Estimating brain periodic sources activities in steady-state visual evoked potential using local fourier independent component analysis

Objective: Steady-State Visual Evoked Potentials (SSVEPs) are widely used in Brain-Computer Interfaces (BCIs) applications. This research aims to estimate the brain sources' activities which are corresponding to these signals. Methods: While in SSVEPs, the stimulus frequency could modulate the EEG signal, sparsity in the frequency domain distribution of the signal may be seen. Consequently, the Fourier-ICA method may be appropriate for SSVEP source estimation. Moreover, because the stimulus is visual, the physiological information of the visual cortex region is exploited. So the Local Fourier Independent Component Analysis method is introduced in this research. Results: In order to assess the proposed method, two online available SSVEP datasets such as "Dataset BCI EEG SSVEP for four classes of stimuli" and SSVEP part of "EEG dataset for three BCI paradigms" were utilized. K means clustering of independent components extracted using the Fourier-ICA algorithm revealed brain activity in occipital and parietal regions in response to flickering lights. The Local Fourier-ICA algorithm showed acceptable performance with the maximum average classification accuracies of 98.4% and 99.95% for the first and second datasets, respectively. Conclusions: The results in the channel domain do not have a remarkable difference from the ones in the IC domain. Therefore, for stimulus detection and BCI application purposes, using the information of the EEG signals of the channels in the same cluster as Oz is preferred to reduce computational load and save time.