Information Technology Reference
In-Depth Information
13.4 Applications of Passive BCI
In this section, our aim is to present existing applications of passive BCI. These
applications can be categorized into four categories: implicit multimedia content
tagging, error correction, adaptive aiding and automation, and virtual environment
applications as displayed in Table
13.1
.
13.4.1 Implicit Multimedia Content Tagging
Passive BCI have been used for tagging multimedia contents (Shenoy and Tan
2008
; Kapoor et al.
2008
; Koelstra et al.
2009
). Shenoy and Tan used EEG activity
to classify images (Shenoy and Tan
2008
). They used ERP that occur in EEG
activity after image presentation. Their system was able to classify images matching
to human faces versus inanimate objects with a 75.3 % accuracy. For a three-class
classi
cation (human faces vs. animals vs. inanimate objects), an average accuracy
of 55.3 % was obtained. Kapoor et al. (
2008
) used these results and proposed to
combine BCI with a more classical recognition system. The experiment yielded
significant gains in accuracy for the task of object categorization. In the two
aforementioned works, users were not aware of the classi
cation task. They were
assigned
“
distractors task
”
to force them to look at the display. No feedback about
the classi
cation task was provided. This reinforces here the
implicit
property of the
interaction
.
Video content tagging has also been explored (Koelstra et al.
2009
). Koelstra
et al. proposed to use EEG brain activity to implicitly validate video tags. They
demonstrated that incongruent tags could be successfully distinguished by EEG
analysis. Recently, Moon et al. (
2012
) proposed to automatically extract interesting
parts of video clip by using EEG activity. They used the commercial Emotiv device
and one of the proprietary EEG index related to user emotional state named long-
term excitement. In another study related to multimedia content, Scholler et al.
(
2012
) proposed to use EEG activity and speci
cally P300 components to deter-
mine whether a change in video quality of multimedia clip occurred (the process is
done off-line).
13.4.2 Error Detection and Correction
The detection of ErrP provides a promising possibility to correct errors in different
contexts. For instance, the detection of ErrP has been used to correct error during
classical computer interaction task (Parra et al.
2003
) and to increase performance
of active BCI (Ferrez and Mill
n
2008
; Dal seno et al.
2010
).
Parra et al. (
2003
) use the detection of error potentials in brain activity to correct
errors in a visual discrimination task. In this study, the users had to push buttons
corresponding to visual stimuli. The user sometimes failed and perceived error
á
