• Expression of Interest

    Lodz University of Technology, Institute of Electronics

    Instead of using the classical spectrum analysis method to extract steady-state visual evoked potentials (SSVEPs) from the recorded brain activity of the interface user, we developed an alternative based on Canonical Correlation Analysis. This modification allows finding the best bipolar combination of measurement channels (one used as a signal channel, the other as a reference).

    The proposed detection algorithm improves classification accuracy in the groups of subjects with the average and poor results (rated based on classical spectrum analysis method). In the group of users with the best results, there was no clear improvement of the SSVEP detection accuracy. At the same time, average detection times were not significantly better, but the measured overall information transfer rates were in many cases higher for the proposed method, which is due to the higher classification accuracy of this method. It should be noticed that only a short off-line calibration session was necessary to achieve such results.

    Due to a high user SSVEP variation, BCI illiteracy phenomenon, and a low communication speed many BCI systems are still at the stage of laboratory demonstrations or small-scale clinical trials with promising results for patients with various neurological diseases. In our opinion, these limitations may be eliminated in the following years by the improvement of portable EEG technologies (enabling reliable and effective use of an EEG recorder in home environments) as well as further development and optimization of BCI algorithms, spatial filtering and detection methods so that they can work reliably in uncontrolled environments.

    {Empty}