Efficient Algorithms for a Brain Computer Interface: Performance Studies

The design of a reliable Brain Computer Interface (BCI) is a very interesting topic in the Assistive Technology field because it would allow to people with very severe limitations to interact with his environment using electronic devices. This paper describes an architectural proposal for a BCI along with the results of a full test set. Proposed architecture has a three step structure: preprocessing, feature extraction and classification; tests have been conducted with twelve subjects of different gender and ages. Analysis of the data obtained in tests shows that the best performance in almost every circumstance can be achieved with a combination of Fast Fourier Transform and Wavelets (FFT-WT) as feature extraction step and Support Vector Machines (SVM) as classification step.