Time Delay Neural Network with Fourier transform for multiple channel detection of Steady-State Visual Evoked Potentials for Brain-Computer Interfaces

In BCI (Brain-Computer Interface) systems, brain signals must be processed to identify distinct activities that convey different mental states. For example, in BCIs that rely on Steady-State Visual Evoked Potentials (SSVEP) recognize EEG activity that reflects attention to a particular stimulus. In this paper, we propose a new technique for the classification of SSVEP-activity for non-invasive BCI. The proposed method is based on a Time Delay Neural Network that includes a Fourier transform in order to switch between the analysis of layers in the time domain to layers in the frequency domain. The first step allows the creation of different channels. The second step is dedicated to signal processing of these channels in the frequency domain. The last step is the classification. The presented results on offline processing correspond for electroencephalographic (EEG) signals obtained with 6 electrodes on 3 subjects. We compare our method on two time segments with the average combination method. These results outperform the average combination method and are promising for online processing.