44
7/31/2019 Packet Error Rate http://slidepdf.com/reader/full/packet-error-rate 1/44 Video Communication over Wired and Wireless CDMA Networks with Optimal Bandwidth Allocation Yushi Shen Advisors: Prof. Pamela C. Cosman Prof. Laurence B. Milstein 2005 June 20th University Qualifying Exam 1

Packet Error Rate

Embed Size (px)

Citation preview

Page 1: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 1/44

Video Communication over Wired

and Wireless CDMA Networkswith Optimal Bandwidth Allocation

Yushi Shen

Advisors: Prof. Pamela C. Cosman

Prof. Laurence B. Milstein

2005 June 20th University Qualifying Exam 1

Page 2: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 2/44

Introduction• Area: Video communications

• Channel model:Tandem wired and wireless CDMA networks– Wired component: Packet erasures

– Wireless component: Burst errors• Direct-sequence (DS) CDMA technology

Three components:

– Source Coding– Channel Coding– Spread Spectrum

2005 June 20th University Qualifying Exam 2

Page 3: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 3/44

Introduction

2005 June 20th University Qualifying Exam 3

Page 4: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 4/44

Introduction• Problem:

Bandwidth allocation among source coding, channelcoding and spreading, with a target frame rate, under thetotal bandwidth (chip rate) constraint:

 R s: source bit rate (bit/s);  r

 c: channel coding rate;

 M : spreading gain; W : total bandwidth (chip rate, chip/s).

• Adaptive allocation: – Changing channel characteristics

 – Different source content

2005 June 20th University Qualifying Exam 4

Page 5: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 5/44

Talk Outline Introduction

2. Computation of Packet Error Rate3. Tradeoff between Channel Coding and

Spreading

4. Source Encoding with Optimal Mode

Selection

5. Optimal Bandwidth Allocation6. Simulation Results

7. Future Work

2005 June 20th University Qualifying Exam 5

Page 6: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 6/44

Computation of Packet Error RatePacket Error Rate (P):

• The key factor for the performance

• Determined by:

– Packet erasure rate introduced by wired link: Pe

– Packet drop rate due to uncorrectable bit errors by

wireless link: P p

• Total packet error rate:

P = Pe + P p - Pe x P p

2005 June 20th University Qualifying Exam 6

Page 7: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 7/44

Computation of Packet Error RatePacket Drop Rate due to Wireless Link (P p):

– y : fixed packet length

– Ad : weight distribution of block codes

– : received SNIR (signal-to-noise-plus-interference-ratio)

r γ 

2005 June 20th University Qualifying Exam 7

Page 8: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 8/44

Computation of Packet Error RateComputation of Weight Distribution ( Ad ):

• Computed based on the transition matrix sequence

• Computed

– For each RCPC code (characterized by rate  r c)

– Given fixed packet length ( y)

2005 June 20th University Qualifying Exam 8

Page 9: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 9/44

Computation of Packet Error Rate

Computation of received SNIR ( ):

=

– : channel gain;

K : # of users;  L: # of resolvable paths;

 E b: energy per source bit;  N 0: energy of Gaussian noise;

: exponential multipath intensity profile

– Result for sufficiently large number of users

• Multi-access-interference is asymptotically Gaussian

• Self-interference is negligible

2005 June 20th University Qualifying Exam 9

(forward)

α 

λ 

Page 10: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 10/44

Computation of Packet Error RatePacket Drop Rate due to Wireless Link (P p):

– Weight distribution ( Ad ), which is determined by

• Structure of RCPC code (characterized by r c)

• Fixed packet length ( y)

– Received SNIR ( ), which is determined by

• System parameters (K , L, E b, N 0)

• Current channel conditions ( )

• Channel coding rate (r c ) and spreading gain ( M )

α 

2005 June 20th University Qualifying Exam 10

Page 11: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 11/44

Computation of Packet Error RateConclusion: Given system parameters, and conditioned on

channel characteristics, packet error rate is fully determinedby channel coding rate (r c ) and spreading gain ( M ). (forward)

