Article / motionSense: Early Detection of Depression and Anxiety via Multimodal Emotional AI with Wearable and Social Media Data

Depression and anxiety, affecting over 400 million individuals globally, often remain undetected until critical stages, delaying life-saving interventions. This study introduces EmotionSense, an innovative emotional artificial intelligence (AI) framework designed for the early identification of these disorders. By integrating physiological signals (e.g., heart rate variability, sleep metrics) from wearable devices, behavioral patterns (e.g., app usage, digital footprints) from smartphones, and linguistic markers (e.g., sentiment, emotional valence) from social media, EmotionSense leverages advanced natural language processing (BERT-based models) and ensemble machine learning techniques (XGBoost) to achieve robust detection. Evaluated on a diverse cohort of 750 participants over 60 days, the system demonstrated accuracies of 92% for depression and 88% for anxiety, with an AUC-ROC of 0.95. These results position EmotionSense as an adaptable, privacy-first tool for real-time mental health monitoring, offering transformative potential for personalized interventions and public health strategies.