A novel stimulation method for multi-class SSVEP-BCI using intermodulation frequencies

Abstract

Objective. Steady-state visual evoked potential (SSVEP)-based brain–computer interface (BCI) has been widely investigated because of its easy system configuration, high information transfer rate (ITR) and little user training. However, due to the limitations of brain responses and the refresh rate of a monitor, the available stimulation frequencies for practical BCI application are generally restricted. Approach. This study introduced a novel stimulation method using intermodulation frequencies for SSVEP-BCIs that had targets flickering at the same frequency but with different additional modulation frequencies. The additional modulation frequencies were generated on the basis of choosing desired flickering frequencies. The conventional frame-based ‘on/off’ stimulation method was used to realize the desired flickering frequencies. All visual stimulation was present on a conventional LCD screen. A 9-target SSVEP-BCI based on intermodulation frequencies was implemented for performance evaluation. To optimize the stimulation design, three approaches (C: chromatic; L: luminance; CL: chromatic and luminance) were evaluated by online testing and offline analysis. Main results. SSVEP-BCIs with different paradigms (C, L, and CL) enabled us not only to encode more targets, but also to reliably evoke intermodulation frequencies. The online accuracies for the three paradigms were 91.67% (C), 93.98% (L), and 96.41% (CL). The CL condition achieved the highest classification performance. Significance. These results demonstrated the efficacy of three approaches (C, L, and CL) for eliciting intermodulation frequencies for multi-class SSVEP-BCIs. The combination of chromatic and luminance characteristics of the visual stimuli is the most efficient way for the intermodulation frequency coding method.

A Hybrid Speller Design Using Eye Tracking and SSVEP Brain–Computer Interface

“…However, recent advances in SSVEP-based BCIs have developed techniques and algorithms to overcome these obstacles. Several studies have proposed different stimulus design techniques to realize large number of targets with only few frequencies, but they have also suffered limitations [21,38,43,50,57,58,90]. Furthermore, a few researchers combined SSVEPs with P300 to generate more targets with less frequencies, but this was at the cost of task complexity, which eventually affected the performance of the system.…”

“…Nakanishi et al developed a task-related component analysis technique to develop a high speed speller with an average accuracy of 89.93% and an average ITR of 325.33 bits/min [29]. In a recent study, Maye et al proposed a multi-target SSVEP-based BCI system that only uses single flickering stimulus [37].The stimulation frequencies ranging between 5 and 90 Hz could be used to elicit SSVEPs, but only few frequencies could be used due to the technological constraints of the current systems [38]. For instance, the conventional frequency coding method can only generate specific frequencies due to the limitation posed by the monitor refresh rate [39].…”

“…These methods include phase coding techniques [44,45], dual frequency stimuli [47,48], variable frame rate stimuli [49], multiple frequency sequential coding [50], amplitude modulation techniques [51,52], intermodulation frequencies [38,53,54], varying duty cycles [41,55], interpolation techniques [39], and joint frequency and phase modulation [28], but these techniques also have limitations. For instance, the number of phase lags are also dependent on the number of frames of the stimulus frequency in the phase tagging method, consequently resulting in a limited number of targets [38]. In the case of multiple frequency coding techniques, the number of targets increases as the cycle period increases, and the classification time window is proportional to the amount of stimulation frequencies that further decreases the overall BCI performance [38,50].…”

“…For instance, the number of phase lags are also dependent on the number of frames of the stimulus frequency in the phase tagging method, consequently resulting in a limited number of targets [38]. In the case of multiple frequency coding techniques, the number of targets increases as the cycle period increases, and the classification time window is proportional to the amount of stimulation frequencies that further decreases the overall BCI performance [38,50]. Furthermore, it has also been shown that low-frequency stimuli can cause photosensitivity-based epileptic seizures and high-levels of visual fatigue and discomfort, especially when presented with large number of targets and longer periods [21,43].…”

“…The stimulation frequencies ranging between 5 and 90 Hz could be used to elicit SSVEPs, but only few frequencies could be used due to the technological constraints of the current systems [38]. For instance, the conventional frequency coding method can only generate specific frequencies due to the limitation posed by the monitor refresh rate [39].…”