Single-Subject vs. Cross-Subject Motor Imagery Models

Abstract

This paper compares the performance of machine learning algorithms trained and tested on single-subject EEG data compared to nine-person cross-subject EEG data from the BCI IV 2a dataset. To compare the performance of single-subject and cross-subject EEG models, we implement eight machine learning algorithms and test them on EEG motor imagery data. Single-subject models had higher average accuracies compared to cross-subject trained models for 7 out of 8 machine learning models.