X axis: Received SNIR

(dB)

Y axis: Probability of packet error

Rate 1/3 RCPC code

Rate 2/3 RCPC codeRate 8/9 RCPC code

No FEC used

2005 June 20th University Qualifying Exam 11

Page 12: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 12/44

Talk Outline Introduction

Computation of Packet Error Rate

3. Tradeoff between Channel Coding and

Spreading

4. Source Encoding with Optimal Mode

Selection

5. Optimal Bandwidth Allocation6. Simulation Results

7. Future Work

2005 June 20th University Qualifying Exam 12

Page 13: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 13/44

Tradeoff between Channel Coding and Spreading,

under a Target Packet Error Rate

• r c and M  the packet error rate, given the system

parameters and conditioned on the channelcharacteristics.

• r c and M  the source bit rate ( Rs), under the

bandwidth constraint.

• Basic idea:

– Choose the (r c , M ) for each packet that maximizes Rs, among all the (r c , M ) combinations that achievethe target packet error rate.

2005 June 20th University Qualifying Exam 13

Page 14: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 14/44

Tradeoff between Channel Coding and Spreading,

under a Target Packet Error Rate

Algorithm:

• Pre-calculate the relationship between packet error andreceived SNIR, and solve the SNIR threshold under the

target packet error rate, for each RCPC code. (example)

• During transmission

– Estimate the current channel characteristics;

– For each r c , calculate the smallest M that achieves the

SNIR threshold for the target, and get a (r c , M ) pair;

 – Among all the (r c , M ) pairs, choose the one that

minimizes the ( M / r c) ratio to maximize Rs.

2005 June 20th University Qualifying Exam 14

Page 15: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 15/44

Tradeoff between Channel Coding and Spreading,

under a Target Packet Error Rate

It can be shown the encoder will actually select the

same r c , and only vary M , most of the time.• For each r c , denote the SNIR threshold for thetarget packet error rate

• Solve M in the SNIR equation and plug in ( M / r c): (see)

2005 June 20th University Qualifying Exam 15

• To minimize ( M/ r c ) is equivalent to maximize

• RCPC code with lowest rate has the highest ratio

≈cr 

 M 

Page 16: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 16/44

Tradeoff between Channel Coding and Spreading,

under a Target Packet Error Rate

Conclusion:

• The tradeoff between channel coding andspreading:

– Achieve the target packet error rate

– Maximize the source bit rate

– Satisfy the bandwidth constraint

• Under the scenario of larger number of users(self-interference becomes negligible), FEC ismore important than spreading

2005 June 20th University Qualifying Exam 16

Page 17: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 17/44

Talk Outline Introduction

Computation of Packet Error Rate

Tradeoff between Channel Coding and

Spreading

4. Source Encoding with Optimal Mode

Selection

5. Optimal Bandwidth Allocation6. Simulation Results

7. Future Work

2005 June 20th University Qualifying Exam 17

Page 18: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 18/44

Page 19: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 19/44

Source Encoding: Intra-Coding Mode

CurrentFrame

PreviousFrame

INTRA Coding

Macro

Block 

2005 June 20th University Qualifying Exam 19

Page 20: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 20/44

Source Encoding: Inter vs. Intra Coding

2005 June 20th University Qualifying Exam 20

• Inter-Mode coding:

 – High compression efficiency –  Propagates past errors

 – More suitable for good channel conditions

• Intra-Mode coding: –  Stops error propagation

 – Costly in bits in general

 – More suitable for bad channel conditions

• It is desired to switch between two modes according

to channel conditions and video content

Page 21: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 21/44

Fixed-length Packet• Fixed-length packet

• Re-sync per packet• A MB is reconstructable at the decoder if:

– All packets containing this MB are received

• Illustration:

