DocumentCode
493010
Title
Simple media-aware packet discard algorithms
Author
Guedes, Bruno ; Pereira, Ricardo Lopes ; Varela, António ; Vazao, Teresa
Author_Institution
Dept. of Comput. Sci. & Eng., Inst. Super. Tecnico, Porto
fYear
2009
fDate
21-24 Jan. 2009
Firstpage
1
Lastpage
5
Abstract
IPTV solutions are emerging in the market today, both competing with traditional distribution mediums, such as cable TV, and creating new markets, such as mobile TV viewing using cellular networks. The distribution of IPTV is usually performed using a single stream, resorting multicast. This is not compatible with the growing number of devices which may be used to watch TV, from PDAs to laptops, desktops and high definition TVs, each with different processing and networking capacities. This has prompted research into scalable video coding and adaptation techniques, where a single video stream may be used for a large number of differentiated clients. In this paper we propose and evaluate two different metrics for IP packet discard by a media-aware network element, allowing it to reduce the bandwidth of an original video stream in case of congestion on a downstream link. The first metric consists in discarding the last IP packet of P frames. The second consists in discarding entire P frames. Through experimental evaluation, using subjective viewing scores, we conclude that both of our proposals allow for a much better viewing experience than random packet dropping, although quality still deteriorates rapidly as the drop rate increases.
Keywords
IPTV; video coding; IPTV solutions; cellular networks; media-aware network element; media-aware packet discard algorithms; mobile TV; scalable video coding; single video stream; Cable TV; HDTV; IPTV; Land mobile radio cellular systems; Mobile TV; Multicast algorithms; Personal digital assistants; Portable computers; Streaming media; Watches;
fLanguage
English
Publisher
ieee
Conference_Titel
Information Networking, 2009. ICOIN 2009. International Conference on
Conference_Location
Chiang Mai
Print_ISBN
978-89-960761-3-1
Electronic_ISBN
978-89-960761-3-1
Type
conf
Filename
4897298
Link To Document