Vargas, Guilherme VettorazziLeite, Sarah Negreiros de CarvalhoBoccato, Levy2022-10-102022-10-102022VARGAS, G. V.; LEITE, S. N. C.; BOCCATO, L. Analysis of the spatiotemporal MVDR filter applied to BCI-SSVEP and a filter bank extension. Biomedical Signal Processing and Control, v. 73, p. 103459-103469, 2022. Disponível em: <https://www.sciencedirect.com/science/article/pii/S1746809421010569>. Acesso em: 29 abr. 2022.1746-8094http://www.repositorio.ufop.br/jspui/handle/123456789/15664Artifacts inevitably permeate brain signal acquisition by electroencephalography (EEG). Hence, brain-computer interfaces based on steady-state visually evoked potentials (BCI-SSVEP) frequently require a filtering to increase signal-to-noise ratio (SNR) in an attempt to improve its ability to identify a command selected by the user. By combining the signals from different electrodes, the spatiotemporal filtering technique based on the minimum variance distortionless response (MVDR) attenuates undesired frequency components while preserving the spectral content at the visual stimuli frequencies. In this study, we revisit the MVDR filter, further evaluating its behavior with respect to critical factors in a BCI-SSVEP system: proximity amid stimulation frequencies, number of stimuli and stimulation window-length. Additionally, the main parameters of the filter were also varied, such as the order and the number of electrodes to be combined. The experimental analysis confirmed the effectiveness of the MVDR filter for the majority of the scenarios. However, it also revealed a significant difficulty that the MVDR filter has when dealing with short-length time windows, especially when compared with classical filtering techniques, such as CAR and CCA. So, in order to mitigate this limitation, we propose a filter bank MVDR (FBMVDR), where each element is a MVDR filter designed to preserve a single stimulation frequency or harmonic components. This new approach provided an increase of more than 5% in relation to the standard MVDR, reaching a performance of 92.6% in scenarios with 4 visual stimuli and 1s window-length and achieved competitive results with the state-of-the-art technique filter bank CCA (FBCCA).en-USrestritoBrain-computer interfaceSteady-state visually evoked potentialMinimum variance distortionless responseSpatiotemporal filteringArtigo publicado em periodicohttps://www.sciencedirect.com/science/article/pii/S1746809421010569https://doi.org/10.1016/j.bspc.2021.103459