Evaluation of the Bremen SSVEP based BCI in real world conditions

Abstract

A brain-computer interface (BCI) provides the possibility to translate brain neural activity patterns into control commands without user's movement. The brain activity is most commonly measured non-invasively via standard electroencephalography (EEG), i.e., with electrodes placed on the surface of the scalp. In this article, we evaluate a BCI system based on steady-state visual evoked potentials (SSVEPs) in real world conditions. Although the performance of this type of BCI has already been proved by several research groups with healthy users in laboratory settings assisted by scientific researchers, there are still many difficulties in changing from demonstration systems to practical BCIs. The Bremen-BCI was evaluated in this case study with 37 naive subjects (without any SSVEP-BCI experience), including 8 handicapped users, at the international rehabilitation fair RehaCare2008. In spite of unsuitable environment conditions on the fair, the spelling tasks were successfully completed by 32 participants with a mean accuracy of over 92% and an average information transfer rate (ITR) of 22.6[bits/min]. No significant dependency of the physical disability of participants on the ITR could be observed.

Behavior-Based SSVEP Hierarchical Architecture for Telepresence Control of Humanoid Robot to Achieve Full-Body Movement

“…SSVEPs prominently appear over the visual cortex in the occipital region of the scalp [25]. The brain signal, , evoked by the SSVEP stimulus at the time is described by [26] ( 1) where is the flickering frequency of the visual stimulus, is the total number of stimuli, is the number of considered harmonics, and are the amplitude and the phase of each sinusoid, and includes noise, artifacts and any components irrelevant to the SSVEP response.…”

Riemannian geometry for EEG-based brain-computer interfaces; a primer and a review

“…The study of the applicability of BCI technology to clinical populations (e.g., [6][7][8][9][10][11][12][13][14]), for instance, in comparison (or combination) to other Assistive Technology (AT) such as eyetracking, electrooculography and contactors [15,16].…”

“…Since Cov(x1k, x2k) = Cov(x2k, x1k), the matrix Ck is symmetric, thus 14 it is determined by (N(N+1))/2 of its elements, three in this case. Therefore, we can represent covariance matrices Ck as data points in a 3D space with coordinates along the axes Var(x1), Var(x2) and Cov(x1, x2).…”

“…The mixing matrix depends on the dipole position and orientation in the brain, the physical properties of the head and on the position of the electrodes on the scalp. Once estimated B, the source process are estimated by (14), out of the usual scaling and permutation ambiguities [90,91]. Now, let Si and Sj be the covariance matrix of the unknown source process for any two trials.…”

“…where DA, DB and D are PxP diagonal matrices with positive diagonal elements dA,p, dB,p and dp, respectively, with p={1,…,P} and 1<P<N (8), thus the P most extreme eigenvalues forming this distance are associated with the eigenvectors forming the CSP filter F. The elements dA,p and dB,p are the variances of the data filtered by the CSP filter as per (14). Each associated eigenvector p f projects the data covariance in a mono-dimensional space where the ratio of the variance in the two classes is maximized (19), that is, the points are aligned so as to be maximally separated; the first P/2 vectors explains the maximum of the variance of class A and the minimum of class B, log ,…”