Systems, Man & Cybernetics - January 2016 - 12

Trials

alv. 1

alv. 2

ret. 1

ret. 2

velar 1

velar 2

Cues

Fp1
50 µV
Fp2
T7
T8
O1
O2

Finally, we are also expecting that
t he su r veyed GK P pat ter n s ca n be
exploited for the silent speech recognition (SSR) tech n ique [14]. SSR is a
speech recognition technique that does
not use acoustic information from the
user's speech [25]. Instead of acoustic
sound recorded from a microphone, it
records and analyzes speech-related
bioelectric potentials. Prev ious SSR
research has mainly focused on EMG
analysis to trace articulatory muscle
activities. However, if GKP analysis is
combined with conventional SSR techniques, the SSR interface could exploit
information about not only the muscle
activities, but also the tongue position,
which is a useful feature for discriminating the category of consonants.

Conclusion
In this a r ticle, we summa r ized our
efforts to reveal the hidden potential of
GKP as a new tool to trace tongue moveFigure 4. the gKP response for language-related tongue movements
ments. We reviewed GKP by describing
(alv.: alveolar, ret.: retroflex). Phonetic symbols were given to the
subject, and he was asked to articulate the presented symbols. the
its pattern for horizontal and languageupward movement to articulate the alveolar consonant /n/ and the
related tongue movements. We believe
backward movement to articulate the velar consonant /g/ evoked
that explained GKP patterns can be utithe potential decrease in the frontal regions (Fp1 and Fp2) and
lized in various ways to develop new
the potential increase in the occipital regions (O1 and O2). On the
ma n-machine inter face techniques.
other hand, when the tip of the tongue curled back to articulate the
retroflex consonant /1/, the potential levels in the frontal regions
First, it can be exploited to develop a
increased, and the potential levels in the occipital regions decreased. new assistive technology for persons
with limb motor disabilities. The tongue
is
directly
connected
to the brain by cranial nerves and
Second, the investigated GKP patterns can be utilized
generally escapes severe damage in spinal cord injuries
to develop a sophisticated GKP removal technique for lan[2]. Therefore, the TMI can be applied to many tetrapleguage-related EEG analysis. The use of EEG in language
gic patients who still have motility of the tongue. Second,
production research has been widely studied by neurosciit can provide a new interface device for ordinary people,
entists [19], and they investigated and discovered various
which has fine-controllability able to recognize smooth
language-related event-related potentials [20], [21]. To
and accurate tongue movements. Finally, it can assist
avoid contamination from the artifacts, various experispeech recognition by providing information about lanmental procedures have been developed to replace the
guage-related tongue movements. We believe that there
overt speech in a covert manner, such as phoneme monicould be more fascinating applications that we have not
toring [22], and a covert (or tacit) naming [23]. The covert
conceived yet that can greatly improve the quality of
speech approach can avoid artifacts from speech movehuman-computer interaction.
ments, but it is reported that there is a difference between
the brain activities for covert and overt speech in [19], [24].
Acknowledgments
Vanhatalo et al. exploited a more direct GKP removal techThis research was supported by Basic Science Research
nique using a latex sheet to cover the surface of the tongue
Program through the National Research Foundation of
for insulating the negative charge of the tongue [7]. Their
Korea (NRF) funded by the Ministry of Education
method can successfully eliminate GKP in an inexpensive
(2014R1A6A3A03059354), Institute for Information and Comand practical way, but adhering the sheet into parts of the
munications Technology Promotion (IITP) grant funded by
mouth can cause discomfort, and it may not be hygienic.
the Korea government (MSIP) [B0101-16-0307, Basic SoftWe believe that the GKP patterns explained in this article
ware Research in Human-level Lifelong Machine Learning
can be utilized to develop a sophisticated spatial filter to
(Machine Learning Center)], and National Research Founactively remove GKP from the recorded EEG signals for
dation (NRF) of Korea (NRF-2013R1A2A2A01067464).
overcome these limitations.
6

18
Time (s)

IEEE SyStEmS, man, & CybErnEtICS magazInE Janu ar y 2016

Table of Contents for the Digital Edition of Systems, Man & Cybernetics - January 2016