Artificial Neural Network Based Stress Level Detection System using Physiological Signals

Abstract

This study presents a method of detecting the stress level of a person. The objective is to design a portable device using galvanic skin response, body temperature, and heart rate as input parameters. Additionally, the study aimed to implement artificial neural network algorithm as the classifier in measuring the stress level. The system was trained and tested by inducing physical and mental stress stimuli on a group of engineering students as participants, to classify it as low stress, moderate stress, and high stress. The accuracy of the system was verified by comparing the induced stress level to the stress level detected by the device. It was concluded that the device could successfully detect the person's stress level with an accuracy rate of 91.67%. It was also concluded that the stress stimuli used on the study was enough to differentiate the stress level of a person.