Multimedia over CDMA Mobile Wireless Networks: A Joint Source CodingPower Control Approach

1
Multimedia over CDMA Mobile Wireless Networks A
Joint Source Coding-Power Control Approach

• Yee Sin Chan
• Center for Image Processing Research
• and
• Electrical, Computer, and Systems Engineering
  Department
• Rensselaer Polytechnic Institute
• Troy, NY 12180

2
Overview

• Issues and Challenges
• Joint Source Coding-Power Control (JSCPC)
  Approach for Video Delivery
• Delivery of Single layer Video
• Prioritized Transport of Scalable video
• JSCPC with Adaptive Channel Coding
• Rate-Compatible Punctured Convolutional (RCPC)
  Codes
• Rate-Compatible Punctured Turbo (RCPT) Codes
• An End-to-End Embedded Approach for
  Multicast/Broadcast of Video
  
  
  
  

3
Multimedia over Wireless NeworksChallenges

• Highly Heterogeneous Traffic Multirate and
  Diverse QoS.
• Network Resources (BandwidthPower) Expensive and
  Limited.
• Mobile Wireless Channels Are Extremely
  Error-prone Due to Multipath Fading Effects.
• Highly Time-varying Channel Conditions.
• Interference/congestion, Latency and Delay
  Jitter.
• Questions How to Maximize
• Capacity (in Terms of Number of Users Supported),
• Network Utilization,
• End-user QoS.
• Focus on Video Traffic, One of the Most Important
  Applications in Emerging Multimedia Mobile
  Wireless Networks.

4
Real-Time Video Traffic
Difficult!!!!!

• Video Requires Extremely Large Bitrate
• Large Bandwidth Requirement for Video Sources vs.
  Limited Resources of Wireless Networks.
• High Compression Ratio Bitstreams
  Extremely Sensitive to Channel Errors and Network
  Impairments.
• Stringent Latency Requirements.
• Compressed Video Requires Error Protection
• Most Existing Video Compression Standards
  Originally Not Designed for Lossy Channels.
• Appropriate Error Protection Schemes an
  Important Research Topic.

5
Direct Sequence Spread-Spectrum Multi-user
Networks (CDMA)
Narrowband Message
S(t)
d(t)
Wideband Signal
c(t)
fc
fc
c(t)
Synchronized Wideband PN Sequence
Wideband PN Sequence
Up-Conversion to fixed Carrier Frequency
Down- Conversion
despreading
Spreading

• All Users Communicate Simultaneously on a Given
  Carrier.
• Other Users Signals are Considered as Noise
  Interference Limited Property
• Desirable to have a High Processing Gain to take
  Advantage of CDMA characteristics and Maintain a
  Good Correlation Properties Bandwidth-Limited
  Property

6
Issues and Approach

• Optimal System Performance
• Consider Both Interference- and Bandwidth-Limited
  nature.
• Simultaneous Maximize the Delivered Video Quality
  and the System Capacity.
• Breakdown of the Problems
• Consider the Interference-Limited Nature
• Extend to Bandwidth-Limited Nature

7
Evaluating the End-to-End Video Quality
Average Peak Signal-to-Noise Ratio (PSNR)

• Factors Affecting End-user QoS
• Source End Quantization Errors.
• Network Packet Loss and Bit Errors.
• Receiver End Error Concealment and
  Post-processing.

8
Testing Sequences
Susie a studio based videophone sequence, no
background motion
Carphone a mobile videophone sequence,
continuous camera motion and a structured
background
9
A Joint Source Coding-Power Control Approach
forVideo Delivery

• Jointly Optimize Source Coding and Power
  allocation
• Utilize Unequal Power Assignment/Unequal Error
  Protection (UPA/UEP)
• Incorporate Passive Error Recovery (PER) Scheme

10
System Model

• A Single-cell Reverse Link in Which All Users
  Communicate Asynchronously Through the Same
  Physical Channel (AWGN) Using coherent BPSK.
• Perfect Power Control Assumed.
• Consider Uncorrelated Bit Errors Achieved Through
  Sufficient Interleaving Binary Symmetric
  Channel Model.
• Constant Chip Rate for All Users, hence Multirate
  Implies Variable Processing Gain.
• Real-time Video User - FEC Only

11
Capacity Analysis for Heterogeneous Traffic

• Total Bandwidth, WT.
• Number of Users, N.
• Consider user i, at power level SiEiRi, where Ei
  is the energy per information bit and Ri is the
  information rate.
• Use the same approach as Gilhousen1 et al. to
  evaluate single-cell capacity.
• The energy per information bit to multiple-access
  interference (MAI) ratio of user i will then be

1K. S. Gilhousen et al., On the Capacity of a
cellular CDMA System, IEEE Trans. on Vehicular
Tech., Vol. 40, no. 2, pp. 303-312, May 1991.
12
Capacity for Heterogeneous Traffic (Cont.)

