1. A Quality-Driven Cross-Layer Solution for MPEG Video Streaming Over Wi-MAX Networks
Presented by Guided by:-
SPARSH GOEL, KARAN CHAUHAN Dr.Jonathan C.L. Liu
Graduate Student, CISE Associate Prof.(CISE)
University of Florida University of Florida

2. Need of Quality driven solution?

3. Satisfying users with different QoS requirements while optimizing the resource allocation is a challenging problem.

4. In this paper, the authors discuss the challenges and possible solutions for transmitting MPEG video streams over WiMAX networks.
A cross-layer solution for enhancing the performance of WiMAX networks with respect to MPEG video streaming applications is explained

5. Other Challenges High quality and bitrate. Low Latency.
Less Complexity.
Traffic maintenance.
Other Performance parameters.

6. Excessive bandwidth demand is there to provide mobile users with high speed network connectivity in order to support variety of applications.

8. Introduction to Wi-Max
WiMAX networks are based on the IEEE standards.
IEEE standards defined different QoS classes to support a broad range of applications with varying service requirements.
In WiMAX networks, each traffic flow is mapped into an appropriate service class

9. Selecting appropriate service classes with proper parameters to support the required QoS while not wasting the scarce resources can increase the utilization of network.
An application driven, traffic aware service classification will provide the WiMAX subscriber stations (SSs) with better video quality.

10. The WiMAX technology is promising to provide broadband wireless access to mobile users in the near future.
Video streaming will be a very attractive application for the rapid deployment of WiMAX networks.

11. Previous Work
Network adaptive (Feedback based) video rate control design for enhancing the overall received video quality.
The server receives some information such as:-
The available bandwidth
Loss rate
Buffer size at the receiver
End-to-End delay to adapt to the optimum video coding rate.

12. Drawbacks Channel variations in the wireless networks.
Sending feedback is not a feasible option in some multicasting applications such as IPTV or Mobile TV.
Increases the computational complexity at the video server which results in overloading the video streaming servers.

13. Transmission of same content to multiple receivers with different physical channels.
Complicated to attain the flexibility and efficiency in the case of heterogeneous receivers.
Unable to track the fast changes in radio channel conditions and adapt to the optimum rate accordingly.

14. Other Approaches Scalable video coding (SVC)-
The goal of this method is to encode high quality video streams into some group of bit streams including one base sub layer and multiple enhancement sub layers.
A channel based, rate adaptive solution for video streaming in WiMAX network.

15. Active frame dropping approach-
In active frame dropping approach the base station drops a frame if it does not have enough confidence about successfully delivering a video frame within the application delay limit.
Frame discarding solution-
A frame discarding solution based on the packet lifetime. This method, frames that cannot meet the deadline are dropped by the video server or the intermediate routers.

16. In this paper, the authors have suggested a method that exploits the characteristics of MPEG traffic to give priority to the more important frames and protect them against dropping.
In this method, no real-time feedback is sent to the video server. Thus, a video server will be able to support multiple clients simultaneously, and this makes the solution flexible for multicasting applications as well

17. Author’s Approach
The characteristics of MPEG traffic are used to give priority to the more important frames and protect them against dropping.
Video streaming traffic model.
Quality-of-service support in WiMAX networks.
Proposed solution for video streaming in WiMAX.
4) Simulation results.

18. 1) Video streaming Traffic model
MPEG-2 and MPEG-4 are used for reducing the redundant information.
Consecutive scenes have few differences and the information in their pictures is highly correlated.
Data compression is achieved by coding the small differences between the scenes.
Much less data needs to be transported.
Bitrate is application specific

19. MPEG-4 encodes the input video into a sequence of frames called group of picture (GOP).
The number of GOPs generating a scene is a function of the scene complexity and compression ratio.
I (intra coded) frames are used as the reference frame in each GOP and for synchronization of all frames in a GOP.
If a GOP is lost or corrupted, the next GOP will be built based on its I frame which is not coded using any other frames.

20. P (predicted) frames are built by predicting the changes from the closest match in the preceding I or P frames of their GOP.
B (bi-predictive) frames use previous I or P frame and the next P frame to predict the changes in the picture.
Therefore, loss of I or P frames will affect some other frames in their corresponding GOP and this will degrade the perceived quality.
These frames maybe fragmented over network.

21. Traffic Modelling
A scene is a short part of the movie that does not contain sudden changes in the view while it can possibly include some zooming or object movement.
I frames exhibited different behavior at different time scales.
At the shorter time scales of a few seconds, the bitrate varies a little around a mean level. However, the mean level varies tremendously at larger time scales.

