UTILIZING VIDEO ENCODING FOR POWER MANAGEMENT

OVER WIRELESS NETWORKS

C. Bouras

1,2

, K. Stamos

1,2,3

, S. Stroumpis

1,2

and G. Zaoudis

1,2

Research Academic Computer Technology Institute, Patras, Greece

Computer Engineering and Informatics Department, University of Patras, Patras, Greece

Technological Educational Institute of Patras, Patras, Greece

Keywords: Cross-layer, TFRC, Power management, Wireless, Video transmission, H.264.

Abstract: Power management is especially important in wireless networks because of the emergence of energy

constrained mobile devices. This paper examines the ways that video encoding information can be used for

more efficient power management. In particular, we propose and evaluate cross layer mechanisms that fine-

tune transmission power depending on the video frame being encoded and we examine how this mechanism

may be combined with algorithms that adjust transmission power based on feedback reports when TCP

Friendly Rate Control (TFRC) is used as the transport protocol. Since typical video encoding uses frames of

varying importance for the overall quality, this approach can improve received video quality and make

better usage of available power.

1 INTRODUCTION

Packet losses in wired networks mainly occur due to

congestion in the path between the sender and the

receiver, whereas in wireless networks the packet

losses mainly occur due to corrupted packets. That is

a result of the low Signal to Noise Ratio (SNR), the

multi-path signal fading and the interference from

neighbouring transmissions. A second difference

between wired and wireless networks is the

“mobility factor”. Mobility in wireless networks

introduces a number of additional barriers in

multimedia data transmission.

H.264/AVC is an essential component in

wireless video application due to its excellent

compression efficiency and network-friendly design.

Typical video encoding standards define various

types of frames (Stockhammer T., 2005) with

varying importance in terms of information and

compressibility. The main picture types as presented

in Figure 1 are I-frames, P-frames and B-frames. I-

frames are independent of other frames, P-frames

are dependent on previous frames, and B-frames are

dependent both on previous and future frames.

Therefore, a video stream is expected to suffer more

quality degradation when an I-frame is lost or

delayed instead of a B-frame. The latest standard

H.264/MPEG-4 defines slices instead of frames,

which are more fine-grained elements that make part

of a video picture.

Figure 1: Dependence of various video frame types.

Cross layer design refers to protocol design done

by actively exploiting the dependence between

protocol layers to obtain performance gains. This is

unlike layering, where the protocols at the different

layers are designed independently. In particular,

cross layer techniques allow the exchange of

information between different network layers for the

sake of improved performance. Power management

in particular can significantly benefit from

application layer information related to the type of

frame being encoded and its importance in the

decoding of the video stream at the receiving end of

the transmission. A theoretical discussion of the

cross-layer problem framework can be found in

(Van der Schaar, 2005).

The trade-off between increased power

consumption and improved signal strength has been

explored by various researchers studying

Transmission Control Protocol (TCP) modifications

(Tsaoussidis, 2000, Zhang 2001, Jones 2001) trying

to combine reduced power consumption with

increased data throughput. Wireless standards such

Bouras C., Stamos K., Stroumpis S. and Zaoudis G..

UTILIZING VIDEO ENCODING FOR POWER MANAGEMENT OVER WIRELESS NETWORKS.

DOI: 10.5220/0003519700570062

In Proceedings of the International Conference on Data Communication Networking and Optical Communication System (DCNET-2011), pages 57-62

ISBN: 978-989-8425-69-0

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

as IEEE 802.11 specify power saving mechanisms

(IEEE 802.11 PSM Standard), although studies have

shown that PSM (Power Saving Mode) and other

similar mechanisms carry a significant performance

penalty in terms of throughput (Molta, 2005, Chen,

2004, Anastasi, 2004, Simunic, 2005). The goal of

this paper is to present a mechanism for cross layer

power management for video transmission over

wireless 802.11 networks utilizing the information

about video encoding.

The rest of this paper is structured as follows:

Section 2 describes the proposed mechanisms that

aim to achieve improved quality and power

consumption trade offs. Section 3 presents the

simulation testbed that was used for evaluating the

proposed mechanisms and section 4 discusses the

obtained results. Finally section 5 concludes the

paper with a summary of our proposal and ideas for

future work in this area.

2 POWER MANAGEMENT

UTILIZING VIDEO FRAME

INFORMATION

In this section we describe how the information

about video encoding can be utilized for the

adjustment of power levels at the transmitting node

and how this mechanism can be combined with a

power management mechanism we refer to as

Binary (Bouras, 2011).