• Simplifying this expression yields
• Assuming a large number of users so that the MAI
  experienced by any user due to total traffic from
  all the other users is approximately equal,
• For large N, 1/N? 0,

The product , equivalent
bandwidth, represents the effective bandwidth
resources consumed by user i and is to be
optimally allocated subject to the above
constraint.
13
Real-Time Video Traffic

• The For a given Weq, how can we tradeoff Rs and
  Eb/N0 effectively so as to maximize the
  end-to-end performance defined by average Peak
  Signal to Noise Ratio (PSNR) ?
• That is we want to solve
• subject to the constraint Weq RsEb/No
• For homogeneous users ( i.e., fixed Rs) the
  problem is reduced to finding the optimum data
  rate as Eb/N0 is the ratio of the processing
  gain to the number of users.

14
Proposed CDMA Wireless System Supporting Video
Traffic
Source Coding
Video Source
Modulation
Rate-Matching Spreading
Channel Coding
Adjust Rs
Adjust Power Level Eb/N0
Estimate the Resource available for an
individual user
CDMA Networks
Centralized Admission Control Resource Scheduler
User Requirements
Monitor the Total Power Level
Despreading
Demodulation
Channel Decoding
Source Decoding
Video Display

• A centralized admission control/resource
  scheduler unit allocates the resource (Weq) for
  an individual user.
• Question How to jointly allocate Rs and Eb/N0 to
  minimize overall distortion for a fixed network
  commitment?

15
H.263 Basics
Video Sequence
P
P
I
P
I
P
Group of Block (GOB)
MB 1
MB n
MB 2
MB 3
GOB 1
GOB 2
2
1
Y1
Y2
8 lines
GOB n-1
Y3
Y4
Cr
Cb
GOB n
57
64
Macroblock (MB)
Picture Frame
8 pels
Block
16
Source Coding

• Use ITU-T H.263 as the video source coder
• low bitrate video compression standard
• supports a wide range of custom picture formats.
• Source coding rates adjusted through the choice
  of the corresponding quantization parameters
  (QPs).
• Each packet corresponds to one GOB.
• Incorporate passive error concealment scheme
  (TMN8 PER).
• Utilize universal distortion-rate characteristics
  PSNR(Rs,Pb) to evaluate end-to-end quality.

17
Typical Universal Distortion-Rate
Characteristics, PSNR(Rs, Pb)(Single-Layer
H.263 Coder)

• Can be used to compute overall distortion for a
  given source coding rate and channel conditions.
• Increasing Rs actually degrades performance for
  large Pb in the presence of channel induced
  errors.

18
Illustration of JSCPC for an Uncoded System
PSNR vs. Weq for fixed Source Coding Rate System
Optimized end-to-end PSNR vs. Weq
By jointly optimizing both source rate and power
level for a given Weq, we can avoid the
saturation effect associated with the use of a
fixed Rs while at the same time avoiding the
precipitous drop in PSNR as Weq decreases.
19
Illustration of JSCPC for a Coded System
End-to-end PSNR vs. equivalent bandwidth Weq for
a coded system Rc1/2 convolutional code with
constraint length K9 using an H.263
single-layer coder.
20
Illustration of JSCPC for a Coded System with
Fixed Channel Coding Rate Special
Case-Homogeneous Users(WT 20 MHz, Rc1/2, K9 )

• Notice the precipitous drop in performance with
  increasing number of users associated with a
  fixed source coding rate.
• JSCPC procedure extends the useful capacity of
  CDMA network while exhibiting a more graceful
  degradation pattern under increasing load.

21
Original
Typical Results
22
Prioritized Transport of Scalable Video

• Error-Resilient Scalable Source Coding
• Jointly Optimize Source Coding Rate and Power
  Assignment (JSCPC) Across Layers of Different
  Priority Class

23
Scalable Video Coding

• Scalable coding produces multiple bitstreams
• a high-priority base-layer (BL)
• a number of low-priority Enhancement-Layers
  (EL).
• BL carries the most important information and can
  be used to generate video with a minimum
  base-level quality.
• BL should be transmitted with highest priority
  level.
• Correctly received EL can provide refined Video
  Quality.
• Unequal Power Assignment (UPA)/UEP across
  different layers.
• Concentrate on a 2-layer H.263 SNR scalable mode.

