Authors: Dr. Radha Shirbhate, Zaidkhan Pathan, Aditya Gude, Vishal Joshi
Abstract: Mental health plays a vital role in determining overall well-being, productivity, and social interaction [1], [2]. However, diagnosing mental disorders like depression and anx- iety often relies on self-reporting and therapist observation, which may introduce subjectivity and delay treatment. This paper presents an AI-based facial expression recognition (FER) framework that analyzes human emotions from visual cues to assist early mental health assessment [3], [4]. The proposed system uses Convolutional Neural Networks (CNNs) trained on the FER-2013 dataset, combined with MediaPipe for real-time facial landmark extraction and OpenCV for image preprocessing. The model recognizes seven basic emotions: happy, sad, fear, anger, disgust, surprise, and neutral. Real-time video streams are processed, and the detected emotional states are visualized on a dashboard that can track emotion trends over time [5]. The system demonstrates promising performance with accuracy above 92% on validation data and real-time latency under 40 ms/frame [6]. The integration of FER technology into mental health analysis offers an innovative, non-invasive, and continuous monitoring tool that complements traditional clinical methods.
