P1258764311cLZyB

1
Video transmission schemes using Multiple Input
Multiple Output techniques with Multiple
Description Coding
28th January 2005
Yasushi Takatori, Frank Fitzek, and Ramjee Prasad
2
Outline

• Background and Problems
• Introduction of key techniques
• MIMO
• MDC
• Proposed method
• Performances in i.i.d. channel
• Conclusion and future works

3
Next generation wireless systems
The spectrum efficiency improvement in the
physical layer will be saturated. We need to
anticipate the services of the next generation
wireless systems especially for physical layer
development.
4
Dominant services
Target services
Since very short delay is required, it is
difficult to apply strong FEC for this service.
Voice
Video services
Game
Requirement for short delay
File transfer
Text data
Very strong FEC, e.g. Turbo Codes, can be applied.
Required data rate
Problem How to support video services with
limited frequency resources?
5
New wireless system design approach
Conventional physical layer design
Minimizing the outage probability
Requirement for all services
Required PER, delay
Required SNR
Since the target service is determined, design
method becomes,
Service requirement Video quality
Minimizing the outage probability
MIMO
MDC
Proposed method
6
Outline

• Background and Problems
• Introduction of key techniques
• MIMO
• MDC
• Proposed method
• Performances in i.i.d. channel
• Conclusion and future works

7
Introduction to MIMO
What is MIMO systems? MIMO transmission
without Channel State Information (CSI)
Spatial Multiplexing Space Time
Transmission Diversity MIMO transmission with
CSI

8
MIMO Systems
Effects
MIMO transmission techniques
Space Time Coding Spatial division multiplex
Transmission diversity Beamforming Antenna
selection
Spatial multiplexing High data
rate Diversity gain Array gain (Spatial
equalization) (Interference suppression)
High spectrum efficiency
All effects cannot be achieved at the same time.
Smart design method is needed for each system.
9
Scenario 1 CSI is unknown at a transmitter
Spatial multiplexing effect
Diversity effect Scenario
2 CSI is known at a transmitter
10
MIMO systems without CSI at a transmitter
Rx
Tx
MIMO transmission techniques
Effects
Space Time Coding Signal Division Multiplex
Transmission Diversity
Spatial multiplexing High data
rate Diversity
11
Scenario 1 CSI is unknown at a transmitter
Spatial multiplexing effect
Diversity effect Scenario
2 CSI is known at a transmitter
12
Spatial Multiplexing Effect
12dB
BER
SNR dB
2 bits/symbol
4 bits/symbol
8 bits/symbol
QPSK
QPSK
16QAM
16QAM
256QAM
256QAM
Tx
Rx
To double the data rate
6db is required !
12db is required !
If existing system uses higher order modulation,
unacceptable huge transmission power is required
to double the data rate.
13
Spatial Multiplexing Effect
Coherent phase
QPSK
Dfp

w1
Tx
No signal
No signal
-
w2
Tx
To double the data rate, required transmission
power does not depend on the modulation order.
14
Scenario 1 CSI is unknown at a transmitter
Spatial multiplexing effect
Diversity effect Scenario
2 CSI is known at a transmitter
15
Diversity Effect- Space Time Block Coding
(STBC)-
h11
S1
S2
S1
S2
Tx
T

We can easily decode the data.
S1
S2
S1 S2
-S1 S2
h21
-
S2
-S1

w
Tx
2 S2
2 S1
By doing this procedure, diversity effect can be
expected even for transmitters. Additional
diversity effect can be expected by increasing
antennas at the receiver.
16
Which method performs better?
2 X 2 MIMO
BER
STTD, 2X Mbps
SISO, X Mbps
SDM, 2X Mbps
STTD, X Mbps
SNR
The crossing point depends on the following many
parameters. With higher modulation order, the
cross point largely shifts to right. Number
of antenna branches at transmitter / receiver
Fading condition Receiver type ZF, MMSE,
SUC, MLD, etc.
17
Scenario 1 CSI is unknown at a transmitter
Spatial multiplexing effect
Diversity effect Scenario
2 CSI is known at a transmitter
18
MIMO systems with perfect CSI at a transmitter
Rx
Tx
Estimated CSI
MIMO transmission techniques
Effects
Beams are generated in both sides.
Probe signals are transmitted from each
antenna. CSI is estimated at the receiver.
Estimated CSI is fedback to the transmitter.
Signal distribution in space Beamforming
Spatial multiplexing High data rate Array
gain
Complicated procedure is required while it
achieves the best performances.
19
Summary

1. Suitable MIMO effects should be selected for your
   system. Then, we need to find best combination
   method to achieve the selected effects.
2. MIMO techniques without CSI at transmitters is
   very simple transmission scheme comparing to
   perfect CSI scenario.
3. Spatial multiplexing approach and transmission
   diversity approach have different
   characteristics. It is very difficult to select
   the suitable approach.

