Exploring BMI recognition through resting state, free imagination, and visual stimulation-induced EEG

Abstract

In recent years, the development of Brain-Computer Interface (BCI) technology has attracted significant attention, particularly in the acquisition and application of electroencephalogram (EEG) signals. Most pertinent studies have utilized EEG to monitor neural activity states. Given that EEG does not provide direct insights into the physical states of body structures, many researchers have refrained from employing EEG to assess structural body states such as Body Mass Index (BMI). This study, for the first time, employs portable EEG detection devices combined with machine learning algorithms such as Logistic Regression (LR), Naive Bayes (NB), and Support Vector Machine (SVM) to process EEG features, enabling the application of EEG in identifying BMI. We collected EEG data from three different states: resting state, free imagination, and viewing pictures. The results demonstrate that these EEG signals can effectively differentiate subjects with higher BMI. Particularly under visual stimulation, EEG features exhibited optimal performance when subjects viewed images of high-calorie foods, with the model achieving an average accuracy of 80% and an F1 score of 75%. From the perspective of frequency bands, within a single frequency band, the β band performed notably well in classification. However, the model constructed by integrating features from various frequency bands yielded the best classification performance. These comprehensive paradigms are expected to fill the current research gap in the application of EEG and provide new methodological perspectives for further studies.