Classification of imagined geometric shapes using EEG signals and convolutional neural networks

Abstract

Brain-Computer interface systems allow the recognition of neuronal activity to create a direct communication channel between the brain and the outside world without using the peripheral nervous system. Many of the paradigms used are based on the detection of motor imagery and evoked potentials. In this work we study the use of visual imagery of geometric shapes to build a non-invasive BCI using convolutional neural networks and the black hole search algorithm. Our algorithm computes, for seven people, an average of 35% correctly when classifying 7 geometric shapes, and 70% when classifying the best two ones (line and parallelogram). The same technique has been tested on a set of data collected on eleven subjects when classifying 2 geometric shapes, and a maximum accuracy of 82% was obtained.