24
Proposed JSCPC Approach for Delivery of Scalable
Digital Video
Source Coding
Video Source
Modulation
Rate-Matching Spreading
Channel Coding
Adjust (Rs(1),Rs(2))
Adjust Power Level (Eb(1)/N0, Eb(2)/N0)
Estimate the Resource available for an
individual user
CDMA Networks
Centralized Admission Control Resource Scheduler
User Requirements
Monitor the Total Power Level

• 2-layer video encoder producing 2 priority
  bitstreams with rates, Rs(Rs(1),Rs(2)) and the
  corresponding bit energy to MAI ratios,
  Eb/N0(Eb(1)/N0,Eb(2)/N0), it follows that the
  total bandwidth requirement of a scalable video
  user is,
• Question How to maximize the overall performance
  by allocating both the bit-rate and the power
  level within and between bitstreams of different
  importance of a scalable video coder subjected to
  the allocated resources (Weq) ?

25
Typical Universal Distortion-Rate Characteristics
2-layer SNR scalable codec
Single-layer codec
Quality, measured in terms of PSNR, is of
different sensitivity to bit-errors in the BL and
EL.
Increasing Rs actually degrades performance for
large Pb
26
Unequal Power Assignment (UPA) (2-Layer Coded
System, Rc1/2, Rs 35 Kbps)

• In the presence of channel-induced impairments,
  2-layer coding and UPA/UEP approach shows
  substantial performance advantages over EPA/EEP,
  especially for small values Weq.
• For Large Weq, 2-layer coding suffers performance
  penalty compared to single-layer system due to
  the additional overheads associated with scalable
  coding.
• Optimized UPA/UEP system still suffers from a
  precipitous drop in performance for small Weq and
  exhibits a saturation effect for large Weq.

27
Illustration of JSCPC for a Coded System
Optimized end-to-end PSNR vs. equivalent
bandwidth Weq for a coded system Rc1/2
convolutional code with constraint length K9
using a 2-layer H.263 codec.
By jointly optimizing both the source coding
rates and the power level for a given Weq, it is
possible to avoid the saturation effects
associated with the use of a fixed Rs while at
the same time avoiding the precipitous
degradation in PSNR as Weq decreases.
28
Joint Source Coding-Power Control (JSCPC) with
Adaptive Channel Coding

• Jointly Optimize Source and Channel Coding Rate
  (JSCC) employing Rate-Compatible Codes
• Jointly Optimize Source Coding Rate and Power
  Assignment (JSCPC)
• Consider a Combined JSCPC/JSCC Approach

29
Why JSCPC JSCC?

• Support Diverse Multimedia Services
• Limited and Fixed Channel Transmission Rate

• System has to Support a Wide Span of Coding Rates
• Unequal Error Protection (UEP)
• Improving Spectral Efficiency

Multirate Diverse QoS
JSCC
CDMA Spread-Spectrum System

• Unequal Error Protection (UEP)
• Improving Power and Spectral Efficiency

Interference Limited Diverse QoS
JSCPC
30
Proposed CDMA Wireless System

• PSNR(Weq, RT) max PSNR(Rs, Eb/N0,Rc)
• Constriants Rs Eb/N0 Weq
• Rsc Rs/Rc ? RT

31
Adaptive Channel Coding RCPC codes

• Existing System Fixed Rate Convolutional Codes
  Symbol Repetition/Puncturing
• Rate-compatible Punctured Convolutional (RCPC)
  Codes
• Provides a Wide Span of Coding Rate for Multirate
  Multimedia Services in CDMA Networks.
• Provides Unequal and Adaptive Error Protection
  Capability for Different Multimedia Services.
• Same Viterbi Decoder Can Be Used for All
  Punctured Codes From a Given Mother Code.

Input sequence
An RCPC encoder of Rc4/7 code derived by
puncturing a mother code of rate Rc1/2.
32
Performance of RCPC Codes over AWGN Channel
RCPC codes Rc1/4, K9, p8

• A wide span of available coding rates.
• Codes with different error correcting
  capabilities to match the prevailing channel
  conditions for a specified level bit error
  probability.

33
Performance of a Coded System with Total
Transmission Rate 576 Ksps
Solid RCPC codes Rc1/4, K9 and p8 Dotted
Convolutional codes with symbol
repetition/puncturing Rc1/2, K9

• Using JSCPC in tandem with JSCC and RCPC codes
  provides substantial improvement in end-user QoS
  and resource utilization.
• Results in improved overall capacity.

