1. Kyoungwoo Lee, Minyoung Kim, Nikil Dutt, and Nalini Venkatasubramanian
Error-Exploiting Video Encoder to Extend Energy/QoS Tradeoffs for Mobile Embedded Systems
Kyoungwoo Lee, Minyoung Kim, Nikil Dutt, and Nalini Venkatasubramanian
Department of Computer Science
University of California at Irvine

2. DIPES ’08 Strategy 25 mins talk and 5 min QnA Very broad audience
10 to 15 mins for motivation and problem background
High level presentation
Technical depth is not that high, and deep technical highlights are in the backup slides

3. Outline Motivation and Problem Statement Our Solution Experiments
Conclusion

4. Energy Reduction is Essential
Battery-Operated Mobile Embedded Systems
Energy reduction is essential in battery-operated mobile embedded systems
Mobile video applications demand high energy consumption
Complex video encoding algorithms incur high overheads in terms of performance and power
Network
Mobile Video Applications

5. Active Error Exploitation
Network f1 f4 f3 f2
Active Error Exploitation – Intentional Frame Dropping
Skip the expensive video encoding algorithm  Energy saving
Degrade the video quality
Inherent error-tolerance mitigates the impact of frame drops on video quality

6. Frame Drop Types
Intentional Frame Drop (one way to actively exploit errors) can result in energy reduction for each operation
FDT-1 affects the following components with respect to power, performance, and QoS in mobile video applications
Error-prone Networks
Mobile Video Application
Enc Tx Rx
Dec
CPU
WNI
WNI
CPU
FDT-1
FDT-2
FDT-3
FDT: Frame Drop Type
Enc: Encoding, Dec: Decoding
WNI: Wireless Network Interface
Packet Loss

7. Inherent Error-Tolerance of Video Data
Spatial and temporal correlation among consecutive video frames
Lossy video encoding
(e.g.) High Quantization Scale
Energy Reduction and Error-Tolerance
Error-tolerance can be used for energy reduction
(e.g.) partial ME vs. Full ME
One frame loss may not be noticed by users
(e.g.) One frame loss out of 30 frames per second
Mobile Video Encoding ME
DCT Q EE
ME – Motion Estimation
DCT – Discrete Cosine Transform
Q – Quantization
EE – Entropy Encoding

8. Frame Losses due to Packet Losses
Mobile Video Applications f1 f4 f3 f2 f3 f2 f1
Error-Prone
Network
f4 is lost
Error-Induced Video Data
Inherent Error-Tolerance of Video Data
Error-Resilient/Error-Concealment Techniques

9. Error-Induced Video Data
Error-Resilience f1 f4 f3 f2 f3 f2 f1
Error-Prone
Network
f4 is lost
Error-Induced Video Data
Error-Resilient Techniques
Insert Intra-frames (I-frames) periodically
(e.g.) GOP-10 inserts I-frame every 10 P-frames
Intra Refresh video encoding techniques
(e.g.) PBPAIR (Probability Based Power Aware Intra Refresh) encodes video data resilient against 25% frame loss rate (1 frame out of 4 frames)

10. Error-Induced Video Data
Energy Efficiency f1 f4 f3 f2 f3 f2 f1
Error-Prone
Network
f4 is lost
Error-Induced Video Data
Energy-efficient error-resilient video encodings
(e.g.) PBPAIR or Probability-Based Power Aware Intra Refresh [Kim, MCCR06]
It may improve not only the video quality but also energy consumption

11. Outline Motivation and Problem Statement Our Solution Experiments
Error-Exploiting Video Encoding
EE-PBPAIR
Experiments
Conclusion

12. Our Proposal Error-exploiting video encoder
Intentional frame dropping + error-resilient video encoding
Extends the tradeoff space for energy consumption / QoS

13. Error-Resilient Video Encoder
Data
Original
Video
Data
Error-Resilient
Encoder
Parameters

14. Error-Exploiting Video Encoder
Injected
Video
Data
Error-
Aware
Video
Data
Original
Video
Data
Error-Injecting Unit
Error-Canceling Unit
Error
Controller
Error-Resilient
Encoder
Constraints
Parameters
& Feedback

15. Intentional Frame Dropping and PBPAIR
Error-Induced Video Data f1 f4 f3
f4 is lost
Error-Prone
Network
Intentional
Frame Dropping
PBPAIR
EE-PBPAIR
Error-Injecting Unit
Error-Canceling Unit
f2 is dropped
Energy Efficiency
Frame Dropping
(e.g.) f2 is dropped
PBPAIR
Quality Management
Error-Resilience
(e.g.) EE-PBPAIR encodes video data resilient against f2 and f4
Error-Tolerance

16. EIR – Error Injection Rate
EE-PBPAIR
Error-
Aware
Video
Data
Original
Video
Data
Error-Injecting Unit
Error-Canceling Unit
Frame Dropping
PBPAIR
Quality
Constraint
and Quality
Feedback
Parameters
EIR
EIR adjusts the rate of intentional frame dropping
EIR is translated for PBPAIR (considering it as PLR)
Feedback-based quality adjustment
High EIR increases energy saving but degrades video quality

