BCI is a technology that provides people with a new communication tool controlled by brain activity instead of body movements.
BCI uses voluntary, reactive, and continuous brain activities to control external devices (computers or machines).
BCI helps with the rehabilitation of motor function and can be used for entertainment, such as brain games.
BCI requires a long amount of training time to collect training data for classification models. It significantly reduces user comfort.
BCI performance varies across subjects, which means some portion of subjects cannot operate BCI applications due to its very low performance even though they performed a long training phase.
Developing an end-to-end BCI for a larger user pool requires addressing long training time and performance variations
BCI performance predictors
Zero-training BCI
Stimulation for enhancing BCI performance
Won et al., 2019
BCI performance predictors could be used for pre-screening subjects whose performance is expected low before doing a long calibration phase.
BCI performance predictors could contribute to understanding the underlying neural mechanisms and be targeted for neuromodulation approaches.
P300 based speller performance predictors were developed from P300 multi-feature during the pre-performed RSVP task.
Won et al., 2021
Due to high subject-to-subject variability, pre-trained models from different subjects significantly reduced the classification performance.
In motor imagery BCI, previous studies reported that high- and low-performers have different neurophysiological characteristics, so using selective subjects can benefit zero-training BCI performance in a simple way.
Low-performers can be treated differently rather than applying a new feature extraction method (e.g., external stimulation for entraining relevant brain activities).
Experimental design (In preparation)
There are people who have difficulties in learning BCI and generating discriminative features, yielding as low BCI performance as random chance.
Brain activity entrainment could be another solution instead of extracting hidden features using a complex feature extraction algorithm.
Sensory stimulation and tDCS are compared in terms of brain activity changes during motor imagery