Fountain Coding-based Video Transmission System over Heterogeneous Wireless Networks

1. Fountain Coding-based Video Transmission System over Heterogeneous Wireless Networks Presented by Hyunchul Joo POSTECH chul1978@postech.ac.kr

2. Contents • Introduction • Proposed Video Transmission System • Experiment Results • Demo

3. Introduction • Many wireless networks with different features • WLAN networks • Cover small area but support high transmission rate • UMTS networks • Cover wide area but support low transmission rate

4. Introduction • Vertical handoff • Switching technology between different networks to support seamless service • MN can only access the single network at a time • Needs sophisticated architecture and mutual agreement among network service providers • Some transition time is inevitable

5. Introduction • Path diversity • MN establishes multiple paths over multiple wireless networks • Advantages • Parallel transmission • Increase the total bandwidth • Disperse the network load • Eliminate the vertical handoff delay • Without the cost of cooperation on the network infrastructures

6. Introduction • We assume that UDP (User Datagram Protocol) has to be used for delay sensitive video streaming • The distortion caused by packet loss is more severe than source encoding rate decrease • The network path with high packet loss rate is useless until we deploy FEC (Forward Error Correction) coding such as the fountain code

7. Proposed System • Fountain Code • Rateless & Erasure code • Generate endless encoded symbols from finite source symbols • If a sufficient number of encoded symbols are available, all source symbols can be recovered although some encoded symbols are lost • Type • Online code • LT code • Raptor code • Etc..

8. Proposed System • The original data have to be divided into source blocks • The basic unit of fountain coding • Source block size = GOP • One source block impair can affect one more GOPs if the source block is not synchronized with one GOP • Each source block is divided into source symbols and then applied to the fountain encoding

9. Proposed System • Based on the feedback of receiver, the transmission rate of each path and the fountain code rate are dynamically determined to maximize video encoding rate subject to delay and source block loss constraints • Simple weighted round-robin scheduling is enough to distribute packets among possible paths because all packets has the same priority BW, delay, PLR

10. Experiment Results • Experiment set-up • 802.11b and 802.11g networks are used as access technology for each path • The JM reference software of H.264/AVC and original JM rate control mechanism are used • CIF format HARBOUR video sequence is adopted as the test sequence • The GOP structure is IPPPPPPPPPPPPPP and the number of frames in a GOP is set to 15 802.11b 802.11g

11. Experiment Results • The number of successfully received packets • The number of received packets through proposed system are always higher than K’ • Performance of the proposed system is superior to other cases that utilize the single path GOP number

12. Experiment Results • PSNR comparison • In RS code systems • When the single reliable path is used, decoding is always successful with low PSNR value • When two paths is used, decoding is sometimes incomplete • Proposed system can provide the smooth video with high quality because source symbols are always decoded with high PSNR value GOP number

13. Demo • Test-bed overall structure • Java is used to implement the proposed system

14. Demo