Fuzzy Logic Enabled Stress Detection Using Physiological Signals

Abstract

Stress is attributed as the natural response of body towards unwanted and challenging conditions of society and environment. The consequences of stress are in general very harmful. However, in some specific situations, stress can be highly serious to the human health as it impacts the cardiovascular system in addition to other body sub systems. Timely Detection of stress through physiological sensors may help in arresting the disease. As the stress results in change of heart rate, skin conduction, temperature, blood pressure and oxygen levels, hence, measurement and assessment of these parameters is useful in deciding the stress levels into a candidate. In this paper, a fuzzy logic based automatic stress detection system is designed to timely detect the patients passing through this silent killer disease. The physiological signals of human body are used to assess the current state of person. The parameters used for the purpose are heart rate, Galvanic Skin Response (GSR), and temperature. The fuzzy logic is an exciting branch of artificial intelligence that is used to make useful decisions under uncertain and incomplete input knowledge. A synthetic data set of 500 patients is also developed to train and test the proposed architecture. The algorithms used for this purpose are Hill Climbing (HC) and Simulated Annealing (SA) for fuzzy membership functions training. The results show that Simulated Annealing gives a testing accuracy of 89% in comparison to Hill climbing with 81% accuracy. Future work is directed towards using new methods like type-2 fuzzy logic and deep learning etc.