2005 June 20th University Qualifying Exam 21

Page 22: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 22/44

Page 23: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 23/44

Rate-Distortion Optimization• Every MB is tentatively encoded using both

intra/inter mode and 31 quantization steps• For each possibility, distortion ( D MB) and rate

usage ( R MB), thus the chip used for the MB (W  MB),

are calculated

• To minimize D MB under the constrained of frame

rate target. This is given by the Lagrangian

2005 June 20th University Qualifying Exam 23

Page 24: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 24/44

Source Encoding: Algorithm Summary• Computation of expected distortion at the encoder

is recursive on a per-pixel basis

• Uses the estimated distortion within a rate-

distortion framework for optimal mode-switching

• Optimal mode selection algorithm

 – Given the packet error rate – Given the target frame rate

2005 June 20th University Qualifying Exam 24

Page 25: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 25/44

Talk Outline Introduction

Computation of Packet Error Rate

Tradeoff between Channel Coding and

Spreading

Source Encoding with Optimal ModeSelection

5. Optimal Bandwidth Allocation6. Simulation Results

7. Future Work

2005 June 20th University Qualifying Exam 25

Page 26: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 26/44

Optimal Bandwidth Allocation• Problem:

Allocation bandwidth among Rs, r c and M , tooptimize the overall performance, under:

– Bandwidth constraint

– Frame rate target (transmission time)

• Effects:

– Changing channel characteristics

– Different source content

2005 June 20th University Qualifying Exam 26

Page 27: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 27/44

Optimal Bandwidth AllocationAlgorithm:

• Predetermine a set of target packet error rates

• For each packet: 2-step tradeoff (1) For each target packet error rate

Trade off channel coding and spreading, to maximize

the source bit rate ( Rs) and achieve the target packet errorrate (P p).

If no (r c, M ) pair can achieve a certain P p, skip this P p.

If none of the packet error rates can be achieved, reportdeep fading and temporarily send nothing (except pilotbits for channel estimation).

Result: a set of (P p , Rs , r c , M ) 4-tuples

2005 June 20th University Qualifying Exam 27

Page 28: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 28/44

Optimal Bandwidth AllocationAlgorithm (cont.):

• For each packet: 2-step tradeoff 

(2) Trade off source coding rate and packet error rate

For each (P p , Rs , r c , M ) 4-tuple, using the video

encoder encodes the current content ( MB by MB),

until the encoded bits fill the fixed length packet.

Choose the tuple that optimizes the performance:

minimizing the expected distortion per time unit.

2005 June 20th University Qualifying Exam 28

Page 29: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 29/44

Optimal Bandwidth AllocationAlgorithm (cont.):

• Illustration of a typical packetization after source encoding

• Select the 4-tuple that minimizes the distortion per time

unit, which is approximately determined by

2005 June 20th University Qualifying Exam 29

Page 30: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 30/44

Optimal Bandwidth Allocation• Computation complexity:

Let N 1 denote the number of candidate packet error rate, N 2 the number of  MBs per packet, the encoder needs to

tentatively encode ( N 1x N 2 ) MBs for each packet

• Choices of P p candidates:

 –  Too large (P p >5%): error rate goes up quickly to unity

 –  Too small (P p

<0.1%): under the same Rs

, performance

gain diminishes dramatically

– Our Choices of  P p : [0.2%, 0.6%, 1%, 1.5%, 3%]

2005 June 20th University Qualifying Exam 30

Page 31: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 31/44

Optimal Bandwidth AllocationSummary:

Optimal bandwidth allocation among source coding,channel coding and spreading for video

communications over tandem channels

 – Adaptively at the packet level

 – Incorporates the effect of changing channel

conditions– Incorporates the effect of current video content

2005 June 20th University Qualifying Exam 31

Page 32: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 32/44

Talk Outline Introduction

Computation of Packet Error Rate

Tradeoff between Channel Coding and

Spreading

Source Encoding with Optimal ModeSelection

Optimal Bandwidth Allocation

6. Simulation Results

7. Future Work

2005 June 20th University Qualifying Exam 32

Page 33: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 33/44

Page 34: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 34/44

2005 June 20th University Qualifying Exam 34

Simulation Result: PSNR vs. Chip RatePSNR performance versus target transmission chip rate,

Carphone QCIF, and E b /N 0 =4dB.