22. Fig :- Source Model.

23. The model considers two independent components for defining the size of the nth I frame of the video stream, located at the Kth scene:-
is the mean activity of scene and represents the large variations.
is constant for all I frames in scene while it will be different for other scenes.
represents the small variations of the I frames around the mean level of each scene.

24. The is modeled by an order two autoregressive process:-
is a normal random variable with zero mean and constant variance for each stream (depends on content of video).
The model generates 2 I frames per second, 8 P frames per second, and 20 B frames per second.
The average bitrate for each frame type are :- 273Kbps , 588Kbps , and 1094Kbps.

25. Fig:- MPEG Model parameters used.

26. 2) Quality of Service (QoS) in WiMax
Video streaming applications generate variable bitrate traffic which is real-time and delay sensitive.
Such applications require the network to allocate network resources to handle the corresponding traffic within a limited period of time.
It is observed that the real-time polling service (rtPS) class of WiMAX standard is suitable for supporting video streaming traffic.

27. In the rtPS class, each traffic flow is characterized by a few parameters such as the minimum reserved traffic rate and the maximum sustained traffic rate.
An rtPS traffic flow will not get admitted into the network if the BS cannot guarantee it the requested minimum reserved bitrate.
The BS periodically polls the rtPS queues and assigns resources based on the bandwidth requests and connection parameters

28. 3) Multilevel Service Classification
The video server indicates the type of each frame in the Type of service(ToS) field of the IP header.
BS can distinguish the frame type of each packet.
MPEG Frames are mapped into three different rtPS classes with different minimum reserved bandwidth parameters(determined by SS or video server).
Video stream is projected into three different traffic flows.

29. Fig:- Multilevel service classification.

30. As I frame is the base frame in GOP, I Frames are protected against dropping since loss of I frame propagates through the GOP.
This prevents other frames also from being corrupted.
Similarly P frames affects proceeding P and B frames and are protected.
The loss of B frame doesn’t propagate and result in smaller quality degradation.

31. The DL (Downlink) traffic from the video server to a WiMAX SS is en- queued at its corresponding DL queue in the BS.
If the traffic flow is admitted as an rtPS flow, the BS has to guarantee the requested minimum reserved bitrate for that.
If the low bitrate is requested then Video quality will degrade - DL queues becoming full because input traffic rate higher
While requesting higher bitrate for flows will decrease the network utility as less number of flows will be admitted.
There exists a trade-off between the video quality and network utility which this is optimized by choosing the optimum minimum reserved bitrate value for each video stream.

32. Inspired by the fact that I frame has the most effect on video quality, more minimum reserved bitrate is allocated to the rtPS flow corresponding to the I frames .
In order to minimize the probability of dropping I, a bitrate that is higher than the average bitrate of I frames is used.
minimum reserved bitrates for the P and B frames are also requested.

33. Fig:- Network Architecture.

34. Simulation Results
OPNET simulator and its WiMAX package is used to perform the simulations.
The PHY and MAC layers of WiMAX are fully simulated In order to achieve more accurate results in the simulations.
The simulations are run under highly traffic loaded conditions of the network and similar PHY characteristics and video traffic for each SS for both the schemes.

35. Conventionally, each SS asks for one rtPS traffic flow with the minimum reserved bitrate equal to 884 Kbps.
In the multilevel traffic classification scheme, each SS asks for three rtPS flows corresponding to three different MPEG frame types.
The SS will request 384 Kbps for I frames, 300 Kbps for P frames, and 200 Kbps for B frames.

37. Fig:- BS Performance 1

38. Fig:- BS Performance 2

39. Fig:- Client Performance 1

40. Fig:- Client Performance 2

41. Fig:- Multilevel Classification Performance 1

42. Fig:- Multilevel Classification Performance 2

43. Fig:- Avg. number of Frames received by all SSs

44. Conclusion
A cross layer design to enhance the quality of MPEG video streaming for the end user in WiMAX networks is proposed.
Unlike other approaches, more priority is given to more important frames and protecting them against dropping without increasing the complexity of the video server.
The proposed solution for multicasting applications elaborates QoS classification feature of WiMAX and it is reliable, flexible and scalable.

45. By providing the BS with information about the type of video frame and by map I, P, and B frames into three different rtPS service classes with different service requirements, the overall number of frames delivered to each SS increases.
This results in enhanced quality of video at the end users.

46. Thank you