Citation

Sultana, Saima and Alam, Muhammad Mansoor and Prasad, Mukesh and Mustapha, Jawahir Che and Mohd Su ud, Mazliham (2026) Key frame snippet influence on Robot vision in health hazardous circumstances using modular skeleton joints sampling technique. PeerJ Computer Science, 12. e2859. ISSN 2376-5992

Text peerj-cs-2859.pdf - Published Version Restricted to Repository staff only Download (5MB)

Abstract

Robot vision is one of the viable technologies for health-hazardous situations. Robots are good coworkers of humans. So, human action recognition is one of the focused areas of Robot vision. The more accurate identification during a medical emergency is a very crucial issue. For identification accuracy, the current empirical research study considers a novel approach. This study proposes a novel fusion framework model based on Convolutional Neural Network and Recurrent Neural Network to identify and classify health-hazardous situations through human actions. A novel technique of the Modular key frame selection is proposed to bring innovation to the research. To verify the effectiveness of the proposed model, experiments are performed on datasets of the NTU RGBD-60 Skeleton dataset. The research observes and compares the effect of key frame selection on the recognition of actions during medical emergencies by considering the least available frames frequently. To the best of our knowledge, the key frame selection is never considered for action recognition of medical situations. By focusing on this vital issue, observation of key frame selection is conducted precisely. The features have been extracted in three different frame rates per second 3 fps, 6 fps, and 10 fps. While the data is available in 30 fps originally, the results of the empirical study reveal the impact of key frame selection on Skeleton data. The empirical study results from 90.56, 91.12, and 92.01, respectively. In comparison with state-of-the-art research, the current empirical study findings are much better for accurate classification, in circumstances when the data is not available in a constant fashion.

Item Type:	Article
Uncontrolled Keywords:	Robot vision, Neural networks, Recognition of human action, Recognition of medical conditions, Key frame selection in video data
Subjects:	Q Science > QA Mathematics > QA71-90 Instruments and machines
Divisions:	Faculty of Computing and Informatics (FCI)
Depositing User:	Ms Suzilawati Abu Samah
Date Deposited:	04 May 2026 04:49
Last Modified:	04 May 2026 04:49
URII:	http://shdl.mmu.edu.my/id/eprint/15871

Downloads

Edit (login required)