An Intelligent Machine Learning System for Real- Time Stress Management Based on a Mini-Xception Algorithm and Deep Neural Network Models

• Published in: 2023 IEEE International Conference on Contemporary Computing and Communications (InC4) Vol. 1; pp. 1 - 6
• Main Authors: Varsha, S K, Sri, R. Leena, Anuvidhya, K
• Format: Conference Proceeding
• Language: English
• Published: IEEE 21.04.2023
• Subjects: Anxiety disorders; C N N; Deep learning; Games; Human factors; Image processing; Open CV; Probability; Real-time face emotion recognition and Stress deduction; Real-time systems; Streaming media; Visualization
• DOI: 10.1109/InC457730.2023.10263028

The Massive Open Online courses that students take up to enhance their knowledge other than academics can be isolating, with spending long hours in front of a computer screen without social interaction that comes with in-person classes. This can lead to feelings ofloneliness and disconnection. Online courses often require self-motivation and self-discipline, which can be daunting for some students. The fear of falling behind or not meeting expectations can be a significant source of stress. This Stress that is customary among students can cause unpropitious mental and physical side effects if left undeterred. So, in this project, we propose a system to monitor and reduce stress levels. So, the main goal of our product is Effective stress management that escalates the user's level of self-confidence, improves their concentration, enhances their productivity, ameliorates the frames of mind, and finally lowers despondency and anxiety. The project has been completed with a functional GUI called `Hakuna Matata.' The stress level is calculated with the help of eyebrows contravening and supplantation from the mean position from the real-time video of the user by using an exponential function after several independent computations, the binary probabilities are normalized between 1 and 100 to get the stress level. With this computed stress value combined with the emotion predicted by analyzing the user's face using functions from the Histogram of oriented gradients algorithm and Convolutional Neural Networks, we declare if a person is stressed or not, and based on the mid-stress level a small game will be axiomatically opened for the user to play and get relaxed. Our project is within the scope of Sustainable Development Goal 3: Good Health & Wellbeing since the present work aims to solve the societal need by reducing the increasing stress levels, which comes under the umbrella of Health Issues.