Could the self-similarity exponent be an useful feature extraction method for electroencephalographic (EEG) signals in Brain Computer Interfaces (BCIs)?