The rapid growth of mental health concerns in modern society necessitates effective and accessible tools for stress detection. Traditional approaches rely heavily on clinical evaluation or self-reporting, which are often subjective, time-consuming, and unsuitable for real-time assessment. This research proposes a dual-modality artificial intelligence system for stress detection that leverages both text analysis and facial recognition. The system utilizes Natural Language Processing (NLP) techniques with LSTM and CNN models to analyze user-input text, while Convolutional Neural Networks (CNNs) process facial expressions captured in real-time video streams. By integrating these modalities, the system achieves higher accuracy and reliability compared to single-modality approaches. The model is implemented using Python, TensorFlow, and OpenCV, and deployed through an interactive Streamlit interface for real-time user interaction. Experimental evaluations demonstrate the system’s potential to deliver accurate, efficient, and user-friendly stress detection, offering valuable applications in healthcare, education, and corporate environments for proactive mental health monitoring and early intervention.
Noorun Nehar (Thu,) studied this question.