Multimodal Stress Detection Using Deep Learning: Integrating Physiological and Behavioral Signals

Abstract

The Stress is a big deal nowadays, really taking a toll on how people feel mentally and physically. Traditionally, figuring out someone's stress levels uses questionnaires, but those can be pretty off since they depend on the person's memory and mood at the time. This paper suggests something new model that hooks up physiological signals like heart rate, sweat stuff, and brain waves with how you look and sound to tell if you're stressed. This model use Convolutional Neural Networks and LSTM networks together to spot patterns in your body’s responses, both in the moment and overall. The evaluation results obtained clearly highlights that the proposed model worked better than other traditional methods that only focus on one kind of signal. This Proposed model can consistently check for stress in real-time, keep track of health, helps to boost work performance, and set up modified relaxation platforms. Furthermore, the results obtained showed that combining multiple types of signals like behavioral and physiological improves accuracy and stability. This proposed hybrid deep learning technique out performs traditional techniques in terms of all performance evaluation metrics like accuracy, precision, recall, f1-score.