34
JSCC/JSCPC using Turbo Codes

• Turbo Codes
• Introduced by Berrou et al. in 1993.
• A Novel Combination of Iterative Decoding With
  Soft-input/soft Output (SISO) Algorithms and
  Recursive Systematic Convolutional (RSC) Codes in
  Parallel Connected by an Interleaver.
• Achieve Performance Near Shannon Limit on AWGN
  Channels.
• Extensive Research in Progress Including
  Interleaver Design, Decoding Algorithms and
  Applications.
• Turbo Codes have the potential for significant
  improvement in overall system capacity when using
  JSCPC/JSCC Approach.

35
Structure of Rate-Compatible Punctured Turbo
(RCPT) Encoder

• Formulation follows RCPC codes almost exactly.
• Higher-Rate Codes Obtained by Puncturing an
  Rate-1/n Mother Code.
• Decoding Through the Use of a Bank of
  Soft-Input/Soft-Output (SISO) Decoders and the
  Associated Iterative Decoding Structure

36
Performance of Short-Length (L1024) Turbo Codes
over AWGN Channel
RCPT Rc1/3, K 5, g(1,23/35), Block Length
1024, Number of Iteration 12
RCPT provides Outperforms both fixed rate
turbo/convolutional codes with symbol
repetition/puncturing
37
Performance Analysis for JSCC/JSCPC Employing
RCPT codes with RT288 Kbps

• JSCC Schemes Using RCPT Codes Provide Substantial
  Improvement Over Systems Using Convolutional
  Codes for Similar Complexity
• Results in Improved Overall System Capacity.

38
An End-to-End Embedded Transmission Scheme for
Multicast/Broadcast of Digital Video

• Using Scalable Video Encoder
• Consider a Combined JSCPC/JSCC Approach Employing
  Rate-Compatible Punctured Turbo Codes
• Employing Adaptive MPSK Modulation

39
An End-to-End Embedded Transmission Scheme for
Multicast/Broadcast of Digital Video

• Development of future generation cellular systems
  is targeted at building a general platform for
  various types of services.
• Traditional approach Performance constrained by
  the least capable of the intended receivers.
• Fails to take advantage of the fact that some of
  the intended recipients may be more capable than
  the others.
• Theoretical Investigation initialized by Cover2,
  suggested optimal broadcast scenarios could be
  achieved by a scalable/embedded approach.
• Required transmission techniques that can
  simultaneously deliver a basic QoS to each
  receiver and and additional end-user refinement
  for more capable receivers.

2T. Cover., Broadcast Channels, IEEE
Information Theory, Vol. IT-18, pp. 2-14, Jan
1972.
40
System Model
Layout of the System
Basic- and Enhanced-Quality Coverage Area

• Consider forward multicast/broadcast channel
• Only consider interference from 1st and 2nd tier
  cells (K19)
• Users within the same User Group (UG) share the
  same PN sequence.

41
SNR Distribution
Path Loss 3.0, 1/? 100, Antenna Height 5,
Cell Size 100
Path Loss 3.0, Antenna Height 5, Cell Size
100

• The received SNR depends on the geographical
  distance from the base station.
• Also Depends on the Power Allocation (?)

42
Block Diagram of the Embedded System
43
Adaptive Modulation based Upon Nonuniform MPSK
Nonuniform QPSK
Nonuniform 8-PSK

• Adaptive modulation in conjunction with
  multi-layer video source coder.
• Base-layer sent as basic message.
• Enhancement layer sent as additional message
• Consider QPSK due to its universal deployment.

44
Performance of the System Using Uniform
Modulation Scheme

• The Proposed System Using Uniform Modulation
  Scheme is Unable to Take Advantage of More
  Capable Receivers of the Subscribers Closer to
  the Base Station.

45
Performance Analysis for the Embedded System

• Simultaneously deliver basic video-quality for
  less capable receivers and enhanced video-quality
  for more capable receivers with proper power
  allocation.

46
Tradeoff Issues of the Embedded Transmission
Schemes
Tradeoff of Coverage Area with Allocated Power
Tradeoff of Enhanced Video-Quality with Allocated
Power
47
Summary

• Developed a JSCPC Approach for Multimedia
  Transmission Over Spread-Spectrum CDMA Networks.
• Extended the Approach to Prioritized Delivery of
  Scalable Video.
• Described a Methodology for Combining JSCPC and
  JSCC Using RCPC/RCPT Codes for Channel Adaptation
  and Unequal Error Protection.
• Proposed an End-to-End Transmission Scheme for
  multicast/broadcast of Digital Video Over CDMA
  Networks.