Multimodal Depression Detection Using Audio, Visual and Textual Cues: A Survey

Abstract

In the recent past, mental health has become a global concern. COVID-19 has further caused a rapid surge in depression. Depression is a serious mental illness that is impacting the lives of individuals of all ages all around the world. Depression affects a person's physiological well-being as well as their emotional state. Now days, Depression is the most common element experienced by the human beings irrespective of their age factor and professional life. To detect the depression status among the persons, the system uses different approaches by using the sensor technology. The automatic identification of depression at early stages or immediately helps the clinical studies to cure the people accurately. In this proposed research, the system aims to identify the depression using facial expressions, voice, live video capturing, by analysing their tweets, status, posts in the social media. By applying computer vision integrated with ML and DL techniques, the entire capturing and analysis process gets automated and the complexity involved in the model designing gets reduced because the system focuses more on extracting the statistical features involved in movements and behaviour of the human being. Most of the existing research works focuses on the unimodal development which focuses on the single component analysis but the proposed research aims to focus on the multi modal with a fusion of different modalities of learning approaches involved in detection of depression, this survey provides an overview of numerous methodologies that have been created with the goal of employing emotion recognition to analyse depression.