EEG Channel Selection using Fractal Dimension and Artificial Bee Colony Algorithm

The development of Brain Computer Interfaces (BCI) has attracted the attention of several research groups for solving different kind of problems in the field of medicine, device control, gaming, etc. However, there are several challenges that must be solved in order to have more feasible BCI applications. One of these challenges is related with the high dimensionality of the EEG recordings due to they are acquired from several channels to preserve high spatial accuracy. However, it is necessary to carefully select the channels that provide the most relevant information as well as the feature extraction technique in order to guarantee an acceptable accuracy. On the other hand, Swarm Intelligence techniques are based on biological processes of some species for their survival such as the food search task or natural selection process. These techniques have been widely applied in several optimizations problems obtaining acceptable results. In this paper, we described the high dimensionality challenge as an optimization problem in order to applied a swarm optimization technique, the Artificial Bee Colony (ABC) algorithm, to determine the set of channels that are more useful to discriminate different mental tasks. Furthermore, during the feature extraction stage, we applied fractal dimension methods to build the feature vector that will be used to train a classifier. Finally, the accuracy of the proposed methodology is tested classifying different motor tasks using the data set IVa from BCI international competition III.