Development of a brain computer interface interface using multi-frequency visual stimulation and deep neural networks

This work proposes a Brain Computer Interface based on using multi-frequency visual stimulation and deep neural networks for signals classification. The use of multi-frequency stimulation, combined with a new proposed coding method codifying up to 220 commands, which could be used to create a large multi-command brain computer interface. The advantages this method for commands codification and classification performance is analyzed in a five commands Brain computer interface. The classification of the electroencephalographic signals used in the interface was performed using several algorithms. The outcomes reveal that the best classification algorithm is a deep neural network, which gives a classification accuracy of 97.78 %. This algorithm, also, allows establishing the most relevant features of the electroencephalographic signal spectrums for the classification and information extraction from the evoked potentials.