A comprehensive benchmark dataset for SSVEP-based hybrid BCI

Abstract

The use of hybrid brain–computer interface (HBCI) systems have received much attention in recent years. However, there are few published datasets for HBCI systems. This paper provides a benchmark dataset for these systems. The provided dataset consist of data corresponding to three speller systems, including SSVEP-based BCI system and HBCI systems based on SSVEP-EMG and SSVEP-EOG. These spellers have been implemented using electroencephalogram (EEG) and other biosignals, including electromyogram (EMG) and electrooculogram (EOG). The virtual keyboard of spellers was composed of 9 visual flickers with stimulation frequencies ranged from 5.58 Hz to 11.11 Hz. The stimuli presentation was controlled by the psychophysics toolbox of Matlab. We conducted three experiments, using three speller systems under the similar situation. Ten subjects participated in a copy-spelling task experiment. For each system, the data included ten sessions corresponding to 10 times, which the subject spells the determined phrase. These HBCI systems were validated using NASA-TLX workload index and performance evaluation parameters, including the accuracy and the information transfer rate. Results showed that combining two control signals only when signals are derived from one organ of the body, might increase the workload. These datasets can be used as a comprehensive dataset to evaluate the performance of different HBCI systems. They are freely available from Http://maleki.semnan.ac.ir/.