Abstract— The steady-state visually evoked potential (SSVEP) is a particular response of the brain observed as an oscillating wave induced by repetitive visual stimulation. This paper focuses on the use of this paradigm for the construction of a brain-computer interface (BCI) system using different numbers of visual stimuli in different frequencies. The responses were analyzed for healthy subjects and stroke patients and show that a suitable choice of stimulation frequencies can be of paramount importance insofar as the system performance is concerned. I. INTRODUCTION A brain-computer interface (BCI) is a communication and control system by means of which external devices receive commands directly from the cerebral cortex without involving the conventional output pathways of peripheral nerves and muscles [1]. BCIs have several applications, ranging from assistive technologies for users with motor disabilities and neurologic diseases to immersive games [2]. The steady-state visually evoked potential (SSVEP) is a largely used noninvasive paradigm for BCI [3] based on electroencephalography (EEG). Figure 1 schematically shows a BCI-SSVEP system, in which specific patterns are evoked in the visual cortex by means of a set of stimuli. Through an analysis of the generated brain signals, it is possible to identify the stimulus on which the subject was focusing his/her attention at a given instant. The characteristics of these visual objects such as form, color and flicker frequency are important in defining the quality of the evoked responses [4]. In the case of BCIs, if different stimuli appear simultaneously and each one is associated with a command, *Research supported by the Brazilian government agencies: FAPESP (São Paulo Research Foundation), FINEP (Funding Authority for Studies and Projects), CNPq (National Counsel of Technological and Scientific Development) and CAPES (Coordination for the Improvement of Higher Education Personnel). S. N. Carvalho is an Assistant Professor at the Federal University of Ouro Preto, and a PhD student at School of Electric and Computing Engineering (FEEC) of University of Campinas (UNICAMP) (phone: +55- 19-3521-3857; e-mail: sarah@dca.fee.unicamp.br). T. B. S. Costa is an MSc student at FEEC – UNICAMP. L. F. S. Uribe is a PhD student at FEEC – UNICAMP. D. C. Soriano is an Associate Professor at the Federal University of ABC. S. R. M. Almeida is a post-doc researcher at School of Medical Sciences (FCM) – UNICAMP. L. L. Min is a Professor at FCM – UNICAMP. G. Castellano is an Associate Professor at Gleb Wataghin Physics Institute of UNICAMP. R. Attux is an Assistant Professor at FEEC – UNICAMP. the user could control an application by focusing his/her gaze on the stimulus linked with the intended command. Due to personal idiosyncrasies related to physiological, morphological and cognitive qualities, the SSVEP responses present an inter-subject variability, even for the same stimulus and environment [5]. Hence, each user needs a custom BCI system for efficient operation. In this study, we investigated the variation of the VEP response among subjects testing six different visual stimuli. The main goals were to compare the performance of SSVEP- BCIs built according to custom and generic stimuli interfaces, and to investigate the accuracy of the classifier when the number of commands increases. Figure 1. Diagram of a BCI based on SSVEP. This study was carried out using a database generated with healthy and stroke patients, according to the experimental setup described in Section II. Section III brings an overview of the feature extraction, attributes selection and classification approaches; Section IV presents the results and the discussion and Section V contains the conclusions and final remarks. II. EXPERIMENTAL SETUP Eight volunteers (six males and two females) with an average age of 26.3 ± 3.3 years and two 69.5 ± 0.5 year old stroke male patients took part in the experiment. One patient (P1) was evaluated 27 days after ictus. He had had an ischemic stroke in the frontal lobe. The other patient (P2) was evaluated 89 days after the stroke. He had had an ischemic stroke in the frontal, parietal and temporal lobes in the right hemisphere, with motor deficit of hemiplegia. For both patients, stroke severity was assessed using the National Effect of the combination of different numbers of flickering frequencies in an SSVEP-BCI for healthy volunteers and stroke patients Sarah N. Carvalho, Thiago B. S. Costa, Luísa F. S. Uribe, Diogo C. Soriano, Sara R. M. Almeida, Li L. Min, Gabriela Castellano, Romis Attux