Stress Detection Using Physiological Signals Based On Machine Learning

Stress can be defined as the body's attempt to control itself in response to changes in the environment. Due to stress work performance may suffer and the risk of neurological issues such as hypertension and psychological illnesses such as anxiety disorder may rise. In today's world, an increasing number of people are experiencing some form of stress. Comprehension of stress cognition is required along with the capacity to build systems with stress cognition characteristics. A methodology of stress detection using physiological signals based on machine learning is presented in this paper. Physiological signals such as respiration, sweat gland activity on the skin of hands, heart rate, and electromyogram were recorded while driving from multiple healthy participants in various situations and locations. The signal is then segmented for various time intervals such as 100, 200, and 300 seconds, depending on the levels of stress. Statistical features were retrieved and made available to the classifiers namely Support Vector Machine (SVM) and k-Nearest Neighbor (KNN) algorithm. We achieved the highest accuracy of 96% with 100 and 200-second long signal, and 98% with 300-second long signal.