20
Outline

• Background and Problems
• Introduction of key techniques
• MIMO
• MDC
• Proposed method
• Performances in i.i.d. channel
• Conclusion and future works

21
Introduction to MDC
22
Attractive merits of MDC
300 K bytes
Path-1
1 M bytes
300 K bytes
Path-2
MDC decoder
MDC encoder
Video Source
Decoded Video
300 K bytes
Path-3
3 Descriptions, Overhead20
By increasing the number of successfully received
descriptions, higher quality is achieved.
It has very strong diversity effect as well as
spatial multiplexing effect.
23
One example of MDC
This example guarantees high picture quality,
while the frame rate can not be guaranteed.
24
Overhead of MDC
Single Description Coding (SDC)
I
P
P
P
I
P
P
P
P
Only the differences are transmitted.
MDC
Description A
I
P
P
I
P
Description B
I
P
P
I
P
I frame period becomes double. Therefore, the
total number of I frames is equal to SDC.
The difference gets larger as the number of
descriptions increases. This results in the
overhead of MDC.
25
One example of MDC overhead
The overhead depends on the video sequences and
video coding methods. In this example, 5 -20
overhead is observed for two descriptions.
26
Performance of FEC in IEEE802.11a systems
18Mbps / 12Mbps 1.5 (Overhead50.0)
54Mbps / 48Mbps 1.125 (Overhead12.5)
M.H. Manshaei, et. al,A Media-Oriented
Transmission Mode Selection in 802.11Wireless
LANs, WCNC 2004, pp.1228-1233.
12.5 redundancy for FEC
Less than 0.5dB SNR improvement is expected.
50.0 redundancy for FEC
Around 1.0dB SNR improvement is expected.
For two descriptions scenario, MDC overhead
becomes 5-20. SDC achieves 0.5-1.0dB improvement
by employing stronger FEC instead of MDC.
27
Summary

1. In MDC, higher quality is achieved by increasing
   the number of successfully received descriptions.
2. It works like a combination of Spatial Division
   Multiplexing (SDM) and Space-Time-Transmission-Div
   ersity (STTD).
3. Only 0.5-1.0dB SNR improvement is expected in
   AWGN channel by employing the strong FEC instead
   of MDC.

28
Outline

• Background and Problems
• Introduction of key techniques
• MIMO
• MDC
• Proposed method
• Performances in i.i.d. channel
• Conclusion and future works

29
Basic concept
Transmission signals
Received signals
MIMO generates multiple spatial channels.
with MDC
Proposed
Transmission quality
Video quality
Conventional (SDM)
time
Conventional (STTD)
time
STTD or STC
MIMO-MDC achieves both effects simultaneously.
Video quality
SDM
time
30
Comparison of the video transmission scheme
Conventional video data transmission scheme
Wireless access
Strong FEC is applied here.
Video data source
MIMO-channel generator
Serial-parallel converter
SDM decoder
Video decoder
Video encoder
Internet
APs
WT
MIMO-MDC video data transmission scheme
Wireless access
Each description should be identified here.
Video data source
MIMO-channel generator
SDM decoder
MDC decoder
Access Controller
MDC encoder
Internet
FEC
APs
WT
31
Operation for the channel state variances in a
conventional scheme
MIMO-channel generator
Serial-parallel converter
SDM decoder
Video decoder
Video encoder
Internet
APs
WT
Feedback channel
Decrease the data rate!
MIMO-channel generator
Serial-parallel converter
SDM decoder
Video decoder
Video encoder
Internet
APs
WT
32
Operation for the channel state variances in
proposed MIMO-MDC scheme
MIMO-channel generator
SDM decoder
MDC decoder
Access Controller
MDC encoder
Internet
APs
WT
MIMO-channel generator
SDM decoder
MDC decoder
Access Controller
MDC encoder
Internet
APs
WT
WT discards the erroneous descriptions. Thus, It
does not require feedback procedures to the Video
data sources.
33
Summary

• Combination of MIMO with MDC achieves spatial
  multiplexing effect and transmission diversity
  effect simultaneously.
• The discarding description scheme is introduced
  at WT.
• Adaptive video quality change is possible without
  feedback.

34
Outline

• Background and Problems
• Introduction of key techniques
• MIMO
• MDC
• Proposed method
• Performances in i.i.d. channel
• Conclusion and future works

