Decoding Steady-State Visual Evoked Potentials From Eeg Signals: Towards An Eeg-Triggered Fes System To Restore Hand Grasp Function

It has been widely reported that patterns of EEG generated when a person performs a mental strategy can be recognized by signal processing algorithms. Among those mental strategies are the EEG-based brain-computer interface (BCI) paradigms. Furthermore, recognized patterns can be used as a source of information for communication to operate devices of BCI. Steady-State Visually Evoked Potentials (SSVEP) is a BCI paradigm that uses EEG brain responses when a subject focuses on a visual stimuli (flickering stimuli). Decoding SSVEP signals refers to identify what stimulus the user focuses on, which could be used as a command for communication or control. The minimum energy combination (MEC) and canonical correlation analysis methods (CCA) have been used in SSVEP-based BCIs due to its high efficiency, robustness, and simple implementation. In the last years, variants of CCA-based SSVEP methods have been reported in literature to improve classification and usability such as filter bank canonical correlation analysis (FBCCA). This paper evaluates the MEC, CCA and FBCCA methods for decoding commands from EEG signals in a SSVEP-based BCI application. It was carried out a set of experiments with five subjects which consist of four flickering stimuli (6.66, 7.5, 8.57 and 10 Hz) showed on a LED monitor. The results showed, for an epoch of 3 s, that CCA and FBCCA methods were able to detect SSVEP with high accuracy: 92.6% for FBCCA and 91.4% for CCA. The classification accuracy was 86.1% for MEC. As future work, FBCCA method will be used to decode user intention to control a closed-loop system based on EEG-triggered FES to restore hand grasp function.