Performance Comparison for Skeletonbased Action Recognition by Machine Learning Algorithm

JIITA, vol.6 no.4, p.644-648, 2022, DOI: SOON

Vasanth Ragu 1), Ho-Gun Ha 1), Rock-Hyun Choi 1), Hyunki Lee *

1) Division of Intelligent Robotics, Daegu Gyeongbuk Institute of Science & Technology, Daegu, South Korea

Abstract: Regular exercise gives strong and healthy body in day-to-day life.
Most of the people have lack of knowledge of exercise and they didn’t think their doing exercise pose whether it is right or wrong. While doing different kind of exercise sometimes mistakes are happened, which mistakes can break our bones or muscle spasm, etc. To overcome with this, we need to create awareness of exercise and teach to human whether they are doing exercise right position or not. This study deals with this issue and to solve with the help of artificial intelligence. In OpenPose algorithm, camera has taken as input and to produces the output of classifying human exercise recognition.
The procedure of the system is, data collection, step 2 is preprocessing, step 3 is feature extraction and classification using machine learning algorithm.
The following detection accuracy of skeleton-based action recognition was discussed in result and conclusion.

Keywords: Skeleton-based action recognition, OpenPose algorithm, Machine learning algorithm.

Scroll to top