Mental health challenges demand innovative, non-invasive interventions to reduce stress and enhance emotional stability. Music has long served as a therapeutic medium; however, existing approaches often rely on generic playlists that lack personalization and adaptability to an individual's psychological state. This paper presents a novel framework that combines webcam-based facial expression analysis with questionnaire-based self-reports to achieve robust emotion detection. The proposed system employs deep learning models to extract emotional cues from visual data, while structured self-assessments provide subjective validation of user states. A fusion mechanism integrates both modalities to enhance the accuracy and reliability of emotion recognition. Based on the detected emotional profile, personalized music recommendations are generated and visualized through interactive Power BI dashboards. This multimodal, AI-driven approach bridges traditional music therapy with modern data analytics, enabling adaptive, accessible, and user-centric mental health support. The experimental results highlight the potential of this method to enhance emotional well-being, alleviate stress, and increase access to personalized therapy.
Gnanasambanthan et al. (Tue,) studied this question.