Wearable Physiological IoT System with Neural Network to Detect an Individual IT Professional’s Stress / Emotion in Working Environments

Abstract

Stress in the workplace has a major impact on people’s lives. IT industries are pushing the boundaries by introducing cutting-edge technologies and products but the high levels of stress among IT professionals remains as a concerning issue. Research has shown that stress can have an adverse effect on cognitive performance and physical health. This research study presents a novel neural network based wearable physiological IoT system to detect the stress and emotion levels of individual IT professionals in working environments. The proposed system is composed of a wearable device, mobile application, and neural network. The wearable device is equipped with three physiological sensors, namely heart rate, skin temperature, and electromyography (EMG) sensor. The mobile application collects the sensed data from the wearable device and then sends it to the neural network for analysis. The neural network is designed to detect the individual’s stress and emotion levels based on the sensed physiological data. In addition, the mobile application provides the user with visual feedback and can be used to set goals and track progress. The proposed system is evaluated through a user study in an IT professional work environment. The results show that the system has the capability to detect the individual’s stress and emotion levels with an accuracy rate of 91.6%. In order to minimize the potential risks of stress and its related effects, it is important to recognize these emotions and take appropriate steps to alleviate them. This work presents a system that can detect stress using facial analysis, a pulse oximeter, and temperature sensors. A camera is used to capture the person’s front view. The proposed system utilizes a Convolutional Neural Network (CNN) Algorithm to analyze an individual’s facial expression in an image frame and determine their stress level. A prototype was developed to detect if someone is under stress based on their heart rate and blood pressure variations. Stress cannot be measured, so mapping stress through facial expressions and heartbeat can provide an indication of the level of stress. In case of a high pulse rate of the stressed person, both the organization and the person himself are notified about the stress condition. All recorded data are stored and maintained in the database for future reference. Results show that the algorithm used detects stress with 95 % accuracy, showing high correlations between measured (sensors) and recorded (image) stress events. The proposed system is expected to provide valuable feedback for IT professionals to self-manage their stress and emotion levels in their work environments.