Anxiety And Depression Detection Using Deep Learning Technique

Abstract: The majority of individuals cope with stress on a regular basis in varied situations in their daily activities. However, sustained tension or a high level of stress will compromise our safety and interfere with our regular activities. Many physical issues linked to stress can be avoided through early detection of mental stress. There are noticeable changes in a variety of physiological and psychological characteristics, such as facial emotion, speech emotion, etc. when a person is under stress. Data from these features can predict whether a person is stressed or depressed. By using all these and some standard questionnaires, the system's probability of predicting anxiety and depression will increase. The system was evaluated on a dataset of individuals with and without anxiety and depression and achieved promising results with high accuracy and sensitivity. In our proposed system, we have used three modules: facial emotion, speech emotion, and standard questionaries. For the detection of facial emotion, we have used VGG16, for the detection of speech emotion, we have used DNN, and for the standard questionaries, we have used SVM. Finally, we integrated all these modules by using a soft voting mechanism to get the desired outcome. This proposed approach has the potential to provide a non-invasive and efficient method for early detection of anxiety and depression in individuals.

Keywords: SVM, VGG16, facial emotion, speech emotion.