Depression Estimation from Audio Recordings of Clinical Interviews

Author

Veselin Roganović, Faculty of Technical Sciences, University of Novi Sad

Project Overview

This project focuses on the estimation of depression severity from audio recordings of clinical interviews. It includes analysis of sound patterns and the words spoken with popular machine learning models. Additionally, the learned representations in the models will be investigated to better understand the characteristics associated with depression.

Project Objectives

• Acoustic Analysis: Implement and evaluate machine learning architectures for processing audio features to estimate depression severity.
• Linguistic Analysis: Develop and compare text-based models for analyzing interview transcripts.
• Interpretability: Investigate model representations to identify key acoustic and linguistic characteristics associated with depression.
• Comparison: Benchmark different modeling approaches to determine effective strategies for clinical severity estimation.

Implementation & Results

Will be updated later :)