Depression Detection and Analysis Using Deep Learning: Study and Comparative Analysis

Depression is one of the most serious and quite commonly diagnosed mental disorders. It affects not only the sufferers but also their families, friends and even the society in general. With rapid advancements in Artificial Intelligence and Machine Learning, there have been some recent developments that aim to predict the severity of depression in an individual by analyzing certain parameters extracted from their video sample. This paper discusses some of the recent and most noteworthy researches in the domain of depression analysis and compares their different frameworks like CNNs, RNN, etc. and algorithms used. The severity of depression is identified by using the Beck Depression Inventory-II (BDI-II) that scales from 0 to 63. The performance of all these models is compared based on the framework (architecture and algorithms) adopted and based on the MAE & RMSE scores obtained after testing them on AVEC 2013 & AVEC 2014 datasets. Apart from the comparative analysis, this paper also discusses a couple of approaches that study some non-conventional parameters that might prove effective in the task of depression detection and prediction from video data.