Optimizing Facial Expression and Head Dynamics Data Processing to Enhance Depression Detection with Cutting-Edge AI Models

In this paper, we present an optimized approach for processing facial expression and head dynamics data to improve depression detection using state-of-the-art AI models. Depression is a significant global mental health concern, necessitating accurate, efficient, and non-intrusive diagnostic tools. We evaluate multiple machine learning algorithms, including Naïve Bayes, Decision Tree, Random Forest, XGBoost, K-Nearest Neighbors (KNN), and Neural Networks, to determine their effectiveness in identifying depression. Among these, the Decision Tree model achieves an average accuracy of $\mathbf{8 3. 9 6 \%}$. To further enhance detection performance, we propose a hybrid model combining Extra Trees Classifier, XGBoost, and Logistic Regression as a meta-classifier. This ensemble approach significantly improves classification accuracy, achieving an average of $99.55 \%$. The results highlight the potential of integrating multiple AI techniques for more reliable and precise depression detection. Our findings emphasize the importance of leveraging facial and head movement analysis as a robust, automated, and accessible screening method. Future research will focus on refining feature selection, incorporating additional behavioral cues, and improving realtime deployment for clinical and telemedicine applications.