Audio-centered Approach for Building a Multimodal Predictive AI Agent to Detect Depressive Behaviors

Depression remains one of the most serious and widespread mood disorders, affecting more than 280 million people worldwide. Current diagnostic approaches rely heavily on subjective self-reports and clinical interviews that are still failing to reliably capture the true patient’s behavior in natural, everyday settings. This master’s thesis introduces a novel approach for automated mental health monitoring. Particularly designed around an acoustic-based approach for depression detection, designed specifically as a software application for IoT- enabled private households. Using passive sensing techniques, the system focuses on the detection of potential depressive behavior to allow timely intervention.