Microsoft PowerPoint - IIT - IPTV Performance Measurement

26
IPTV Performance Measurement Alan Clark Telchemy

description

 

Transcript of Microsoft PowerPoint - IIT - IPTV Performance Measurement

Page 1: Microsoft PowerPoint - IIT - IPTV Performance Measurement

IPTV Performance Measurement

Alan ClarkTelchemy

Page 2: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Outline

• IPTV Performance • Performance Measurement Approaches• Analyzing Encrypted Video Streams• Telchemy’s IPTV Test Lab

Page 3: Microsoft PowerPoint - IIT - IPTV Performance Measurement

IPTV Performance

• Transmission Impairments– Packet loss

• Codec configuration– Quantization level– GoP Length– Slice Structure– Bandwidth limitation

• Interaction of loss with video frame structure

Page 4: Microsoft PowerPoint - IIT - IPTV Performance Measurement

IP Video basics – 1 – I Frames

I B B P B

GoP Structure- I frames are typically large ( 00’s to 000’s of IP packets per frame)- P and B frames are typically much smaller (0’s to 00’s of packets

per frame)- GoP length – typically 15-250 frames in length

Page 5: Microsoft PowerPoint - IIT - IPTV Performance Measurement

IP Video basics – 2 – I Frames

Macroblocks

Page 6: Microsoft PowerPoint - IIT - IPTV Performance Measurement

IP Video basics – 3 – I Frames

Slices

Page 7: Microsoft PowerPoint - IIT - IPTV Performance Measurement

IP Video basics – 4 – P/B Frames

Motion encoded

Page 8: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Impact of lost packet

I B B P B

Error in I frame willpropagatethrough following P and B frames Error in

B frame willonly impact this frame

Error in P frame willpropagatethrough following P and B frames

Page 9: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Example – I frame vs B frame loss

• I frame example – 161 packets lost– MOS-V = 3.3– Loss Rate = 0.27%

• B frame example – 161 packets lost– MOS-V = 3.7– Loss Rate = 0.27%

Page 10: Microsoft PowerPoint - IIT - IPTV Performance Measurement

IPTV Performance Measurement Approaches

• Full Reference– ITU J.144, PEVQ, PSNR…

• Zero Reference– “Bitstream” algorithms (e.g. IP based)– Decoded video analysis (pixel based)

• Partial Reference– Feature extraction/ comparison– Augmented zero reference algorithms

Page 11: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Industry Standards Activities

• ATIS IPTV Interoperability Forum (IIF)– QoS Metrics Task Force/ ATIS 0800008 Metrics– Test & Interoperability Task Force

• ITU– SG9 – Standards for objective video test (J.nnn)– IPTV Focus Group– SG12 – Lead SG on QoS

• Video Quality Experts Group (VQEG)– Testing of objective video quality algorithms

• DSL Forum– WT135 – IPTV STB management

Page 12: Microsoft PowerPoint - IIT - IPTV Performance Measurement

IP Based “Bitstream” algorithms

• Efficiently measure the performance of high bandwidth IP Video streams

• In IPTV services - content of RTP/MPEG-TS is typically encrypted

• Ideally – report– Video MOS– Audio MOS– I/B/P frame statistics– RTP/MPEG TS statistics– IP Statistics

Page 13: Microsoft PowerPoint - IIT - IPTV Performance Measurement

VQmon/HD

UDPRTP/ MP2T

StreamDemultiplex

VideoAssessment

AudioAssessment

IPStatistics

Page 14: Microsoft PowerPoint - IIT - IPTV Performance Measurement

VQmon/HD – Video Assessment

ContentAnalysis

FrameDetection

Per-FrameQuality

Computation

PerceptualQualityModel

ContentProblemDetection Content

Alerts

VideoMOS

TR101290Metrics

Page 15: Microsoft PowerPoint - IIT - IPTV Performance Measurement

VQmon/HD – Frame Detection

PictureHeader

Decoding

TransportPacket

Processing

HeuristicFrame

Detection

I/B/PFrameMetrics

Scrambled/Encrypted

Unscrambled

Per-FrameLoss

PatternMetrics

StreamDescriptors

(BW..)

