Page 146 - Handbook of Biomechatronics

Biomechatronic Applications of Brain-Computer Interfaces 143

Thus, the EOG can be considered a form of eye tracker. Furthermore, blinks
are easily identifiable as very large, brief changes in the signal value.
Perhaps the most common use of EOG is to remove blink artifacts from
EEG data using methods such as regression and independent component
analysis (Hong and Khan, 2017). However, many other interesting EEG-
EOG fusion approaches have been developed. For example, since eye mea-
sures such as blink frequency are correlated with workload and fatigue, they
can be used together with EEG-based workload indicators to obtain a more
accurate estimate of a person’s workload or fatigue (Khushaba et al., 2013;
Novak et al., 2015). Alternatively, EEG and EOG can be used as two inde-
pendent control channels: one command (e.g., raise/lower robotic arm) is
performed using EOG while the other (e.g., open/close robotic hand) is
performed using EEG paradigms such as motor imagery (Hortal et al.,
2015; Ma et al., 2014).
EEG and EOG can even be combined without the use of dedicated
EOG electrodes: since eye artifacts appear in the EEG, it is possible to esti-
mate EOG “traces” from EEG electrodes. For example, Ramli et al. (2015)
developed a wheelchair controller where EOG traces in EEG are used to
estimate whether the eyes are open or closed. If the eyes are closed, no
wheelchair movement is allowed; if the eyes are open, the wheelchair is con-
trolled based on the EEG. However, while this approach reduces the num-
ber of required electrodes, it is currently unclear whether the increase in
convenience is large enough to outweigh any decreases in BCI accuracy
caused by not having access to a “true” EOG signal.

1.4.3 EEG and Electromyography
EMG is the measurement of electrical signals generated by individual
muscles. Such electrical muscle activity frequently acts as a source of noise
in EEG: for example, EEG electrodes placed near the back of the head are
frequently contaminated by neck muscle EMG while EEG electrodes
placed near the front of the head are contaminated by jaw EMG. As with
EOG, the most common use of EMG in BCIs is thus to remove muscle
artifacts from the EEG. However, other sensor fusion methods exist and
are similar to those used to combine EEG and EOG. For example, one
input channel of a device can be controlled using EEG while the other
can be controlled using intentionally generated jaw EMG (Foldes and
Taylor, 2010).