TYPE Editorial

PUBLISHED 09 February 2023

DOI 10.3389/fnbot.2023.1140508

Editorial: Brain-computer interface

OPEN ACCESS and its applications
EDITED AND REVIEWED BY
Alois C. Knoll,
Technical University of Munich, Germany Duo Chen1*, Ke Liu2 , Jiayang Guo3 , Luzheng Bi4 and Jing Xiang5
*CORRESPONDENCE 1
School of Artificial Intelligence and Information Technology, Nanjing University of Chinese Medicine,
Duo Chen
Nanjing, China, 2 School of Computer Science and Technology, Chongqing University of Posts and
[email protected]
Telecommunications, Chongqing, China, 3 National Institute for Data Science in Health and Medicine, Xiamen
RECEIVED 09 January 2023 University, Xiamen, China, 4 Institute of Mechatronic Systems, Beijing Institute of Technology, Beijing, China,
5
ACCEPTED 19 January 2023 Magnetoencephalography (MEG) Center, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH,
PUBLISHED 09 February 2023 United States

CITATION
Chen D, Liu K, Guo J, Bi L and Xiang J (2023)
KEYWORDS
Editorial: Brain-computer interface and its
applications. Front. Neurorobot. 17:1140508. brain-computer interface (BCI), deep learning, machine learning, artificial intelligence,
doi: 10.3389/fnbot.2023.1140508 Electroencephalogram (EEG)
COPYRIGHT
© 2023 Chen, Liu, Guo, Bi and Xiang. This is an
open-access article distributed under the terms
of the Creative Commons Attribution License
(CC BY). The use, distribution or reproduction
in other forums is permitted, provided the
Editorial on the Research Topic
original author(s) and the copyright owner(s) Brain-computer interface and its applications
are credited and that the original publication in
this journal is cited, in accordance with
accepted academic practice. No use,
distribution or reproduction is permitted which Currently, brain-computer interface (BCI) is the research focus and hotspot in the field
does not comply with these terms.
of neuroscience. Related technologies are widely used in various scenarios such as clinical
use, rehabilitation, engineering, and daily life. The BCI uses different brain signals, recording
methods, and signal-processing algorithms to build a link between the brain and external
software/hardware platforms. With the development of hardware (e.g., BCI chip, wearable
device) and algorithms (e.g., machine learning, deep learning), BCI is becoming more practical
and stable.
We publish this Research Topic to collect the latest research worldwide in BCI. Researchers
from all over the world actively participate and contributed a lot of manuscripts. After carefully
and professionally reviewing all submissions, 14 high-quality manuscripts are accepted.
In this topic, several of the contributions focus on the use of deep learning in EEG
decoding for BCI, among which convolutional neural network (CNN) is the most widely used.
Zhang et al. propose a Multi-Scale 3D-CNN Approach for EEG-based Identity Authentication.
The experimental results show that the classification performance of the proposed framework
is excellent, and the multi-scale convolution method is effective to extract high-quality
identity characteristics across feature domains. Qiu et al. use Electroencephalogram (EEG) and
Functional Near Infra-red Spectroscopy (fNIRS) to track the brain activities evoked by neutral
and preferred music. The authors conclude that music can promote brain activities, especially in
the prefrontal lobe with preferred music. Deng et al. propose SparNet, a CNN composed of five
parallel convolutional filters and the Squeeze-and-Excitation Networks (SENet), to learn EEG
space-frequency domain characteristics and distinguish between depressive and normal control.
The computational results indicate that SparNet achieves a sensitivity of 95.07%, and a specificity
of 93.66%. Wang and Cerf combine common spatial pattern (CSP) features and radial basis
function neural network (RBFNN) to classify motor imagery EEGs. The algorithm provided
high variability within- and across-subjects in EEG-based action decoding. The computational
accuracies are higher or close to 90% on two datasets, i.e., BCI competition IV-2a, and -2b. Chen
L. et al. transform the EEG signals into symmetric positive definite (SPD) matrices and captures
the features of SPD matrices by using CNN. Meta-transfer-learning (MTL) is used to avoid the
time-consuming calibration. Chen G. et al. explore the feasibility of an audio-assisted visual BCI
speller and a deep learning-based single-trial event-related potentials (ERP) decoding strategy.
A spatial-temporal attention-based CNN (STA-CNN) is proposed to recognize the single-trial
ERP components. The average classification accuracy of STA-CNN is 77.7% in the EEG dataset
recorded from 10 subjects.

