Joint source and sending rate modeling in adaptive video streaming

被引:11
作者
Colonnese, Stefania [1 ]
Frossard, Pascal [2 ]
Rinauro, Stefano [1 ]
Rossi, Lorenzo [1 ]
Scarano, Gaetano [1 ]
机构
[1] Univ Roma La Sapienza, Dept Informat Elect & Telecommun DIET Engn, I-00184 Rome, Italy
[2] Ecole Polytech Fed Lausanne, Signal Proc Lab LTS4, CH-1015 Lausanne, Switzerland
关键词
Traffic model; Rate control; Adaptive streaming; ADMISSION CONTROL; HIDDEN MARKOV; MPEG-4; H.264/AVC; NETWORKS;
D O I
10.1016/j.image.2013.01.007
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
This work addresses the modeling of traffic generated by a video source operating in the context of adaptive streaming services. Traffic modeling is a key in several network design issues, such as dimensioning of core and access network resources, developing pricing procedures, carrying out cost-revenue studies. The actual traffic generated during a video streaming session depends on both the video source and the bandwidth variations imposed by lower communication layers. We propose a new traffic model that jointly encompasses these two effects. Specifically, we consider the modeling of the sequence of frame sizes generated by a video streaming source that dynamically adapts its rate to the available communication channel bandwidth using bitstream switching techniques. In order to represent the source rate adaptation to the random network bandwidth variations on the communication channel, we resort to a framework based on Hidden Markov Processes (HMPs). Our HMP model represents the first joint source and sending rate model in adaptive streaming literature. Thanks to effective modeling assumptions on the frame size probability density function (pdf), the HMP parameters can be estimated by means of the Expectation Maximization algorithm. The traffic model is validated by numerical simulations of a mobile adaptive video streaming scenario. We study the model's ability to predict several traffic statistics, including the traffic load of a video streaming source in different network points. Besides, we evaluate the model accuracy in characterizing aggregate video traffic resulting from multiplexing various video sources. In all experiments, we show that the proposed model is able to accurately capture the traffic characteristics. (c) 2013 Elsevier B.V. All rights reserved.
引用
收藏
页码:403 / 416
页数:14
相关论文
共 56 条
[1]   NeXt generation/dynamic spectrum access/cognitive radio wireless networks: A survey [J].
Akyildiz, Ian F. ;
Lee, Won-Yeol ;
Vuran, Mehmet C. ;
Mohanty, Shantidev .
COMPUTER NETWORKS, 2006, 50 (13) :2127-2159
[2]  
[Anonymous], 2005, 3984 IETF RFC
[3]  
[Anonymous], 2004, 26244 3GPP TS
[4]  
[Anonymous], 2007, 26233 3GPP TS
[5]   On modeling MPEG video traffics [J].
Ansari, N ;
Liu, H ;
Shi, YQ ;
Zhao, H .
IEEE TRANSACTIONS ON BROADCASTING, 2002, 48 (04) :337-347
[6]  
Baguda Y.S., 2008, ENG TECHNOLOGY, V24, P344
[7]   A multiscale expectation-maximization semisupervised classifier suitable for badly posed image classification [J].
Baraldi, Andrea ;
Bruzzone, Lorenzo ;
Blonda, Palma .
IEEE TRANSACTIONS ON IMAGE PROCESSING, 2006, 15 (08) :2208-2225
[8]   Bayesian wavelet-based image deconvolution: A GEM algorithm exploiting a class of heavy-tailed priors [J].
Bioucas-Dias, JM .
IEEE TRANSACTIONS ON IMAGE PROCESSING, 2006, 15 (04) :937-951
[9]  
Cai J, 2000, GLOB TELECOMM CONF, P1515, DOI 10.1109/GLOCOM.2000.891893
[10]   A hidden Markov, multivariate autoregressive (HMM-mAR) network framework for analysis of surface EMG (sEMG) data [J].
Chiang, Joyce ;
Wang, Z. Jane ;
McKeown, Martin J. .
IEEE TRANSACTIONS ON SIGNAL PROCESSING, 2008, 56 (08) :4069-4081