“MoodMate” - A Hybrid Approach of Personal Emotional Well-being Recommendation Using Facial and Verbal Emotion Recognition

Abstract

"This research investigates the prevalence of undiagnosed mental disorders by using a novel approach that analyzes self-uploaded videos to detect individuals' moods. Employing deep learning algorithms, the study creates a hybrid model that predicts emotions based on facial expressions, speech tones, and vocal text. Additionally, it incorporates a reinforcement learning agent to suggest personalized activities, aiming to support recovery from mental disorders. The methodology emphasizes the integration of various analysis techniques to improve accuracy and utilizes specific metrics like confusion matrices, AUC-ROC, RMSE, and MSE for evaluation.Initial results are promising, highlighting the model's potential in assisting individuals to recognize and address their mental health issues, suggesting significant implications for further research and development in the mental health domain."