Human Action Recognition using Deep Learning and Supervised Data Fusion Approach

Articles in Press

Document Type : Original Article

Authors

1 Educational deoartment of software computer engineering - Technical- Faculty- shahrekord university -shahrekord

2 Educational Department of Software Computer Engineering - Technical Faculty - Shahrekord University - Shahrekord

Abstract

With the advancement of science and the increasing expansion of artificial intelligence, the recognition of human actions by computers has become an important research field. Action recognition faces numerous challenges due to the complexity of the actions, unstable weather conditions, varying light intensity, and changes in camera viewing angles. To display action, two types of useful data are used: skeletal body structure and color video obtained from Kinect cameras. In this article, body skeletal coordinates are generated with high accuracy using advanced deep learning methods in addition to Kinect cameras. By extracting influential features from each type of data using modern mathematical methods and neural networks, action recognition is performed with each of them. Each type of data alone is capable of providing specific and different information about human action. In the second stage of this research, efficient combination methods with MFA and BGLPCCA supervision are used to utilize the features of both types of data together for human action identification. The proposed method in this research provides a framework for detecting action with single and multiple types of data, and its superiority lies in achieving high accuracy in human action recognition.

Keywords

Main Subjects

Advanced Signal Processing

Articles in Press, Accepted Manuscript
Available Online from 12 October 2024

Files

Share

How to cite

Statistics