A study on performance increasing in SSVEP based BCI application

• Main Authors: Erdem Erkan, Mehmet Akbaba
• Format: Article
• Language: English
• Published: Elsevier 2018-06-01
• Series: Engineering Science and Technology, an International Journal
• Online Access: http://www.sciencedirect.com/science/article/pii/S2215098617316518

People which are afflicted with neurological conditions or neurodegenerative diseases can’t control own muscles by neural pathways. Brain computer interface (BCI) systems offer these people another alternative path from their own neural pathways. This alternative pathway is the direct use of brain signals by a computer without using any vocal muscle. The steady state visual evoked potential (SSVEP) approach currently provides the high performance and reliable communication for the implementation of a non-invasive BCI. In SSVEP based BCI systems, Electroencephalography (EEG) signal detection time (signal window length) and accuracy are the most important performance parameters. Performance is usually measured by Information Transfer Rate (ITR).In the presented paper a SSVEP based BCI robot control application is introduced and system performance is analyzed for different signal window lengths. At first, the number of eye blinks of the subjects is determined by fast eye artifact detection method (FEAD) which based on visual eye blink detection. These eye blink counts are used for system activation. System usability is improved by this control. Two consecutive eye blinks which detecting by FEAD method are used for system activation. System deactivation is also provided by the same command. Synchronous and asynchronous experiments are performed on four healthy subjects for performance analyses. EEG data is analyzed in details by asynchronous experiments. During the synchronous experiments, subjects are tried to complete a predefined route which has twelve steps by navigating the robot (Lego Mindstorms EV3). The minimum energy combination (MEC) and canonical correlation analysis (CCA) methods are applied to EEG segments that are different in length in order to detect SSVEPs in both type experiments. ITR values are calculated for different signal window lengths. The results show that the detection accuracy of the MEC method is similar to that of the CCA method, although it is higher than that of the CCA method in situations where the SSVEP has low strength. In synchronous experiments, using MEC method a system peak ITR of 133.33 bit/min is reached for one subject with a 0.9 s signal window length. This ITR value is higher than previously published studies in the literature for SSVEP based BCI systems. Keywords: Brain computer interface (BCI), Steady state visual evoked potential (SSVEP), Minimum energy combination (MEC), Canonical correlation analysis (CCA)