On the Quality of Delivery for Variable Bit Rate Video

Document type: Licentiates
Full text:
Author(s): Karel De Vogeleer
Title: On the Quality of Delivery for Variable Bit Rate Video
Series: Blekinge Institute of Technology Licentiate Dissertion Series
Year: 2012
Issue: 2
Pagination: 113
ISBN: 978-91-7295-225-6
ISSN: 1650-2140
Publisher: Blekinge Institute of Technology
City: Karlskrona
Organization: Blekinge Institute of Technology
Department: School of Computing (Sektionen för datavetenskap och kommunikation)
School of Computing S-371 79 Karlskrona
+46 455 38 50 00
Authors e-mail: kdv@bth.se
Language: English
Abstract: Several factors affect the perceived quality of a video streamed over a network. Technical aspects related to the codec and the video play an important role besides the delivery of the video content in a timely and error-free manner. If the streamed video frames do not arrive in time, temporal artifacts may become observable during the video playback. This has potentially an adverse effect on the user experience. The study of temporal artifacts during the video playback is the area of concern for Quality of Delivery.

The licentiate work investigates the Quality of Delivery for Variable Bit Rate video sent over wireless technologies. The ability of the network to deliver data to the video’s jitter buffer before its playout deadline is studied. Jitter buffer exhaustions are of particular interest as they tell something about the Quality of Delivery. A number of experiments are conducted where a set of videos with different bit-rates are streamed to a mobile device over a wireless LAN and a W-CDMA network. The video streams are recorded and analyzed based on specific Quality of Service parameters and are related to the states of the jitter buffer. The statistical tools Support Vector Machines and the Mahalanobis distance are applied to the parameters to obtain a model that can classify the jitter buffer states. The performance evaluation indicates that the Mahalanobis
distance can classify the jitter buffer state marginally better than the Support Vector Machines. However, the Support Vector Machines can produce more reliable predictions compared to the Mahalanobis distance. It is also observed
that the perceived Quality of Delivery is not only affected by the behavior of the wireless networks but also by the behavior of the streaming server. Finally, the ability of the Quality of Service parameters to describe the Quality of Delivery is studied. The results indicate that metrics based on the packet arrival-rate and packet inter-arrival time are most suitable in this particular case.
URN: urn:nbn:se:bth-00519