Coded Wireless Video
Broadcast / Multicast:
A Cross-layer Framework With Protections
To Harvest The True Potential of 4G Access Networks
James She, Ph.D.
Research Fellow, Computer Laboratory
Presentation @ The Chinese University of Hong Kong, Hong Kong – Jan 2011
1
Outline
1. Introduction & Background
2. A Preliminary Cross-layer Design
3. Coded Wireless Video Broadcast/Multicast
4. An Information-theoretical Bound of Expected
Distortion
5. Conclusion & Future Work
2
4G/Broadband Wireless Access and
1-to-many/all Video Applications
Wireless Broadcast/Multicast and Problems
In a single-hop wireless network:
• efficient use of spectrum
• higher system scalability
1. Multi-user channel diversity:
 Rate is limited to receiver with the
worst channel  low video quality
2. Error Control:
 Retransmission
 Not efficient or scalable
4
Research Objectives
Limitations of Existing Cross Layer Designs (CLDs):
1. Many for unicast are not applicable
2. Some for multicast/broadcast using erasure/network codings:
– Some statistical number of receivers within a multi/broadcast group
– Multi-hop wired/wireless infrastructure
Research Objectives:
1. Practical and generic cross-layer frameworks (advanced source + channel
coding) for single-hop network
2. Fundamental understanding of the proposed frameworks, using information
theory
3. Possible implementations
5
Outline
1. Introduction & Background
2. A Preliminary Cross-layer Design
3. Coded Wireless Video Broadcast/Multicast
4. An Information-theoretical Bound of Expected
Distortion
5. Conclusion & Future Work
6
A preliminary cross-layer design -
Superposition Coded Multicast
Scalable Video Source (MPEG4/
H.264AVC):
 Bitstream with successive refinable
layers
Layered Channel (Superposition coding):
 Multi-resolution modulated (layered)
broadcast signals
e.g., 2 layered video data (basic + enh. qualities)
layer 1 (BPSK)
layer 2 (QPSK)
layered
broadcast signal
Encoding: superimpose two modulated signals
(i.e., vector additions: x1 + x2  x )
…
… …
… …
Decoding, y:
1. decode the lower order signal (BPSK), y1,
from received signal, y
2. substract it from y for decoding y2.(QPSK)
(i.e., y - y1  y2.)
7
Superposition Coded Multicast (SCM)
Base station
2 quality layers
(base & enhancement)
Receiver(s)
Novelty: exploits the layered properties in scalable source and multi-resolution channel
BS only broadcasts/multicasts a single type of radio signals that contains all layers
decodable by receivers at various channel levels for multiple rate of video delivery.
Simulation Results
Compare achievable video qualities (PSNR):
• Normal Multicast vs. SCM
PS: ‘Normal’ uses the rate everyone supports
poor receiver (low SNR avg.)
good receiver (high SNR avg.)
Conclusion:
Higher video quality regardless of
the avg. channel SNR of a receiver!
9
SCM Summary
Two critical components identified:
1.
Scalable video (source)
2.
Multi-resolution modulation (channel)

Resolved multi-user channel diversity + video quality improvement.
An limitaiton:
• Video quality fluctuates with channel condition at
a receiver regardless of the SNR avg.
 Error control problem!
10
Outline
1. Introduction & Background
2. A Preliminary Cross-layer Design
3. Coded Wireless Video Broadcast/Multicast
4. An Information-theoretical Bound of Distortion
Bound
5. Conclusion & Future Works
11
Proposed - Coded Wireless Video Broadcast/Multicast
Deal with error control & multi-user channel diversity:
 Introduced protections to each successive layer at the source
 Achievable by modifying the Multiple Description Coding based on Reed Solomon
RS(N, K) [18]
Note:
For any layer l, a smaller Kl
value, the higher robustness
to tolerate fading duration
for that layer.
 Each MDC/protected packet (with multiple layers of bitstreams) is sent through SCM as
a multi-resolution modulated signal
[18] P. A. Chou, H. J. Wang, and V. N. Padmanabhan, “Layered multiple description coding,” Proc. PV 13th Int. Packet Video Workshop,
Nantes, France, Apr. 2003.
12
System Model And Error Control Advantages
App. layer
3
Video in
Scalable
Video
Encoder
MAC layer
1
PHY layer
Superposition Modulation
Scalabe bitstream
2
BPSK modulated
analog forms
layer 1 buffer
Multicast
Scalable Error Controls:
addressed
Coded Wireless
Video Broadcast/
+
Receiver recovers its own lost bitstreams of layer l when
successfully
superposed
Multicast signals
Multicast
analog forms
received any Kl “partial” MDClayerpackets
of
layer l.
2 buffer
addressed
16QAM modulated
analog forms
Wireless
Channel
Protection Coder
e.g. Reed-Solomon
4
5
layer 1 buffer
Demodulated
base layer
data
Superposition
Demodulation
+
layer 2 buffer
Video
Decoder
Demodulated
enhancement
layer data
Coded Wireless
Video Broadcast/
Multicast signals
superposed
analog forms
Video out
13
A Quick Video
(00:00:51-00:02:05)
Formulations For Analysis
Performance/video quality measurement:
Total received/recovered bitstreams, Tm, of a GoF by
a receiver m.
• Prob. of receiving/recovering a layer l by receiver m
(i.e., receiving at least Kl partial packets of layer l by
receiver m)


