View
214
Download
0
Tags:
Embed Size (px)
Citation preview
Robust Scalable Video Streaming over Internet with Network-Adaptive
Congestion Control and Unequal Loss Protection
Quan Zang, Guijin Wang, Wenwu Zhu, and Ya-Qin Zang
International workshop for Packet Video, April 2001
Outline
Introduction
BackgroundSource Coding – Scalable Video Coding
Channel Coding – Reed-Solomon Code
End-to-End Architecture
Network-Adaptive Congestion Control
Network-Adaptive ULP
Simulation Result
Conclusion
Introduction
Video over Internet with QoS guarantee faces many challenges due to network bandwidth fluctuation, packet loss, and delay.
From end-to-end application-level QoS adaptation
Two main issues in video streaming scenarioResources for an end-to-end application should utilize
How to cope with the packet loss as well as the varying network condition
Background-Source Coding
Scalable Video CodingPFGS: Progressive Fine Granular Scalable
Base Layer
1st Layer
2nd Layer
3rd Layer
Background - Channel Coding
The idea of FEC across packet is to transmit redundant packets that can be used in the receiver to reconstruct lost packets
Use Reed-Solomon code for FEC
An RS code is defined RS(n ,k)Can correct t = (n-k)/2 symbol errors
Can correct t = n-k symbol erasure
n
k
End-to-End Architecture (cont.)
The Key components : MSTFP bandwidth estimation, MSTFP congestion control, ULP channel coding, target bit allocation
MSTFP bandwidth estimation use the data from Network Monitor module
Four stage:• Estimating packet loss ratio
• Estimating round trip time
• Estimating available network bandwidth
• Adjusting sending rate
Congestion Control
Network-adaptive congestion controlUsually TCP congestion control is based on Additive Increase / Multiplicative Decrease (AIMD)
Instead of using AIMD directly, they adjust the transmission rate related to packet loss ratio
Increment in transmission rate
Slight reduction in the transmission rate
Aggressive reduction in the transmission rate
Network-adaptive ULP
When the network is in good condition , more bits are allocated for source coding and fewer bits are allocated for channel coding
The layered coding framework is suited for prioritized transmission
Unequal Loss Protection are used that protects different layers using different FEC
Fixed Length Packetization
data packet
redundant packet
k information packets
n-k redundant packets
PacketLen1
Largest data packet size of layer i
redundant packet
data packet
k information packets
n-k redundant packets
PacketLen2
Stuffing
Simulation Result (1) (cont.)
The higher bandwidth , the more efficient their scheme with the other two fixed ULP scheme
The better the channel condition is, the smaller difference between their scheme and ULP schemes