1. CS590F Project: Wireless Streaming Protocol Xiaojun Lin Jitesh Nair Samrat Kulkarni

2. Schedule Introduction and motivation Assumptions and goals Techniques: –Differentiating the loss –Quality control using FEC –TCP-friendly rate control Coupling between CC and FEC Contributions & conclusion

3. Introduction Background –Proliferation of Multimedia Content –Increasing rate of Wireless systems 3G:UMTS(384k to 2M), EDGE(384k). –More Scalable Video Coding Technology MPEG – 4 etc. –Industries: www.solidstreaming.comwww.solidstreaming.com Wide Acceptance of Video to handheld is possible!

4. Assumptions and Goals Wired-wireless interaction Wireless channels is at the last hop Competes with other flows on the internet Wireless loss at the last hop Internet wireless Video Source Wireless Gateway Mobile

5. Objective Our Wireless Streaming Protocol will –Maximize perceived quality at the receiver given network conditions. –Live Friendly with TCP on wired Internet –Responsive to changing network conditions.

6. INTERNET DECOUPLE THE LOSSES SEEN BY SENDER INTO WIRELESS AND WIRELINE LOSSES LEVERAGING THE KNOWLEDGE OF THE WIRELESS GATEWAY IN DETECTING CONGESTION LOSSES. DETECTING CONGESTION AND WIRELESS LOSSES

7. GLOBAL_SEQUENCE_ : This is a non negative monotonically increasing number. Each packet has a unique GLOBAL_SEQUENCE_ number. WG_COUNT_ :This is a non negative monotonically increasing number set by the wireless gateway. DIFFERENCE IN THE VALUE OF THE GLOBAL_SEQUENCE_ AND WG_COUNT_ INDICATES A CONGESTION LOSS DIFFERENCE IN THE VALUE OF THE WG_COUNT_ AND THE NET PACKETS RECEIVED AT THE RECIEVER INDICATES THE WIRLESS LOSS GlobalWG_count 100 101 103102 105104 KEY VARIABLES

8. Packet Format From Sender: GLOBAL_SEQUENCE_ WG_COUNT_ –For each block: block indicates a chunk of data which is packetized. This is in the FEC scheme. (think of GOP). FEC_MAX_: block size FEC_MIN_: minimum number of packet required to decode. LOCAL_SEQUENCE_ : This is a non negative monotonically increasing number. Each packet in a block has a unique local sequence number. From Receiver PACKETS_EXPECTED CONGESTION_LOSS WIRELESS_LOSS

9. Quality Control using FEC Quality controller Fair rate Loss Prob. Right amount of redundancy. Scalable Codecs: e.g. MPEG4 FEC code: fec_min_, fec_max_ Quality perceived at the receiver

10. Quality Control using FEC Model for perceived quality at the receiver: Let m be the number of packets arrived at the receiver Maximize “expected quality” over fec_min_ subject to

11. TCP Friendly Rate Control Act on Congestion Loss only Using the approach in Floyd’s paper.

12. The Whole Picture Sender send out FEC-coded stream. Receiver collect statistics related to congestion loss and wireless loss At the sender: –Congestion loss  raw rate –Raw rate + loss  redundancy, layer, net_rate –Raw rate = net_rate * fec_max_ / fec_min_

13. Contribution and Conclusion Combined “Source/Channel Coding” for transmitting multimedia to wireless terminals. require assistance at the wireless gateway/basestation to detect different source of losses. First step towards multimedia to your hand!