N
N

K
j


N

j
j
l (
P

l
o
s
s
(
S
N
R
)
)
(
1

l
o
s
s
(
S
N
R
)
)


l
,
m
l
,
m

l
,
m


j

0
•With the layers dependency, the amount of received/
recovered bitstreams of a layer l in a GoF:
layer l
l P
T

(
bb

)

i
,
m
l
,
m l l

1i

1
Total received/recovered bitstream of a GoF:
L
T

T,
m 
l
1 lm
15
Optimized & Experimental Results
Compare layered broadcast w/o protection (e.g., SCM) and the proposed one:
• 2 layers w/ optimized (searched) parameters
• 2 different standard video sequences (Foreman and Paris)
Note:
SS-1: lowest SNR avg.
SS-10: highest SNR avg.
SS-1
(Foreman)
Coded Wireless Video Broadcast/Multicast
• better video quality even in a poorer channel
• smaller quality difference between receivers with highest and lowest SNR avg.
16
poor receiver
Layered
source+channel
(SCM)
Layered
source+channel
with protection
good receiver
Summary
Novelties:
1. Introduced protections on successive layers over layered broadcast
2. Utilized partial MDC (protected) packets (never discussed in wired
infrastructure)
3. Modified existing MDC for practical implementation
4. An analytical model for analysis and optimization.
 Resolved both multi-user channel diversity and error control
problems which are not possible in all previous and recent works [1-4]
[1] Chris T. K. Ng et al., “Recursive Power Allocation in Gaussian Layered Broadcast Coding with Successive Refinement,”
IEEE Intl. Conf. on Comm. (ICC), Jun 24–27, 2007, Glasgow, Scotland, pp. 889–896.
[2] C. Tian et al., “Successive Refinement Via Broadcast: Optimizing Expected Distortion of a Gaussian Source Over a
Gaussian Fading Channel”, IEEE Trans. on Information Theory, vol. 54, no 7, pp.2903-2918, Jul. 2008.
[3] Y. S. Chan et al., “An End-to-End Embedded Approach for Multicast/Broadcast of Scalable Video over Multiuser
CDMA Wireless Networks”, IEEE Trans. on Multimedia, vol. 9, no. 3, pp. 655-667, Apr. 2007.
[4] Murali R. Chari et al., “FLO Physical Layer: An Overview”, IEEE Trans. on Broadcasting, vol. 53, no. 3, pp. 145-160,
Mar. 2007
18
Outline
1. Introduction & Background
2. A Preliminary Cross-layer Design
3. Coded Wireless Video Broadcast/Multicast
4. An Information-theoretical Bound of Expected
Distortion
5. Conclusion & Future Works
19
Informaton-theoretical Bound of Expected
Distortion
Did you realize that we send less video data? Costs of protections
If / when the proposed framework is better than a similar layered broadcast WITHOUT
protections?
 The expected distortion without layers [3] :
where each source symbol is sent by a channel symbol under symbol error, perr.
Assume a successive refinable source with Gaussian distribution (i.e., L layers in each
source symbol, S)
Apply a generic (n, kl) protection code in a layer l:
i.e. A source symbol, S  n protected layered source symbol, Vi , where i=1,..., n.
Recall: smaller kl value for layer l, more robustness, but less effective data sent
[3] X. Yu and En-hui Yang, “Optimal quantization for noisy channels with random index assignment", Proc. of the 2008 IEEE Intern. Symp. Inform. Theory,
Toronto, Canada, July 6-11, 2008.
20
Bound of Expected Distortion
protected
layered
source
symbols, Vi
x(i)
Each Vi  a layered channel symbol, x(i), for each coded broad/multicast transmission:
A receiver collects n channel symbols, x(1), ... ..., x(n) over n channel symbol durations.
Each x(i) is decoded into

