Table 4.
Important reactive hybrid brain–computer interface studies (from 2010 to 2016).
| Reference | Brain area | Activity | Modality | Application | Analysis type | Classifier | Commands | Accuracy | Window size |
|---|---|---|---|---|---|---|---|---|---|
| Yin et al. (2013) | Parietal and occipital cortex | P300 and steady-state visual evoked potential (SSVEP) | Electroencephalography (EEG) | Speller | Online | SW-LDA | Up to 36 | 93.85% using hybrid paradigm | All rows and columns were flashed in 2.88 s |
| Zimmermann et al. (2013) | Motor cortex | Isometric finger-pinching task | fNIRS + bio-signals (ECG) | Feasibility for BCI | Offline | Hidden Markov model (HMM) | 1 | 88.5% | 5–20 s |
| Li et al. (2013) | Whole brain | SSVEP and P300 | EEG | Wheelchair control | Online | Support vector machine (SVM) | 6 | >80% | 0–0.6 s after a button flash complete for P300 and 3.2 s for SSVEP |
| Xu et al. (2013) | Whole brain | SSVEP and P300 | EEG | BCI speller for target selection | Online | SW-LDA | 9 | 93.3% for P300 + SSVEP-B | 0–0.8 s after the onset |
| Bi et al. (2014) | Parietal and occipital cortex | P300 and SSVEP | EEG | Speed and direction for cursor control | Online | SVM | 4 | >90 | 4 s |
| Aziz et al. (2014) | Frontal and occipital | Eye movements | EEG + electrooculography (EOG) | Automated wheelchair navigation | Online | SVM, HMM | 5 | 98% | 0.5 s |
| Li et al. (2014) | Motor and occipital | Motor imagery and SSVEP | EEG | Wheelchair control | Real time | SVM | 6 | − | − |
| Witkowski et al. (2014) | Motor cortex | Hand-grasping motion assisted with exoskeleton | EEG + EOG | Assistive rehabilitation applications | Online | Sensitivity index | 4 | Average accuracy 62.28% for two conditions | 5 s |
| Putze et al. (2014) | Auditory and visual cortex | Visual and auditory stimuli | EEG + functional near infrared spectroscopy (fNIRS) | Application to patient choice selection | Online | Linear discriminant analysis (LDA), SVM | 2 | 94.7% average | Four window sizes 1, 2, 4, 8, and 16 s |
| Tomita et al. (2014) | Visual cortex | SSVEP-based task | EEG + fNIRS | Optimal window selection for hybrid EEG–NIRS | Offline | − | 1 | 85% average accuracy (in 10 sec optimal window) | 0–10 s |
| Fan et al. (2015) | Parietal and occipital | SSVEP and P300 | EEG | Vehicle destination selection system | Online | LDA | 11 | 99% | 0–0.51 sec from onset for P300 and 8 s for SSVEP |
| Ma et al. (2015) | Parietal and occipital | P300 and eye blink | EEG + EOG | Mobile robot control | Real time | LDA | 9 | 87.3% for average of five trials | ~1.6 s |
| Combaz and Van Hulle (2015) | Whole brain | P300 and SSVEP | EEG | Applications to locked-in patients option selection | Online | SVM | 12 | Maximum achieved > 95% | 200 ms before stimulation to 800 ms after stimulation for experiment 1 |
| Wang et al. (2015) | Whole brain | P300 and SSVEP (shape changing and flickering-hybrid) | EEG | Development of new paradigm with application to devices control | Online | canonical correlation analysis (CCA), Bayesian LDA | 4 | Overall 20% increase in SSVEP classification, 100% for P300 | Flash start to the flash end for SSVEP, single flashes lasting 0.8 s for P300 |
| Ramli et al. (2015) | Motor and occipital | Eye gaze | EEG + EOG | Application to BCI applications (wheelchair control) | Online | Finite-state machine (FSM) | 6 | 97.88% | 0.5 s |
| Yin et al. (2015a) | Parietal and occipital cortex | P300 and SSVEP | EEG | Speller paradigm with applications to BCI systems control | Online | SW-LDA for P300, CCA for SSVEP | Up to 64 | 95.18% | 0.8 s epochs after stimulation |
| Kim et al. (2016) | Occipital | SSVEP and eye movement | EEG + EOG | Turtle movement control | Online | CCA | 4 | 83% for event-related desynchronization (ERD) and 92.7% for SSVEP | 2 s |
| Lin et al. (2016) | Occipital | SSVEP | EEG + EMG | Choice selection | Online | CCA | 2 | 81% | 0.5–5 s |