
Publication date: November 13, 2022
Published in: conference
ISSN: 2308-4553
ISBN: 978-1-61208-995-9
Location: Valencia, Spain
Dates: from November 13, 2022 to November 17, 2022