nus.soc.cs5248-2007 roger zimmermann (based on slides by ooi wei tsang) 1 error recovery

1

NUS.SOC.CS5248-2007Roger Zimmermann (based on slides by Ooi Wei Tsang)

Error Recovery


2

Problems with Internet

Limited BandwidthVarying ConditionsDelay JitterPacket LossDelayHeterogeneity :


3

Network

Encoder

Sender

Middlebox

Receiver

Decoder

You Are Here


4

Overview

Characteristics of the InternetGeneral techniquesError recovery for audioEffect of loss on MPEGError recovery for MPEG

5


Loss Characteristics of The Internet


6

Characteristics of Internet

60-70% of paths do not show any loss

Those with loss have an average of 4.5 – 6% packet loss

[Paxson97] End-to-end Internet packet dynamics


7

Packet Loss Pattern

1000

30

Number of Occurrences

Burst Length


8

Characteristics of Internet

Bursts of loss are typically short (2-3 consecutively loss packets)

Long burst do occur

Burst may occur periodically


9

Wireless Link

Loss rate measured in my office 10 ~ 20%

up to 50% reported!


10

Gilbert Model

GOOD BAD

11


A Survey of Packet-Loss Recovery Techniques for Streaming Audio

Perkins, Hodson and HardmanIEEE Network Magazine 1998


12

Retransmission

1 2 3 4

1 2 4

3

3

X


13

Redundant Data

1 2 3 4

1 2 4

2

2 3

3 4 5

53


14

2 4

Error Concealment

1 2 3 4

1

15


Retransmission

for audio


16

Pros/Cons of Retransmissions


17

Scalable Retransmission

On packet loss T = random(0, RTT) wait for T multicast NACK

On receiving NACK from others suppress own NACK


18

Retransmit when

Group size is small Loss rate is lowLarge latency acceptable


19

Selective Retransmission

Retransmission based on priority of packets

Important/urgent packets are retransmitted first

20


Redundant Data

for audio


21

Parity Forward Error Correction

1011001

1000010

0001001

XOR 0010010


22

Parity Forward Error Correction

1011001

1000010

0001001

0010010


23

Parity FEC Ordering


24

Parity FEC Ordering


25

Parity FEC Ordering


26

Parity FEC Ordering


27

Parity FEC Ordering


28

Reed-Solomon Code

RS(n,k)

RS

n k

(popular: n = 223) (popular: k = 32)

Parity symbolsData symbols


29

Reed-Solomon Code

Data block (length n+k, e.g., 255)

Can correct up to 16 corruptedsymbols per block (32/2)


30

Media Specific FEC

1 2 3 4

1 2 4

2

2 3

3 4 5

53


31

Pros/Cons of Redundant Data


32

Interleaving


33

Pros/Cons of Interleaving

34


Error Concealment


35

Effect of Loss on Audio

Speech Human ears can interpolate Loss up to length of phoneme can

still be tolerable


36

Insertion-based Repair

SpliceSilence SubstitutionNoise SubstitutionRepetition


37

Other Repair Methods

InterpolationRegeneration


38

Error Concealment

More complex error concealment algorithms provide better performance (i.e., playback quality)

39


Colin’s Recommendations


40

Non-Interactive Apps

InterleavingFECRetransmission for unicast only


41

Interactive Applications

Media Specific FEC


42

Error Concealment

Repeat

43


Packet Loss Effects on MPEG Video Sent over the Public Internet

Jill Boyce and Robert GaglianelloACM Multimedia 1998


44

Measurement-based Study

Need to understand the problem before proposing solution


45

Data Gathering Method

From: NYC13 , Austin21, London18

To: Holmdel, NJ


46

Data Gathering Method

Video: Two 5-mins MPEG 30 fps 384 kbps and 1 Mbps IBBPBBPBBPBBPBB QSIF 176x112 and SIF 352x240 One row per slice


47

Average Packet Loss

0

2

4

6

8

10

12

14

Packet Loss Rate

(%)

New York Texas London

1Mbps384kbps


48

Frames Affected by Errors

% Frames in Error

Packet Loss Rate

49


Overview of Error Recovery for Video


50

Methods

Retransmission InterleavingError ConcealmentFECLimiting Error Propagation

Reference Frame Selection


51

Reference Frame Selection

I P B B P

X

I


52

Methods

Retransmission InterleavingError ConcealmentFECLimiting Error Propagation

Reference Frame Selection Changing Temporal Pattern

53


Case Study: HD video conferencing

Korean Air, Inha University (Seoul), University of Southern California, Pratt & Whitney

© PWICE, 2006 54

USC-Inha Two-way ExperimentsUSC (Powell Hall of Engineering)

July 2005

© PWICE, 2006 55

USC-Inha Two-way ExperimentsInha University (Memorial Library)

July 2005

© PWICE, 2006 56

KAL Network Infrastructure

ICNGMP

PUS

ISP (KT)ISP (KT)

