The Brain-Computer Interface (BCI) seeks not only to understand, but also to optimize complex neural processes, establishing communication between the brain and an electronic device. Neuroscience applied to BCI involves studying brain signals to identify patterns associated with specific intentions, allowing the creation of algorithms capable of interpreting these intentions into commands to control devices, the evolution of this promising area stands out for boosting the understanding of brain processes and offering practical solutions, such as improvements in the quality of life for people with motor limitations. BCI systems based on Steady State Visually Evoked Potential (SSVEP) use brain responses to any visual stimulus flashing at a constant frequency as input command to an application or external device, although it is widely used for many applications, there are system characteristics that must be analyzed and discussed to increase the application's performance. This study addresses preprocessing, feature extraction and classification in the digital signal processing steps in an SSVEP-based BCI. The results include comparative analyzes of the extraction of epochs in five different sizes (2s, 1s, 500ms, 250ms, 125ms) for electroencephalogram (EEG) signals in a Brain-Computer Interface in front of stimuli at three different SSVEP frequencies (8Hz, 14Hz and 28Hz). Classification accuracies are presented for each analysis. The results obtained through the system classification reveal that epochs with longer durations present better performance. However, when analyzing shorter duration epochs, they have reasonable performance, offering efficiency for the scenario and providing a greater number of commands applicable in an BCI-SSVEP configuration.
|
