• Ansh Soni
• Aryan Khare
• Devansh Sachan
• Anshalee Choudhary

A video-only deep learning system is presented for real-time stress detection in virtual meetings by fusing facial emotion analysis with attention monitoring. The system utilizes a ResNet18 backbone trained on FER2013 and DAiSEE to jointly predict emotion, engagement, and stress. A key feature is a confidence-based fusion mechanism (threshold >= 0.6) that dynamically adjusts stress scores, particularly for negative emotions (sad/angry). The model achieved 70.47% accuracy for emotion recognition on FER2013 and an overall validation accuracy of 66.15%. This work demonstrates a robust, non-invasive prototype for assessing stress and engagement in remote environments.

Stress Detection, Virtual Meetings, Convolutional Neural Network, Confidence-Adapted Stress Fusion, Confidence-Regulated Fusion, Deep Learning, Facial Expression Recognition, Engagement Analysis

"Stress Detection in Virtual Meetings Using CNN Deep Learning Model", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 11, page no.g251-g255, November-2025, Available :http://www.jetir.org/papers/JETIR2511677.pdf

"Stress Detection in Virtual Meetings Using CNN Deep Learning Model", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 11, page no. ppg251-g255, November-2025, Available at : http://www.jetir.org/papers/JETIR2511677.pdf