17. Outline Motivation and Problem Statement Our Solution Experiments
Conclusion

18. End-to-End Experimental Framework
Mobile video applications such as video conferencing consist of mobile encoding, wireless(and wired) networks, and mobile decoding
They affect each other in terms of energy consumption and QoS
System Prototype and NS2 Simulator
System Prototype
Runs video encoding and decoding on system prototype emulating mobile devices, and returns video quality in PSNR
Estimates the energy consumption of a processor (CPU power)
NS2
Network Simulator
Estimates the energy consumption of WNI (transmission power)

19. Experimental Setup System System NS2 Prototype Prototype
Mobile Video Encoding
Mobile Video Decoding
Encoder
Transmit
Transmit
Encoder
Network
System
Prototype
System
Prototype
NS2
CPU energy for encoding
video quality (frame drop)
CPU energy for encoding
video quality (packet loss)
WNI energy for transmit
WNI energy for receive

20. Evaluation Video Encoding Energy Consumption Video Quality GOP-K
PBPAIR
EE-PBPAIR
Energy Consumption
Enc EC (Energy Consumption for Encoding) + Tx EC (Energy Consumption for Transmission)
Rx EC (Energy Consumption for Receiving) + Dec EC (Energy Consumption for Decoding)
Video Quality
Video Quality at encoder after intentional frame dropping
Video Quality at decoder after packet losses in networks

21. Experimental Results Energy Reduction from Active Error Exploitation
Extended Energy/QoS Tradeoff

22. Extended Tradeoff Space
PLR = 5% and EIR = 0% to 50%
EE-PBPAIR extends interesting tradeoff spaces

23. PSNR: Peak Signal to Noise Ratio
Energy Saving
EC = Energy Consumption
Enc EC = EC for Encoding
Tx EC = EC for Transmission
Dec EC = EC for Decoding
Rx EC = EC for Receiving
PLR = 10% and EIR = 10%
Energy saving occurs at every component in a path from encoding to decoding in mobile video applications
PSNR: Peak Signal to Noise Ratio

24. Energy Reduction at QoS Cost
At 10% cost of video quality, EE-PBPAIR can save the energy consumption of Enc and Tx by up to 49%

25. Outline Motivation and Problem Statement Our Solution Experiments
Conclusion

26. Conclusion Future Work
Intentional Frame Drop is one way to exploit errors actively
Propose an error-aware video encoding (EE-PBPAIR)
Intentional frame dropping and the nature of energy-efficiency of PBPAIR reduces the energy consumption for video encoding
Present a knob (EIR) to adjust the amount of errors considering the QoS feedback
Maintain the video quality using error-resilience of PBPAIR
Future Work
Intelligent Frame Dropping Techniques
Extend Active Error Exploitation to the system level with error-aware architecture and network protocols in distributed embedded systems

27. Any Questions? kyoungwl@ics.uci.edu
Thanks!
Any Questions?
Thank you! Any questions?

28. Backup Slides

29. Intentional Frame Drop and Packet Loss
Error-prone Networks
Packet Loss
Intentional frame drop

30. Error-Exploiting Video Encoder
EE-PBPAIR
Error-prone Networks
Intentional frame drop
Packet Loss
Error-Exploiting Video Encoder
Error-
Resilient
Video
Error-
Aware
Video
Original
Video
Error-Controller
(e.g., frame dropping)
Error-Resilient
Encoder
(e.g., PBPAIR)
EIR

31. Error Controller

32. Error-Induced Video Data
Error-Concealment f1 f4 f3 f2 f3 f2 f1
Error-Prone
Network
f4 is lost
Error-Induced Video Data
Error-Concealment Techniques
Interpolate the lost frame using near frames
Substitute the near frame for the lost one
(e.g.) f2 is copied for f3 (the lost one) in displaying frames

33. GOP (Group of Picture)
Standard H.263 Video Encoder with varying IP-ratio
Higher IP-ratio generates more compressed video output, which consumes more energy
Encoder
GOP
Intra Frame
Static Constraint of Compression Rate
IP-ratio
(KNOB)
Standard video encoding, which is unaware of energy
consumption and error-resiliency

34. PBPAIR
Proactively estimate the probability of the correctness, and adapt the intra_th (KNOB) based on the current network PLR (Packet Loss Rate)
Encoder
PBPAIR
Intra MB
PLR from Network Channel
Intra_Th
(KNOB)
Error-Resilient Encoding, which can satisfy a given PSNR,
and reduce the energy consumption for encoding

35. EE-PBPAIR
EE-PBPAIR introduces another KNOB (intentional EIR) other than Intra_Th, and can further save the energy consumption
Encoder
EE-PBPAIR
Intra MB
PLR from Network Channel
I-FS
Intentional EIR
(KNOB)
Intra_Th
(KNOB)
Error-Introduced Video Encoding, which can still satisfy a given PSNR,
and further maximize the energy saving compared to PBPAIR

36. System Prototype + NS2

37. Adaptive EE-PBPAIR