FitMate: A Real-Time AI-Based Exercise Monitoring and Posture Correction

Authors:

Dr. Rupinder Kaur, Aditya Ujjwal, Abhishek Jain, Ashutosh Kumar Jha, Vinod Joshi

Abstract:

The increasing demand for accessible and intelligent fitness solutions has driven the integration of artificial intelligence and computer vision into personal health applications. This paper presents FitMate, a web-based fitness coaching system that performs real-time posture analysis and exercise monitoring using pose estimation techniques. The system extracts skeletal landmarks from live webcam input and applies geometric computations to accurately track repetitions and evaluate exercise form. Designed as a fully client-side application, it ensures efficient execution, low latency, and enhanced user privacy without reliance on external processing. FitMate supports multiple exercises, including squats, push-ups, bicep curls, jumping jacks, and high knees, and provides immediate visual feedback along with performance metrics such as repetition count, elapsed time, and estimated calories burned. Experimental results demonstrate an average accuracy of approximately 90.4%, indicating reliable performance across diverse exercises. The system enables convenient, home-based guided workouts, promoting consistent fitness habits and highlighting the potential of AI-enabled web platforms for accessible and real-time fitness guidance.

Keywords: Artificial Intelligence, Computer Vision, Pose Estimation, Fitness Tracking, Real- Time Feedback, Web Application, Exercise Monitoring, Human Pose Analysis, Client-Side Processing, Digital Health