FACETRACK: CNN-BASED FACIAL EMOTION RECOGNITION FOR STRESS MONITORING IN SPORTS PHYSIOLOGY
Dr. Deepika Saravagi and Dr. Manisha Saravagi2
Assistant Professor, TransStadia Institute, Mumbai, Maharashtra, India 2Phisiotherapist, Railway Hospital, Kota, Rajasthan, India
ABSTRACT
We present FaceTrack, an efficient convolutional neural network (CNN) model developed for real-time detection of facial emotions, specifically geared toward monitoring stress in athletes during both training and rehabilitation phases. Trained on the FER‑2013 dataset, which includes 35,887 grayscale images of size 48×48 pixels representing seven emotional categories, the model attained an impressive test accuracy of 92.39%. It outperformed several existing approaches in terms of both accuracy and computational efficiency. The model demonstrated particularly high accuracy in recognizing "happy" and "neutral" emotions, while slightly lower results were noted for "fear" and "surprise." Thanks to its lightweight architecture, FaceTrack is well-suited for deployment on edge devices such as smartphones and wearables, offering a practical solution for use in fields like sports psychology and physiotherapy. The complete implementation is available for public access at:
● GitHub: https://github.com/SaravagiDeepika/Stress-Recognition
● Kaggle: https://www.kaggle.com/code/drdeepikasaravagi/stress-recognition
Keywords: facial emotion recognition; CNN; FER‑2013; stress detection; sports physiology; wearable systems
How to cite?
Saravagi, D., & Saravagi, M. (2025). FaceTrack: CNN-Based Facial Emotion Recognition for Stress Monitoring in Sports Physiology. myresearchgo, 1(4), 46.
DOI CrossRef