Page 16: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Frame detection in encrypted streams

• Detect frame boundaries within the packet stream

• Measure frame size• Analyze sequence of frame sizes

– Low motion/ high detail – obvious GoP structure– High motion/ low detail – GoP size less obvious– I frames can be inserted when scene changes

• Assess content type

Page 17: Microsoft PowerPoint - IIT - IPTV Performance Measurement

VQmon/HD – Perceptual Quality

ImpactOf

Loss

Impact ofCodec &Bit rate

IntegrationModel

TemporalMasking

Frame Quality

Perceptual Quality

Content Analysis

SceneChange

Detection

ContentSensitivity

AggregateStatistics

ContentType

Detect

MOS-VMOS-AAbsolute vs Relative

Page 18: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Example 2 – Hero_D02

• Detected GoP size 9, IBBPBB..• Measured bit rate 5542 kbits/sec• Image size 720x480, 29.97 frames/sec• MOS-V = 3.7• MOS-A = 3.7• EPSNR = 31.0 dB• Packet Loss Rate 0.11%• I frame packets = 16108 rcvd, 28 lost• P frame packets = 27551 rcvd, 42 lost• B frame packets = 29984 rcvd, 14 lost• MPEG Transport stats

– Continuity Error count 12– PCR Repetition Error count 285

Page 19: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Example 3 – Hero-05

• Detected GoP size 9, IBBPBB..• Measured bit rate 5541 kbits/sec• Image size 720x480, 29.97 frames/sec• MOS-V = 1.8• MOS-A = 1.8• EPSNR = 21.0 dB• Packet Loss Rate = 1.55%• I frame packets = 15518 rcvd, 602 lost• P frame packets = 26886 rcvd, 483 lost• B frame packets = 30166 rcvd, 56 lost• MPEG Transport stats

– Continuity Error count 163 – PCR Repetition Error count 282– PCR Discontinuity Error count 3

Page 20: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Example 4 - Scrambled vs Unscrambled Video – Chan20_1

• Unscrambled– MOS-V = 3.1– EPSNR = 27.5– VSTQ = 41.5– I frame packets

• Received 6225• Lost 42

– P/B frame packets• Received 20779• Lost 40

– IP Statistics• Loss rate = 0.23%

• Scrambled– MOS-V = 3.0– EPSNR = 27.0– VSTQ = 41.5– I frame packets

• Received 6367• Lost 42

– P/B frame packets• Received 20637• Lost 21

– IP Statistics• Loss rate = 0.23%

Page 21: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Telchemy’s IPTV Test Lab

SAEncoder

AppleDarwin

VLC

TIDaVinci

IPEmulator

SA STB

TIDaVinci

Sigma

ST Micro

Phillips

MediaPlayer

VLC

X

TraceLibrary

TCPReplay

BlackMagic

Capture

IPMulticast

Subjective test- SS DCR/ACR

Objective test- PEVQ- VQM (J.144)- PSNR

Page 22: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Example test - Noticeability

• Introduced impairment– Small number of macroblock errors (1, 4, 8)– Frame freezes

• Varied duration of impairment• Subject test

– Was impairment• Not noticeable• Just noticeable• Definitely noticeable

Page 23: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Frame Freeze

Noticeability of frame freeze

0102030405060708090100

0 50 100 150 200

Duration (ms)

No

tice

ab

ilit

y All

Flowers

Mobile

News

Tempete

Page 24: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Macroblock Errors

Noticeability of Macroblock Errors (high contrast)

0102030405060708090100

0 50 100 150D ura t io n o f e ve nt ( ms )

All

Flowers

Mobile

News

Tempete

Page 25: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Current test project

• Impaired capture files, replayed through ADB Set Top Box to HD monitor

• MPEG-2 and H.264 encoder• H.264 – wide range of encoder configurations

• Single Stimulus, Degradation Category Rating test

• Series of tests – each with 60 x 10 second video clips, wide range of content

• Presentation order randomized for each viewing panel

• At least 4 different viewing panels

Page 26: Microsoft PowerPoint - IIT - IPTV Performance Measurement

Summary

• IPTV Performance • Performance Measurement Approaches• Analyzing Encrypted Video Streams• Telchemy’s IPTV Test Lab