The brain-computer interface (BCI) is an emerging technology that is maturing toward being more practically usable. The goal is to provide a communication channel bridging the human neural systems within the brain to the external world, via interactions with electronic technologies. For example, one such use would be to build communication or control applications for locked-in patients who have no control over their bodies. Recently, the potential application areas have been expanding from communication to marketing, rehabilitation, treatment, mental state monitoring, and entertainment. In the last few decades, machine learning algorithms have advanced BCI technology and the performance levels have been greatly improved in the context of classification accuracy. However, several issues remain to be resolved for BCI to be applicable in the real-world. These issues include the performance variation, long calibration time, and low level of reliability among others. Research focusing on new ‘development’ is expected to bring solutions in this respect.
Along with BCI development, deep learning has been applied in various fields such as computer vision and natural language processing, outperforming traditional machine learning approaches. Consequently, a great number of researchers in engineering, science, and other industries have shown interest in deep learning; convolutional neural network (CNN), recurrent neural network (RNN), and generative adversarial network (GAN).
Researchers in the BCI field are attempting to complement BCI with deep learning algorithms. However, there are some limitations as brain signals are high-dimensional, noisy, and very nonstationary. Additionally, datasets are substantially limited as compared to image data in computer vision fields. Thus, further research focusing on deep learning as applications to BCI and a thorough evaluation of how this application can be used in practice to implement the interface use would be beneficial.
The primary goal of this Research Topic is to create a forum of discussion, bringing together researchers’ contributions allowing for progress in deep learning-based-BCIs. Opinions/survey data addressing the practical issues in the application of deep learning to BCI are also welcome.
We welcome authors to submit Original Research, Review, and Mini-Review articles focusing on, but not limited to, the following subtopics:
1) Deep learning (CNN, RNN, GAN, etc.) for improving BCI performance
2) Transfer learning for BCI with minimum calibration
3) Data augmentation for limited training dataset
4) Explainable Artificial Intelligence making the solution interpretable and beneficial to understand brain dynamics
5) BCI applications that solved practical issues through deep learning techniques
6) Limitations and Critical issues in applying deep learning to BCI
7) Evidence and review of deep learning approaches in BCI
Dr. Ahn holds patents related to BCI systems. The other Topic Editors Dr. Cho and Dr. Yeom declare no conflicts of interest.
The brain-computer interface (BCI) is an emerging technology that is maturing toward being more practically usable. The goal is to provide a communication channel bridging the human neural systems within the brain to the external world, via interactions with electronic technologies. For example, one such use would be to build communication or control applications for locked-in patients who have no control over their bodies. Recently, the potential application areas have been expanding from communication to marketing, rehabilitation, treatment, mental state monitoring, and entertainment. In the last few decades, machine learning algorithms have advanced BCI technology and the performance levels have been greatly improved in the context of classification accuracy. However, several issues remain to be resolved for BCI to be applicable in the real-world. These issues include the performance variation, long calibration time, and low level of reliability among others. Research focusing on new ‘development’ is expected to bring solutions in this respect.
Along with BCI development, deep learning has been applied in various fields such as computer vision and natural language processing, outperforming traditional machine learning approaches. Consequently, a great number of researchers in engineering, science, and other industries have shown interest in deep learning; convolutional neural network (CNN), recurrent neural network (RNN), and generative adversarial network (GAN).
Researchers in the BCI field are attempting to complement BCI with deep learning algorithms. However, there are some limitations as brain signals are high-dimensional, noisy, and very nonstationary. Additionally, datasets are substantially limited as compared to image data in computer vision fields. Thus, further research focusing on deep learning as applications to BCI and a thorough evaluation of how this application can be used in practice to implement the interface use would be beneficial.
The primary goal of this Research Topic is to create a forum of discussion, bringing together researchers’ contributions allowing for progress in deep learning-based-BCIs. Opinions/survey data addressing the practical issues in the application of deep learning to BCI are also welcome.
We welcome authors to submit Original Research, Review, and Mini-Review articles focusing on, but not limited to, the following subtopics:
1) Deep learning (CNN, RNN, GAN, etc.) for improving BCI performance
2) Transfer learning for BCI with minimum calibration
3) Data augmentation for limited training dataset
4) Explainable Artificial Intelligence making the solution interpretable and beneficial to understand brain dynamics
5) BCI applications that solved practical issues through deep learning techniques
6) Limitations and Critical issues in applying deep learning to BCI
7) Evidence and review of deep learning approaches in BCI
Dr. Ahn holds patents related to BCI systems. The other Topic Editors Dr. Cho and Dr. Yeom declare no conflicts of interest.
