Fast Multilevel Mental Stress Identification From Bispectrum-Based Heart Rate Variability Feature

Abstract

Fast and accurate identification of mental stress is critical to reducing its risk to human health and the related consequences in industry. This study proposed a bispectrum-based feature to improve the accuracy and reduce the latency of multi-level mental stress identification from heart rate variability (HRV). The performance was evaluated by two stress induction protocols, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.</i> , Stroop color-word interference test (SCIT) and mental arithmetic calculation (MAC), and the results demonstrated that the bispectrum-based methods significantly outperformed traditional HRV features with windows as short as 15 s. The high-order components found by an established nonlinear test, and the strong feature correlations detected between short and long windows, further validated the effectiveness and the feasibility of the proposed methods on mental stress identification from super-short windows. The results had the potential to promote the applications of HRV-based mental stress management across industries with the Internet of Medical Things (IoMT) devices.