Stress and anxiety detection: deep learning and higher order statistic approach

• DOI: 10.11591/ijeecs.v33.i3.pp1567-1575
• Publication Date: 2024-02-16T09:26:26Z
• Authors: Vaishali M. Joshi; Deepthi D. Kulkarni; Nilesh J. Uke
• Keywords: Statistic

<p>Today's teenagers are dealing with anxiety and stress. Anxiety, depression, suicide rates have increased in recent years because of social rivalry. The research is focused on detecting students due to exam pressure reduce the potential harm a person's wellness. Research performed databases for anxious states based psychological stimulation (DASPS) our own database. measured signal divided into sub bands that correspond electroencephalogram (EEG) rhythms using Butterworth sixth-order order filter. In higher dimensional space, nonlinearities each sub-band analyzed statistics third-order cumulants (TOC). We classified stress support vector machine (SVM), K-nearest neighbor (K-NN), deep learning bidirectional long short-term memory (BiLSTM) network. comparison previous techniques, proposed system's performance BiLSTM quite good. best accuracy this analysis was 87% DASPS database 98% Finally, subjects high levels had more gamma activity than little This could be an important attribute classification stress.</p>