Ruiji Xu, Yan Ling, Keji Mao, Guanglin Dai
Depression detection; multimodal fusion; desktop robot; real-time monitoring; emotional analysis
Depression is one of the most prevalent mental disorders globally, severely impacting individual’s quality of life and socioeconomic burden. Current depression detection methods mainly rely on self- rating scales and clinical interviews, which face significant limita- tions: scales are subject to subjective bias and concealment, while interviews depend heavily on physician expertise and are constrained by the shortage of healthcare professionals. To address these challenges, this study developed EmoEcho, an intelligent desktop robot system that integrates multimodal depression detection with automated monitoring capabilities. The system evaluates depressive mood through three complementary modalities: textual semantics, vocal characteristics, and facial expressions, mirroring the key factors physicians consider during clinical interviews. By implementing a dynamic fusion mechanism and hardware-algorithm co-optimisation, EmoEcho achieved robust performance in uncontrolled environ- ments. Clinical trials conducted at medical centers demonstrated the system’s effectiveness, achieving 91.2% screening accuracy (95% CI: 87.4%–93.8%) and a consistency coefficient of 0.78 with HAM-D scale assessments, significantly outperforming traditional methods. These results suggest that EmoEcho represents a promising advancement in early depression detection and intervention, offering a scalable solution for mental healthcare services.
Important Links:
Go Back
