Improving detection and classification of steady-state visual evoked potentials in brain-computer interfaces