Human activity recognition

• Main Author: Chua, Wilson Wei Jie.
• Other Authors: Jiang Xudong
• Format: Final Year Project
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering; DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: http://hdl.handle.net/10356/52993
• Institution: Nanyang Technological University

The major goal of computer vision research is to be able to interpret human motion and activities. It is a challenging task to be able to implement an accurate human activity recognition system as human activity is complex and diverse. To date, large body of literatures focuses in the field of human activity recognition. While it is easy for a human to interpret some basic action/movement, such as walking or jumping when demonstrated by their peers, it is not possible for computer to distinguish the actions by itself, not without program that gives the computer the ability to do so. Since 1980s, human activity recognition has been gaining tremendous attention in the computer and machine vision world as many researchers are dedicated in this field through these decades. In this project, the aim is to implement an existing method to allow computer to learn the human activity and be able to recognize the human activity by the recognition process. Two methods of recognition process are proposed to determine which methods will give a better recognition rates. This system is written in C++, and an API, OpenCV is used to work on the system. The system will make use of Principle Components Analysis (PCA) to train human activity recognition.