Prediction of Individual User’s Dynamic Ranges of EEG Features from Resting-State EEG Data for Evaluating Their Suitability for Passive Brain–Computer Interface Applications

Abstract

With the recent development of low-cost wearable electroencephalogram (EEG) recording systems, passive brain–computer interface (pBCI) applications are being actively studied for a variety of application areas, such as education, entertainment, and healthcare. Various EEG features have been employed for the implementation of pBCI applications; however, it is frequently reported that some individuals have difficulty fully enjoying the pBCI applications because the dynamic ranges of their EEG features (i.e., its amplitude variability over time) were too small to be used in the practical applications. Conducting preliminary experiments to search for the individualized EEG features associated with different mental states can partly circumvent this issue; however, these time-consuming experiments were not necessary for the majority of users whose dynamic ranges of EEG features are large enough to be used for pBCI applications. In this study, we tried to predict an individual user’s dynamic ranges of the EEG features that are most widely employed for pBCI applications from resting-state EEG (RS-EEG), with the ultimate goal of identifying individuals who might need additional calibration to become suitable for the pBCI applications. We employed a machine learning-based regression model to predict the dynamic ranges of three widely used EEG features known to be associated with the brain states of valence, relaxation, and concentration. Our results showed that the dynamic ranges of EEG features could be predicted with normalized root mean squared errors of 0.2323, 0.1820, and 0.1562, respectively, demonstrating the possibility of predicting the dynamic ranges of the EEG features for pBCI applications using short resting EEG data.

Time series analysis of trial-to-trial variability of MEG power spectrum during rest state, unattented listening and frequency-modulated tones classification

“…Also, we see a lot of possible applications for it. The EEG characteristics prediction is helpful in versatile neurostimulation applications, such as the brain-machine interface based techniques [1, 13, 9] or other EEG-cooperating methods like electrical brain stimulation [36]. Methods of predicting changes in EEG, especially when applied online, can be of a great clinical utility.…”

Deep Neural Regression Prediction of Motor Imagery Skills Using EEG Functional Connectivity Indicators

“…As regards the prediction model, several regression methods are available for prognosticating MI accuracy from neurophysiological variables like simple and multiple linear regression [ 31 , 32 ], stepwise regression [ 33 ], kernel regression [ 34 ], and (kernel) support vector machine regression [ 35 ], among others. Additionally, there is increasing use of regression approaches with neural networks that can be applied to the raw EEG data, simplifying BCI’s design pipelines by removing the need to extract features manually.…”