X axis: Chip Rate W 

(Mcps)

Y axis: PSNR (dB)Optimal bandwidth

allocation

Allocation with fixed

target P p=1%

Allocation with

P p=1% and a fixed M =15

Page 35: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 35/44

2005 June 20th University Qualifying Exam 35

Simulation Result: PSNR vs. Eb /N0PSNR performance versus Eb /N0, Carphone QCIF, chip rate

15Mcps.

X axis: Eb /N0 (dB)

Y axis: PSNR (dB)

Optimal bandwidth

allocation

Allocation with fixed

target P p=1%

Allocation with

P p=1% and a fixed M =15

Page 36: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 36/44

2005 June 20th University Qualifying Exam 36

Simulation Result: Packet Error Rate UsagePercentage of the total packets that employ the corresponding

target drop rate as optimal bandwidth allocation, chip rate 15

Mcps, and E b /N 0 =4dB.

X axis: Target Packet

Error Rate Pp (%)Y axis: Percentage (%)

Carphone QCIF

(high motion)

Mother and

daughter QCIF (low

motion)

Page 37: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 37/44

Talk Outline Introduction

Computation of Packet Error Rate

Tradeoff between Channel Coding and

Spreading

Source Encoding with Optimal ModeSelection

Optimal Bandwidth Allocation

Simulation Results

7. Future Work

2005 June 20th University Qualifying Exam 37

Page 38: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 38/44

Future Work • Channel Estimation, and System Sensitivity to

Estimation Error• Video Communications over Multi-Carrier (MC)

CDMA

– Spreading data stream in the freq domain over multiple

subcarriers

– Subcarriers convey the same information– Better freq diversity to combat selective fading

comparing to signal carrier

2005 June 20th University Qualifying Exam 38

Page 39: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 39/44

Acknowledgements

2005 June 20th University Qualifying Exam 39

Page 40: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 40/44

Appendix: Weight Distribution• Punctured Convolutional Codes:

– Mother convolutional code

– Puncturing table

• Rate-Compatible Punctured Convolutional (RCPC) Codes:

– The higher rate code is embedded in the lower rate codes

2005 June 20th University Qualifying Exam 40

Page 41: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 41/44

Appendix: Weight Distribution (Cont.)• Transition Matrix of Convolutional Codes (C. C.)

Example:

• Transition Matrix Sequence of Punctured C. C.Example: for puncture table a(1):

where

2005 June 20th University Qualifying Exam 41

Page 42: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 42/44

Appendix: Weight Distribution (Cont.)• Product Matrix

Φ=– y is the fixed packet length

• Weight Distribution for Block Codes

– Coming from the corresponding elements of Φ

– Example: Zero-tail method, T(x) is given byΦ1,1

∏=

 y

ii A

1

2005 June 20th University Qualifying Exam 42

Page 43: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 43/44

Appendix: Distortion per Time UnitDenominator: transmission

time of the packet (exactly)

Numerator: distortion of the

packet (approximate due to

the tails)– DCT transform

– Negligible, because tails are

much smaller than y

s

c

 R

 y

 M r 

 y

=

2005 June 20th University Qualifying Exam 43

Page 44: Packet Error Rate

7/31/2019 Packet Error Rate

http://slidepdf.com/reader/full/packet-error-rate 44/44

Appendix: System with Fixed Spreading GainSpreading gain ( M ) is fixed for all packets

Trade off between source coding and channel codingAlgorithm:

• Pre-calculate the relationship between packet error and

received SNIR, and solve the SNIR threshold under the

target packet error rate, for each RCPC code

• Under a target packet error rate (say 1%), use the RCPC

code with highest coding rate (thus maximize Rs) among

those achieve the target error rate.

2005 June 20th University Qualifying Exam 44