Human action recognition by SVM classification using kinect sensors
Siti Wardah Aziz
Date of Issue2018
School of Electrical and Electronic Engineering
Human action recognition has been an interest in the computer vision field. The use of human action recognition has been applied in various applications such as security surveillance, medical and health fitness tracking, just to name a few. An application was created using C# in Visual Studio 2015 to capture, train and test real-time human actions. This application could be used for fitness tracking for users to practice good form while exercising. Kinect sensor of Version 2 was used to capture the data. Machine learning was integrated into this project which has been the crucial component in human action recognition due to its effectiveness and performance. Support Vector Machine is the chosen machine learning classification method that was applied to train and test real-time human action. A set of simple actions were captured by a few volunteers in recording their actions into a dataset. Multiple experiments were done in various conditions to test out efficiency of the application. This application will deduce the accuracy of the machine learning method employed and to analyse its efficiency with the small number of datasets available.
DRNTU::Engineering::Electrical and electronic engineering::Computer hardware, software and systems
Final Year Project (FYP)
Nanyang Technological University