35
Simulation method
36
Picture Signal to Noise Ratio (PSNR)
Original
Decoded picture
Calculate the Mean Squared Error E
37
Simulation method for STTD-SDC
Binary signal generation
Convolutional coding
Inter leaver
STTD Encoder
Gray coding
Mod.
Array antennas
Spatial channel -1
a
b
a
b
Quasi static Rayleigh fading
b
-a
Spatial channel -2
Transmission symbols are distributed to spatial
channels.
Array antennas
BER calculation
Viterbi- decoding
Deinter leaver
STTD Decoder
Gray decoding
Demod.
PSNR evaluation
38
Simulation method for SDM-SDC
Binary signal generation
Convolutional coding
Inter leaver
Signal Distributor
Gray coding
Mod.
Array antennas
Spatial channel -1
a
c
a
b
c
d
Quasi static Rayleigh fading
b
d
Spatial channel -2
Transmission symbols are distributed to spatial
channels.
Array antennas
BER calculation
Viterbi- decoding
Deinter leaver
Signal Combiner
Gray decoding
Demod.
PSNR evaluation
39
Simulation method for proposed systems
Binary signal generation
Signal Distributor
Gray coding
Convolutional coding
Inter leaver
Inter leaver
Mod.
Array antennas
Spatial channel -1
D1
MDC encoder
Quasi static Rayleigh fading
D2
Spatial channel -2
Transmission descriptions are distributed to
spatial channels.
Array antennas
BER calculation
Gray decoding
Viterbi- decoding
Deinter leaver
Signal Combiner
Deinter leaver
Demod.
PSNR evaluation
40
Simulation results
41
Effect of discarding scheme
15dB
EBERlt10-5
D1
D1
5dB
EBERgt10-5
D2
Descriptions reception
Measure SNR
Expected BER
Discarding
10-4
10-4
10-5
10-5
BER for description-2
BER for description-2
PSNRdB
10-6
10-6
10-7
10-7
10-7
10-6
10-5
10-4
10-7
10-6
10-5
10-4
BER for description-1
BER for description-1
By introducing the discarding scheme at the
receiver, the PSNR performance is improved in
case of the imbalanced spatial channels.
42
PSNR comparison
MT2, MR2
symbol
PSNR improvement
STTD-H
SDM
More than 3dB
D1
Required SNR is reduced.
D2
Proposed
The proposed method achieves higher quality than
STTD-H. The proposed method reduces the required
SNR more than 3dB for that with SDM.
43
Conclusion

• For video services, we have proposed the
  combination of MIMO with MDC.
• At WT, we have proposed the discarding scheme.
• By our proposal, we can expect both the spatial
  multiplexing effect and the transmission
  diversity effect.
• The proposed method can automatically change the
  video quality without informing the compression
  rate change to the video server.
• Simulation results confirm that our proposed
  method outperforms conventional SDC-MIMO systems.

44
Other related works

• Theoretical formulation of MIMO MDC for two
  descriptions scenario has been derived.
• MIMO-MDC system design method has been developed.
• Dependency on the video sequences has been
  studied. And it is confirmed that the effect
  maintains for any video sequences.

45
Future works
46
One example of the multiple APs scenario
Multi-cast by MDC
Advertisement
Video streaming. Multi cast video streaming.
AP-1
MDC encoder
AP-2
Adaptive zone generation is used at both AP-1 and
AP-2.
Advertisement
SP conv. FEC
General packets
47
We also want to combine adaptive modulation
Once half of data are lost, it causes a
significant transmission quality degradation.
FEC
16QAM
QPSK
QPSK
Even if half of data are lost, adequate quality
can be expected.
MDC
APC
48
Future works

• Analysis of the combination of MIMO-MDC with
  Multiple access points, adaptive modulation,
  multicast, heterogeneous terminal, and
  heterogeneous APs.
• Development of the testbed for the MIMO-MDC.
• Theoretical analysis of MIMO-MDC for more than
  two descriptors.
• Combination of other Multiple Descriptions.
• Evaluation for other video quality measure.
• Adaptive overlapping zone control method.
• Imperfect CSI scenario.

49
Measurement of MIMO-MDC
D1
Video source
D2
Max. bandwidth 100MHz Center freq. 5GHz
Error pattern Selected mode
MIMO Transmitter
MIMO Receiver
CSI
D1
Decoded video sequence
D2
50
Publications

• Accepted/Published
• 1 Y. Takatori, K. Tsuneksawa, F.H.P. Fitzek and
  R. Prasad, On The Exploitation of Multiple
  Access Points in a Wireless SFN Using
  TDD-OFDM-MIMO Techniques, Wireless Personal
  Multimedia Communications WPMC 2004, page
  217-221. Padova, Italy.
• Submitted
• 1 Y. Takatori, F.H.P. Fitzek, K. Tsuneksawa and
  R. Prasad, Novel Video Streaming Transmission
  Methods for Multiple Input Multiple Output
  Techniques with Multiple Description Coding for
  4G Wireless Networks, JSAC, special issue on 4G.
  
• 2 Y. Takatori, F.H.P. Fitzek, K. Tsuneksawa and
  R. Prasad, Channel Capacity of TDD-OFDM-MIMO for
  Multiple Access Points in a Wireless
  Single-Frequency-Network, Kluwer Journal.
• In preparation
• 1 Y. Takatori, F.H.P. Fitzek, K. Tsuneksawa and
  R. Prasad, Video Streaming Transmission Methods
  for Multiple Input Multiple Output Techniques
  with Multiple Description Coding in Multiple
  Access Points systems, Electronics. Letter.
• 2 Y. Takatori, F.H.P. Fitzek, K. Tsuneksawa and
  R. Prasad, Multiple Input Multiple Output with
  Multiple Description Coding for Video Streaming,
  Mobi-com 2005.
• 3 Y. Takatori, F.H.P. Fitzek, K. Tsuneksawa and
  R. Prasad, New Video Multicast Method for
  heterogeneous terminals using MIMO-MDC, WPMC
  2005.