Stress analysis using physiological signals

Abstract

The growth of technology over the past decade has demanded greater levels of concentration and attention to be divided while multitasking. Carrying out multiple tasks could be challenging and one may experience stress overload. Bad stress management could lead to severe mental health issues in the long run such as anxiety and depression. The main aim of this study is to design a stress recognition system to induce varying levels of stress and thereby identify any pattern for Electroencephalogram (EEG) signals during stress. The stressors used in this experiment are the Stroop Colour Word Test and Mental Arithmetic Test. There are 3 main sections in the experiment: the resting, training, and testing sections. The EEG signals of the test subjects are recorded using a device called Emotiv Epoc+. The GUI of the system is developed using C# Windows Form Application while the signal processing, feature extraction and stress classification was done using MATLAB. Stress features such as bandpower, bandpower asymmetry, bandpower difference and bandpower ratio can also be extracted from the power features. The results obtained from the SVM classifiers are 55.66%, 61.02, and 61.05 for the Stroop Colour Word Test, Mental Arithmetic and both tests combined. In comparison with the previous studies related to stress analysis of EEG, the results obtained in this experiment are marginally lower. Our work focuses on the emotional aspect of the test subject and understanding the varying levels of stress experienced by the subject. Nevertheless, stress is an emotion that is very subjective to an individual’s interpretation and experiences.