Induced Stress Identification Using EEG: A Framework based on MVMD and Machine Learning

The use of EEG for stress detection has driven the development of advanced algorithms to overcome the challenge of visually analyzing subtle changes in EEG patterns. This paper introduces an automated method for identifying induced stress from EEG signals using multivariate variational mode decomposition and machine learning (ML). Various time domain features were derived from the decomposed EEG signals, leading to significant outcomes with the support vector machine classifier across three distinct scenarios of induced stress. The suggested model achieved an impressive accuracy rate of 94.3% for the stroop word-color test with a window size of 7.5 sec. Additionally, an analysis concerning different window sizes of EEG signals is also performed, offering valuable insights into various cases of induced stress and stress identification. Furthermore, a performance comparison between the proposed model and multiple ML-based classifiers demonstrated the superiority of this approach.