Vi
~
up to layer l with prob.
pl
upon the receiver`s instantaneous
channel condition.
21
Bound of Expected Distortion
After n channel symbol durations (or n transmissions),
kL
kl
k1
V1
...
...
Vn
 The Bound of Expected Distortion:
l
k
1
l
i





n

k
R
j
n

j


j
j

j

1
2
1

1

p
p









M
,
i
M
,
i 

j
L
0
i

1





 j



D

p


l
k

1
l
i
l

1


n
j
n

j


 2



1

1

1

p
p










,
i
M
,
i
 M
 l

j
j

0
i

1









 Applicable to a system without protection (i.e., kl =n)
 The discreteness (i.e., binomial CDF terms) can be approximated by
a normal CDF to determine optimal k values for optimization.
22
Numerical Anylysis -1
Fixed symbol error at layer 2
(a) higher pM,1 (k1*=5, k2*=2)
(b) lower pM,1 (k1*=14, k2*=2)
Fixed symbol error at layer 1
(a)higher pM,2 (k1*=18, k2*=1)
(b) lower pM,2 (k1*=18, k2*=3)
23
Numerical Anlysis - 2
Expected distortions of two systems (with and without protections)
under various pM,1 in layer 1 and pM,2 in layer 2.
Simulation Comparisons
Layered broadcast without and with protections under optimized parameters:
Fixed lower, pM,2, in layer 2
Fixed higher, pM,2, in layer 2
Systems with their optimized configurations .
25
Summary of Expected Distortion
Novelties:
 A general closed-form formula for the bound of expected distortion
 Generic to any (n, k) protection code and any number of layers
(source/channel), useful for a new coding design
 More accurate analysis/optimization, instead of using simply using
throughput/bitstream amount.
Outline
1. Introduction & Background
2. A Preliminary Cross-layer Design
3. Coded Wireless Video Broadcast/Multicast
4. An Information-theoretical Bound of Expected
Distortion
5. Conclusion & Future Works
27
Contributions

1st framework using protections for tackling multi-user diversity and
error control.

1st realization through existing codings, as well as the associated
analytical and optimization models.

1st information-theoretical distortion bound for comparisons, and
optimization through a simple search.
Advanced the fields by introducing a new design dimension –
protections, for cross-layer designs that was unapparent in the
past literature.
LESS is MORE sometimes!
28
Future Work
 Extend into cooperative communications under multi-BSs wireless networks (e.g.,
optical-wireless hybrid network) by considering space-time coding
 Promising results from preliminary investigations in EPON-WiMAX access
networks
Final Remark: a cross-layer design with protections is shown to be useful in
cooperative networks for better video broadcast/multicast
29
On-gogin Research and Industrial Collaborations
• WiMAX/LTE BS system and
cooperative broadcasting networks
(prototype and research)
(Taiwan)
(Italy)
Electrical & Computer
Engineering
• SPC chipset and
software-defined
radio platform
(research)
• New scalable source coding
with protection (research)
(Saudi Arabia)
Electrical &
Computer
Engineering
(Ottawa)
• WiFi platform (prototype)
• Coded MIMO
Broadcast/Multicast
(research)
Electrical & Electronic
Engineering
Acknowledgement
Prof. Pin-Han Ho, University of Waterloo
Prof. En-hui Yang, University of Waterloo
Dr. Xiang Yu, Research-In-Motion
Sponsors:
IPMG
Collaborators:
31
Collaborations and students
1. Looking for like-minded researchers and
organizations/industries for collaborations and funs!
2. Looking for smart, creative and entrepreneurial students to
join me as my research interns.
Email: james.she@cl.cam.ac.uk
Web: http://www.cl.cam.ac.uk/~js864
32
The End
THANK YOU
33
Selected Publication From This Research
SCM:
[1] J. She, et al., “IPTV over WiMAX: Key Success Factors, Challenges and Solutions”, IEEE
Communications Magazine, vol. 45, no. 8, pp.87-93, Aug. 2007.
(Top 50 most accessed article in IEEE Xplore 2008, and cited in Wikipedia under MobileTV)
Coded Wireless Video Broadcast/Multicast:
[2] J. She, et al., “A Cross-Layer Design Framework for Robust IPTV Services over IEEE 802.16
Networks”, IEEE Journal of Selected Areas on Communications (JSAC), vol. 27, no. 2, Feb. 2009,
pp. 235-245.
[3] J. She, et al., “A Framework of Cross-Layer Superposition Coded Multicast for Robust IPTV
Services over WiMAX”, Proceedings of the IEEE Wireless Communication and Networking
Conference, pp. 3139-3144, Mar. 2008, Las Vegas, Nevada, USA. (Nominated for the Best Student
Paper Award)
Expected Distortion Comparison:
[4] J. She, et al., “Distortion Comparisons For Protected Successive Refined Over Broadcast Channel
”, submitted to Trans. Multimedia, Jul. 2010.
L-SPC:
[5] J. She, et al., “Logical Superposition Coded Modulation”, submitted to Trans. Wireless
Communication, Nov. 2010.
34