SVC-Based Multisource Streaming for Robust Video Transmission in Mobile Ad-Hoc Networks

1. SVC-Based Multisource Streaming for Robust Video Transmission in Mobile Ad-Hoc Networks Thomas Schierl, Karsten Ganger, Cornelius Hellge, and Thomas Wiegand IEEE Wireless Communications, October 2006.

2. Outline • Introduction • Multisource Streaming Components • Real-Time Media Delivery in MANETs • Scalable Video Coding (SVC) • Raptor Error Correction Codes • Multisource Streaming in MANETs • Media and Channel Coding • Application Layer Protocol for Multisource Media Delivery • Simulation Results • Conclusion

3. Introduction

4. Introduction • Wireless LAN (WLAN): • 802.11a, 802.11b, 802.11g

5. Source D Source E Source C Source A Source B Introduction • Mobile Ad Hoc Networks (MANETs): Client

6. Introduction • However, MANETs’ challenge: • High-quality video transmission • Due to high path-outage probability. • Therefore, this work proposed: • Robust multisource video streaming protocol • Mainly solves the route-loss problem in case of real-time streaming over MANETs. • By using different sources at the same time withdifferent, independent representations of the media layers.

7. Introduction • Scalable Video Coding (SVC): • SVC provides layers with different importance for the video reconstruction and different percentage of the complete stream bit-rate. • Unequal Packet-Loss Protection scheme: • Protects different layers with different importance. • Based on Raptor Codes. • Generates virtually infinite amount of encoding symbols (ESs) from a certain number of source symbols (SSs).

8. Multisource Streaming Components

9. Real-Time Media Delivery in MANETs • In MANETs, each nodes operates as • a data-generating nodes

10. Real-Time Media Delivery in MANETs • In MANETs, each nodes operates as • a data-consuming nodes

11. Real-Time Media Delivery in MANETs • In MANETs, each nodes operates as • a router

12. Real-Time Media Delivery in MANETs • MANETs’ time-variant behavior • The sporadic participation of individual nodes in the network.

13. Real-Time Media Delivery in MANETs • MANETs routing algorithm: Proactive &Reactive. • In this works: • Reactive routing algorithms: • Initiate a routing query only if a packet is to be transmitted to a destination for which it has no active entry in the routing table. • Reduce routing overhead, but might also add some delay. • Dynamic MANET on-demand (DYMO) [8]. [8] I. Chakeres, E. Belding-Royer, and C. Perkins, “Dynamic MANET On-demand (DYMO) Routing,” draft version 04, IETF, Mar. 2006.

14. Real-Time Media Delivery in MANETs • Mobile multihop Ad-Hoc network in client-server setup

15. Scalable Video Coding (SVC) • SVC is an extension to the H.264/MPEG4-AVC video coding standard. • To extend the wide range of: • Temporal Scalability. • Spatial Scalability. • Quality Scalability. • An SVC bit-stream consists of a base layer and several enhancement layers. • The base layer is a plain H.264/MPEG4-AVC bit-stream for backward compatibility.

16. Enhancement Layer Base Layer Scalable Video Coding (SVC) SNR scalable coding Base layer coding Prediction Temporal scalable coding Scalable bit-stream Spatial decimation SNR scalable coding Temporal scalable coding Prediction Base layer coding

17. Scalable Video Coding (SVC) • Temporal Structure of an SVC stream including Progressive Refinement (PR).

18. Scalable Video Coding (SVC) • SNR Scalability (Quality Scalability): • The enhancement layers contain refinement quality information of the base layer in a progressive way.

19. Raptor Error Correction Codes • Raptor Code: • Mainly used in environments with packet losses. • Can produce virtually infinite amount of encoding symbols from a vector of source symbols SV of the length k. • Decoder is capable of reconstructing the source symbols from a number of ES that is only slightly higher than the original length of the SV. • Can be viewed as a serial concatenation of a pre-code and LT Code.

20. Raptor Error Correction Codes

21. E LT Code 5 6 A B D C 1 2 3 4 ‧Modified Inverse Tree-based UEPLT Encoding Graph Source symbol Ψ1 Raptor Error Correction Codes

22. I3x3 Pre-code: Gp [10] Source symbol Raptor Code Parity-Check symbol: Pre-code: Gp [10] Parity-Check symbol LT Code: GLT Raptor Error Correction Codes [10] 3GPP TS 26.346 V6.4.0, “Technical Specification Group Services and System Aspects; Multimedia Broadcast/Multicast Service (MBMS); Protocols and Codecs,” Mar. 2006.