50Mbps50Mbps

50Mbps50Mbps

2Mbps2Mbps

ICN: IncheonGMP: Kimpo (Seoul)PUS: Kimhae (Pusan)

Dedicated LinkPublic Link

10.222.6.222

10.55.33.222

210.105.6.4

210.113.150.1110.71.31.222

• Korean Air network infrastructure between Incheon, Kimpo, and Pusan

• KAL network &external ISP (KT)

• 1 Gbps:ICN-GMP

• 50 Mbps:GMP-PUS

December 2005

© PWICE, 2006 57

Network Measurements

• End-to-end packet loss rates of the network path between Kimpo (GMP) and Incheon (ICN) observed during November 15-16, 2005

• Network is shared with other business traffic• Loss rates are quite low

Data Loss

0.00%

0.20%

0.40%

0.60%

0.80%

1.00%

1.20%

1.40%

1.60%

10

:47

:42

12

:21

:02

13

:54

:23

15

:27

:43

17

:01

:03

18

:34

:23

20

:07

:43

21

:41

:04

23

:14

:24

0:4

7:4

4

2:2

1:0

4

3:5

4:2

4

5:2

7:4

4

7:0

1:0

4

8:3

4:2

4

10

:07

:45

11

:41

:05

13

:14

:25

14

:47

:45

Data Loss

November 2005

© PWICE, 2006 58

Hardware Preparation

• 3 computers purchased in Korea• Computers sent to USC with KAL• Software setup in the IMSC laboratories

• Linux installation• RCS software installation• Retransmission configuration• End-to-end equipment test with local JVC cameras and displays

• Computers sent back to KAL HQ

December 2005

© PWICE, 2006 59

Hardware Test at KAL

• Re-configuration for Sony HDR-HC1 cameras

December 2005

• Local area network tests

© PWICE, 2006 60

Experiments: ICN ◄▬► GMP

• Asymmetric environment: 88 Mb/s & 50 Mb/s• Conclusion: ICN - GMP: sufficient BW for HD video• Sufficient headroom for data applications

December 2005

Bandwidth measurements Visual quality

Required by RCS

© PWICE, 2006 61

Experiments: ICN ◄▬► PUS

• Asymmetric environment: 6 Mb/s & 1 Mb/s• GMP - PUS: initially insufficient BW for HD video• Investigated network routers with KAL IT help• Conclusion: replaced router to improve throughput; to be re-tested

December 2005

Bandwidth measurements Visual quality

© PWICE, 2006 62

Measurements: Packet Loss Rate

• Path: ICN - PUS• Two packet sizes: 564 bytes and 940 bytes• Conclusion: GMP - PUS loss rate very high

December 2005

One-way packet loss rate Two-way packet loss rate

© PWICE, 2006 63

Packet Loss Error Recovery

• Selective retransmission protocol is used to recoverfrom lost data

• Effectiveness of protocol is tested in lab environment• Gilbert Model is used to induce losses into

transmission• Transmission delay is chosen to be 10 ms

(expected latency inKorea)

Sender Receiver

DelayModel

PacketLoss

Model

Retransmission request

Original data+ (retransmitted data)

© PWICE, 2006 64

Packet Loss Error Recovery

• Selective retransmission protocol

10% Loss Rate 5% Loss Rate 1% Loss Rate

With Retransmissions With Retransmissions With Retransmissions

Sept. - Dec. 2005

(ICN-GMP Link)

65


Error Control Techniques for Interactive Low Bitrate Video Transmission over The Internet

Injong RheeSIGCOMM ‘98


66

Basic Idea

“Better Late Than Never!”, orLate packet is still useful


67

MPEG Frame Pattern

I B B P B B P

I P B B P B B


68

H.261 Error Propagation

I P P P P P PX X X X X X

I P P P P P PX X X X

retransmission

loss


69

H.261 Frame Pattern

I P P P P P PX

loss


70

I P P P P P P

H.261 Frame Pattern

X

loss

X X

retransmission


71

PTDD

Periodic Temporal Dependency Distance

Large PTDD Higher Chance of Repair Longer Error Propagations Less Temporal Redundancy


72

QAL

I P P P

I P P P Base Layer

EnhancementLayer


73

QAL

I P P P

I P P P Base Layer + FEC

EnhancementLayer

X


74

QAL + PTDD

I P P P

I P P P Base Layer

EnhancementLayer


75

Error Propagation


76

Recovery from Error Propagation


77

Frame “quality”

PSNR =


78

PSNR vs. Loss Rate

Packet Loss Rate

PSNR

79


Today’s Summary


80

How to recover packet loss

RetransmissionFECError Concealment


81

Limit the damage of error

InterleavingKey frame selectionChange reference frame

nus.soc.cs5248-2007 roger zimmermann (based on slides by ooi wei tsang) 1 error recovery

Documents

roger zimmermann

audioeffect of loss

loss recovery techniques

small loss rate

measurementbased studyneed

loss packetslong burst

mpegerror recovery

length of phoneme