Identification and analysis of human physical exercise postures using computer vision and deep learning / Sriram Krishnamoorthy
The ability of the human beings to perform physical exercise postures is identified by the unique ways and its important categories are gait analysis and muscle expansion of the human body as exercise performance. The postures performed by the human beings are estimated in a dynamic way. Most curren...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2021
|
| Subjects: | |
| Online Access: | http://studentsrepo.um.edu.my/13163/5/Sriram_Krishnamoorthy.jpg http://studentsrepo.um.edu.my/13163/9/sriram.pdf http://studentsrepo.um.edu.my/13163/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Universiti Malaya |
| Summary: | The ability of the human beings to perform physical exercise postures is identified by the unique ways and its important categories are gait analysis and muscle expansion of the human body as exercise performance. The postures performed by the human beings are estimated in a dynamic way. Most current systems which are employed for identifying human exercises propose complex systems. The advancement of the Artificial Intelligence broke the existing complexities, identifies, and analyze the human exercise without complex models. The effectiveness of the Computer Vision and Deep Learning are used to monitor the accurate exercise postures of every human gait to solve the trainer’s vacuum in the exercise environment. In this paper, it has evaluated and developed customized pose model to identify correct human physical postures, to analyze the correct form of physical exercise in a correct way in all type of exercise environments. For correct comparison, typical nodal points are identified (depends on exercise type) from the human body to calculate the estimations in a synchronous way. The estimations are nodal analysis which replicates the correct analyzation. The input frames of nodes are obtained by Webcam to a computer which makes the identification dynamic and naturalistic as Human-Computer interaction. We propose Deep pose estimation model to improve the analyzation of the physical exercises without complex systems. The results demonstrated the possibility of non-invasive identification which can be used for certain physical exercise postures using simple system with better accuracy. |
|---|