Implementation of Pose Estimation as a Foundational Module for AI-Based Dance-Fitness Assistance Systems

DOI

10.2991/978-94-6239-678-4_10How to use a DOI?

Keywords

Pose Estimation; MediaPipe; Temporal Smoothing; Lighting Robustness; Joint Angle Analysis; Motion Stability

Abstract

There is an ever-increasing trend of pursuing newer ways through which physical and mental well-being may be maintained through dance and fitness. Along with the growth of online training and home workouts, there is also an ever-pressing need for systems that are intelligent enough to capture and track human movements in real time. Pose estimation is a computer vision-based technique that infers key body landmarks from video inputs to estimate posture and study motion without the use of wearable sensors. A few of the existing methods are OpenPose, MoveNet, and PoseNet; most of these have a limitation in computational dependence on GPUs, single-person detection, reduced accuracy on rapid movements, and sensitiveness against variation in illumination, thus limiting their usability in dynamic dance and fitness environments. The main aim of this work is to develop a system named the Integrated Real-Time Pose Stability and Analysis (IRPSA) framework. The temporal smoothing and lighting change adjustment to pose estimation in real-time is implemented with the IRPSA and MediaPipe BlazePose on OpenCV (the Open Computer Vision Library) on regular CPU. The framework gives accurate estimates of the joint angles across a wide variety of lighting conditions; stabilizes the motions of a series of frames; and, finally, gives a reliable, lightweight and flexible pose tracking system. AI-powered dance fitness assistance may be based on the pose tracking system. Its primary features of enhancement include improvement of tracking stability, reduction of computational overhead, maintaining accuracy, and being applicable to home and virtual fitness. The system would therefore be sensor-free and efficient, allowing for intelligent fitness monitoring. The performance evaluation was conducted using the pose landmarks produced by MediaPipe BlazePose, which has been trained on Google's internal large-scale human-pose and athletic-pose datasets, in addition to recorded video samples of dancing and fitness under several lighting conditions.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Dussa Sanjana
AU  - Ayesha Butalia
AU  - Reena Pagare
PY  - 2026
DA  - 2026/05/28
TI  - Implementation of Pose Estimation as a Foundational Module for AI-Based Dance-Fitness Assistance Systems
BT  - Proceedings of the 2nd International Conference on Recent Advancement and Modernization in Sustainable Intelligent Technologies & Applications (RAMSITA-2026)
PB  - Atlantis Press
SP  - 115
EP  - 124
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-678-4_10
DO  - 10.2991/978-94-6239-678-4_10
ID  - Sanjana2026
ER  -