Research Article
Application Identification for Virtual Reality Video with Feature Analysis and Machine Learning Technique
@INPROCEEDINGS{10.1007/978-3-030-06161-6_33, author={Xiaoyu Liu and Xinyu Chen and Yumei Wang and Yu Liu}, title={Application Identification for Virtual Reality Video with Feature Analysis and Machine Learning Technique}, proceedings={Communications and Networking. 13th EAI International Conference, ChinaCom 2018, Chengdu, China, October 23-25, 2018, Proceedings}, proceedings_a={CHINACOM}, year={2019}, month={1}, keywords={Application identification Statistical feature Machine learning VR video application}, doi={10.1007/978-3-030-06161-6_33} }- Xiaoyu Liu
Xinyu Chen
Yumei Wang
Yu Liu
Year: 2019
Application Identification for Virtual Reality Video with Feature Analysis and Machine Learning Technique
CHINACOM
Springer
DOI: 10.1007/978-3-030-06161-6_33
Abstract
Immersive media services such as Virtual Reality (VR) video have attracted more and more attention in recent years. They are applications that typically require large bandwidth, low latency, and low packet loss ratio. With limited network resources in wireless network, video application identification is crucial for optimized network resource allocation, Quality of Service (QoS) assurance, and security management. In this paper, we propose a set of statistical features that can be used to distinguish VR video from ordinary video. Six supervised machine learning (ML) algorithms are explored to verify the identification performance for VR video application using these features. Experimental results indicate that the proposed features combined with C4.5 Decision Tree algorithm can achieve an accuracy of 98.6% for VR video application identification. In addition, considering the requirement of real-time traffic identification, we further make two improvements to the statistical features and training set. One is the feature selection algorithm to improve the computational performance, and the other is the study of the overall accuracy in respect to training set size to obtain the minimum training set size.