Detecting and Analyzing Depression: A Comprehensive Survey of Assessment Tools and Techniques

Abstract

Nowadays, clinical depression is a prevalent yet severe mood disorder that occurs with aging. Since sadness has an influence on the mind, it can be hard for the patient to tell the doctor about their situation. Typically utilized diagnostic tools include questionnaires or interview-style evaluations of the symptoms, and also using laboratory tests to see if the depressive symptoms coexist with other severe diseases. In recent years, a variety of approaches have been created to aid the diagnosis of depression, thanks to the development using convolutional neural networks with machine learning. Being a multifactorial condition, depression should be diagnosed using a multimodal approach for an efficient examination. In order to analyze depression using emotion recognition, a number have been created for both unimodal and multimodal approaches. This study reviews these approaches. When compared to multimodal approaches, which combine one or more features, the unimodal approach takes into account just one attribute from the range of facial expressions, voice, etc. for depression identification. This study also discusses many techniques for detecting depression in speech, including spectral, acoustic, and fisher vector algorithms, as well as approaches for extracting face characteristics from speech. The survey includes the current research on emotion recognition that uses auditory and visual information to identify depression. The survey demonstrates that multimodal methods and deep learning techniques outperform unimodal approaches in the study of depression for depression detection.