Enhanced Statistical Modelling For Variable Bit Rate Video Traffic Generated From Scalable Video Codec

Loading...
Thumbnail Image
Date
2016-02
Authors
Ahmadpour, Sima
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
Designing an effective and high performance network requires an accurate characterization and modelling of the network traffic. This work involves the analysis and modelling of the Variable Bit Rate (VBR) of video traffic, usually described as the core of the protocol design and efficient network utilization for video transmissions. In this context, an Enhanced Discrete Autoregressive (EDAR (1)) model for the VBR video traffic model, which is encoded by a Scalable Video Codec (SVC), has been proposed. The EDAR (1) model was able to accurately generate video sequences, which are very close to the actual video traffic in terms of accuracy. The model is validated using statistical tests in order to compare simulated and original traces. The validation is done using graphical (Quantile-Quantile plot) and statistical measurements (Kolmogorov-Smirnov, Sum of Squared Error, and Relative Efficiency), as well as cross-validation. Furthermore, the EDAR (1) model was compared against three different other models using the aforementioned techniques. All four models under the study including the EDAR (1) model have been applied for each movie under the study separately. It is shown that the SSE of EDAR (1) model for one specific movie is less than the SSE of other three models. The same comparison is done for all movies under the study and the SSE of EDAR (1) model resulted in about 11 - 30 % less error compared to the other models. This means that the data generated by the EDAR (1) model is more accurate and close to the actual video traffic than the other models. In terms of cross-validation, the validation has been done for each model in a specific movie separately. The SSE of the EDAR (1) model varied between 8 - 20 % less than the others. It has hereby been shown that the proposed model is effective in deriving a more accurate trace for both video traffic analysis and the network performance evaluation.
Description
Keywords
Video traffic
Citation