23. Multisource Streaming In MANETs

24. Client Media and Channel Coding nl Raptor encodedsymbols Source block SB with kl symbols per layer l … Layer 3 1 k3 … Layer 2 1 k2 … 1 k1 Layer 1 tSB Source 1 Sending nls symbols Multisource transport: Source2 Source 3

25. Media and Channel Coding nl Raptor encodedsymbols Source block SB with kl symbols per layer l … Layer 3 1 k3 … Layer 2 1 k2 … 1 k1 Layer 1 tSB transmission rate: layer byte-rate: code rate:

26. Media and Channel Coding • Client behavior: • Client can influence the number of received symbols per layer, by selecting the number of sources. • Decoding is successful if the number of source symbols per layer l used for encoding is equal or higher than the minimal number of symbols kmin specified in [10]. [10] 3GPP TS 26.346 V6.4.0, “Technical Specification Group Services and System Aspects; Multimedia Broadcast/Multicast Service (MBMS); Protocols and Codecs,” Mar. 2006.

27. Source 2: Source 1: Source 3: Media and Channel Coding • The different sources are using different random seeds i for generating Ψi for the encoding process. • Ensures the generation of independent ESs for each source stream.

28. Application Layer Protocol for Multisource Media Delivery route loss probability for a path going via M intermediate links:

29. Application Layer Protocol for Multisource Media Delivery • The proposed concept: • To increase the number of used sources for enhancing reliability in server availability, while keeping the overall used network transmission rate as small as possible. • The authors assume that: • Nodes are not running in congestion state at any time. • The transmission rate at an intermediate, source, or client node is not higher than the available transmission rate provides by the air interface.

30. Application Layer Protocol for Multisource Media Delivery • The authors further assume that: • The overall probability of having a route from at least one source out of S sources to the client being Pc with having independent network paths per source. • Every source route has the same route-loss probability: Pr

31. Application Layer Protocol for Multisource Media Delivery • The protocol for source monitoring and selection probes available sources cyclically. • The authors assume that the addresses of source nodes available in the Ad-Hoc network area are introduced by an external instance. • Which is not consider in this work. • The monitoring of sources is achieved by sending probing packets (inquiry packets) to all known sources for collecting path quality in formation per source. • Sources are probed continuously during media transmission.

32. Application Layer Protocol for Multisource Media Delivery • Metric information: • The link/route quality information collected by the inquiry process. • The multisource coded network streams are requested from nodes with the best metric. • The metric used is the distance from client to the source node in terms of hops. • Motivated by equation for Pr (route loss probability), • The more nodes are used within a path, the higher the probability is the route may break down.

33. Application Layer Protocol for Multisource Media Delivery • Simplified client scheme for frequent server evaluation: ＊The protocol has been implemented and executed in ns-2.

34. Application Layer Protocol for Multisource Media Delivery • Stream management with resulting layered video quality:

35. Simulation Results

36. Simulation Results • Operation points of SVC/Single-layer media stream: • A repeated Paris sequence (288 frames) with 8640 frames (285 sec), 30 frames per second, CIF resolution, GOP size 32.

37. Simulation Results • Environment: • Area: 1000 x 600m • 40 nodes (moving in random waypoint patterns) • Scenario: • 1 client and 1, 2, 3, 4, 8, and 12 available and randomly selected source nodes. • Each simulation is repeated 60 times in independent random waypoint. • An overall simulation time of 4.75 h per value of available source nodes.

38. Simulation Results • For Raptor encoding, 3GPP-recommended preconditions [10] is adopted. • DYMO [8] is used as routing protocol in combination with an IEEE 802.11b adapter. • The average FEC stream rate is approximately 594 kb/s. • Due to packetization overhead, the effective FEC protection rate: • Multisource coded stream: 84.19 % • Single-layer stream: 86.30 % [8] I. Chakeres, E. Belding-Royer, and C. Perkins, “Dynamic MANET On-demand (DYMO) Routing,” draft version 04, IETF, Mar. 2006. [10] 3GPP TS 26.346 V6.4.0, “Technical Specification Group Services and System Aspects; Multimedia Broadcast/Multicast Service (MBMS); Protocols and Codecs,” Mar. 2006.

39. 1 Simulation Results • Average results for single-source and multisource modes.

40. Conclusion

41. Conclusion • Presented an extended unequal packet-loss protection (UPLP) scheme based on Raptor FEC using different sources for reliable media streaming in MANETs. • Showed the benefits of using linear independent FEC streams with unequal loss protection for multisource streaming in scenarios with high route-loss probability, as is present in MANETs. • This approach has been evaluated with a new protocol for media tracking and delivery in MANETs, which exclusively relies on application layer techniques. Thank you