Personalized Mental Health Interventions Using Generative AI and Multimodal Data

Artificial Intelligence (AI) systems could provide customized mental health interventions based on multiple types of data. The study introduces dynamic graph convolutional neural network (DGCNN) which identifies emotions simultaneously from text signals, as well as voice and face data as a real-time recognition model. The research addresses the gap in effective multimodal integration by introducing a weighted decision-level fusion technique that compensates for modality-specific data degradation. This research implements one-frame-per-second constant video processing of IEMOCAP and MELD video clips through majority voting which serves as a reliable classification technique. By minimizing the differences between various input signals, the weighted decision-level fusion technique optimizes its accuracy performance. The assessment system combines dependable psychological measures which act as infrastructure to support AI-based customized therapeutic services. Key findings reveal that the proposed model achieves improved classification accuracy and supports early detection of emotional distress with minimal computational load. Such emotional evaluation immediately supports clinicians during early detection of psychological stress indicators. Advanced mental health care access is achieved through better patient reach and enhanced health achievement according to the implemented strategy.