Frontiers in Neurorobotics 01 frontiersin.org

Chen et al. 10.3389/fnbot.2023.1140508

Other studies use traditional machine learning and statistical Wang Y. et al. classify the MEG data of patients with complex partial
methods for BCI applications in certain modes. Xu et al. use a seizures (CPS) or simple partial seizures (SPS), using support vector
simulated driving platform with an EEG data collection system machine. The algorithm obtained a classification accuracy higher
for the evaluation of human trust in autonomous vehicles. The than 80%. Cui et al. review the music-emotion recognition and
graphic theoretical analysis illustrates how human trust varies in analysis based on EEG signals.
EEG under semi-autonomous or fully autonomous driving modes. Overall, we hope that this topic can provide some references and
Wang J. et al. use non-linear dynamics of EEG Signals to Classify novel ideas for researchers in BCI. It should be emphasized that for
primary hand movement intent under opposite hand movement. such a rapidly developing research field, the work that has been done
Their experimental results show significant differences in movement- so far is only a drop in the ocean. The manuscripts we collect this time
related cortical potentials between hand movement directions under can only be a small leaf in the Amazon rainforest. For BCI, there is
an opposite hand movement. The results may lay a foundation for still a big gap between the current research and the actual use. Things
the future development of EEG-based human augmentation systems that need to be solved go far beyond.
for both the handicapped and the healthy. Massé et al. investigate
EEG-based alarm detection in the flight simulator. Cognitive fatigue
and cognitive load are manipulated to trigger inattentional deafness, Author contributions
and brain activity is recorded via EEG. The results show that alarm
omission and alarm detection can be classified based on time- All authors listed have made a substantial, direct, and intellectual
frequency analysis of EEG. Triana-Guzman et al. combine filter bank contribution to the work and approved it for publication.
CSP (FBCSP) and regularized linear discriminant analysis (RLDA)
for decoding EEG rhythms offline and online during motor imagery
for standing and sitting. The mean accuracy is higher than 80% in Conflict of interest
offline analysis, and higher than 90% in online experiments. Song
et al. propose a rehabilitative motor imagery BCI system that focuses The authors declare that the research was conducted in the
on rejecting false positive (FP) detection in stroke rehabilitation. A absence of any commercial or financial relationships that could be
two-phase classifier is used to reject the FP. The algorithm achieved construed as a potential conflict of interest.
71.76% selectivity and 13.70% FP rate by using only four EEG
channels in the patient group with stroke. Kwon et al. propose a
new hybrid visual stimuli for steady-state visual evoked potential Publisher’s note
(SSVEP)-based BCI, which incorporate various periodic motions into
conventional flickering stimuli (FS) or pattern reversal stimuli (PRS). All claims expressed in this article are solely those of the
Results demonstrate that FS with sine-wave periodic motion and PRS authors and do not necessarily represent those of their affiliated
with square-wave periodic motion could effectively improve the BCI organizations, or those of the publisher, the editors and the reviewers.
performances compared to conventional FS and PRS. Any product that may be evaluated in this article, or claim that may
Finally, two studies provide clinical and human factor be made by its manufacturer, is not guaranteed or endorsed by the
applications of magnetoencephalography (MEG) and EEG. publisher.

Frontiers in Neurorobotics 02 frontiersin.org