Automated Video and Audio based Stress Detection using Deep Learning Techniques

Abstract

In today's world, stress has become an undoubtedly severe problem that affects people's health. Stress can modify a person's behavior, ideas, and feelings in addition to having an impact on mental health. Unchecked stress can contribute to chronic illnesses including high blood pressure, diabetes, and obesity. Early stress detection promotes a healthy lifestyle in society. This work demonstrates a deep learning-based method for identifying stress from facial expressions and speech signals.An image dataset formed by collecting images from the web is used to construct and train the model Convolution Neural Network (CNN), which then divides the images into two categories: stressed and normal. Recurrent Neural Network (RNN), which is used to categorize speech signals into stressed and normal categories based on the features extracted by the MFCC (Mel Frequency Cepstral Coefficient), is thought to perform better on sequential data since it maintains the past results to determine the final output.