2.1 Binary Mechanism

In this section for the sake of completeness we

present the basic characteristics of the Binary

mechanism which is described in detail in (Bouras,

2011), and which is tested in combination with the

new approaches presented in this work. The

mechanism is used by the sender of the video which

is transmitted over TFRC. The reason TFRC is

selected is because of its attractive properties in

combining TCP characteristics of good behaviour

towards other flows and UDP-type (User Datagram

Protocol) throughput. At the same time, it defines a

set of report metrics that are sent back to the sender

of the traffic. Every time the sender receives a TFRC

report from the receiver changes its state according

to the state it is in and the new data. The mechanism

after receiving the first report, if packet loss is not

satisfactory, defines a region in which it will try to

approximate the optimum power. The optimum

power is that which produces a desired value of

packet loss. After defining the region, the sender

will increase its power to the maximum possible in

that region and send the next TFRC packet with that

power (state A: Expanding Power Region). When

the sender receives the next report, it tests whether

there has been a significant improvement (state B:

Improvement and Constraint testing). If there has

been an improvement and packet loss is below a

predetermined threshold, goes to state C (Lower

Consumption) or else repeats the actions of state A.

In state C, the mechanism sets the power to the

middle of the defined region and the sender goes to

state D (Constraint Testing). In state D the algorithm

tests whether the packet loss constraints are still

satisfied and, if this is the case, it repeats state C. If

this is not the case the algorithm goes to state E

(Backtracking to Last Known Acceptable Value),

where it goes back to the previous known acceptable

power value. The mechanism stays at state E while

the packet loss value is acceptable and, if not, it goes

back to state A.

Figure 2: Binary mechanism.

2.2 Exploiting Video Frame

Information

Since I-frames contain the most important

information compared to the rest of the frames, and

their loss may affect multiple frames before and

after in the frame sequence, it is reasonable to make

sure that they reach their destination. If the receiving

mobile node has moved further away from the

transmitting node, a transmission power increase

may mitigate weak signal reception problems.

However, packet losses may also be due to other

factors, such as channel congestion, and then power

increases offer no benefit. This is where the binary

mechanism is needed: its operation is to quickly

identify the optimal level of power for a given

DCNET 2011 - International Conference on Data Communication Networking

network condition depending on available

information about packet losses. However, since the

identification of an efficient power level

unavoidably has to examine several iterations of

packet loss reports, it is complemented by direct

changes depending on the frame type as discussed

below.

We therefore introduce a modification to the

adaptive algorithm presented above that tries to

heuristically increase power levels only when it is

expected to produce some tangible beneficial effect.

onBackground(BinaryMechanism())

while (true) {

frameType=checkMPEG4FrameType()

currPower=getCurrentPower()

if (frameType == I) {

setPower(P

*currPower)

} else if (frameType == P){

setPower(P

*currPower)

} else {

setPower(P

*currPower)

}

Figure 3: Triple cross-layer power management.

The P

, P

values are fixed for a transmitting

node and quantify the amount of importance that

each type of frame has relative to the rest. It is

therefore imperative that P

>=P

. In section 4

we present the selected values for the type of

encoding that was simulated and tested.

2.3 Triple Cross-layer Approach

The combinations of the above mechanism with the

binary mechanism lead to a new cross-layer design

between Application-Transport-Physical layers

(Srivastava, 2005). This approach is being described

in Figure 4.

Figure 4: Proposed cross-layer design.

3 TESTBED SETUP

For our experiments we have used the Network

Simulator 2 (ns-2.34, www.isi.edu/nsnam/ns/) as a

basic tool for simulating multimedia data

transmission over wireless networks. In order to

simulate MPEG-4 video transmission using ns-2,

another software package is needed, namely

Evalvid-RA (Lie, 2008,

www.item.ntnu.no/~arnelie/Evalvid-RA.htm).

Evalvid-RA supports rate-adaptive multimedia

transfer based on tracefile generation of an MPEG

video file. A typical tracefile provides information

for frame number, frame type, size, fragmentation

into segments and timing for each video frame. The

multimedia transfer is simulated by using the

generated tracefile and not the actual binary

multimedia content. The simulator keeps its own

tracefiles holding information on timing and

throughput of packets at each node during

simulation. Combining this information and the

original videofile Evalvid-RA can rebuild the

videofile as it would have been received on a real

network. Additionally, by using the Evalvid-RA

toolset the total noise introduced can be measured

(in dB PSNR) as well as Mean Opinion Scores

(MOS) can be calculated. An example

implementation is illustrated in (Haukass, 2007).

Objective PSNR measurements can be

approximately matched to subjective MOS

according to the standardized Table 1. The MOS

scores reported below are derived from the

automatic PSNR to MOS mapping according to the

following table.

Table 1: ITU-R Quality and impaired scale (ITU-R, 2002)

and possible PSNR to MOS mapping (Klaue, 2003).

PSNR

[dB]

MOS Impairment

>37 Excellent (5) Imperceptible

31-37 Good (4)

Perceptible, but

not annoying

25-31 Fair (3) Slightly annoying

20-25 Poor (2) Annoying

<20 Bad (1) Very annoying

4 EXPERIMENTS AND RESULTS

In our experiments we used the akiyo sample video

found in http://media.xiph.org/video/derf/ for video

streaming. We transfer H.264 video over TFRC over

UTILIZING VIDEO ENCODING FOR POWER MANAGEMENT OVER WIRELESS NETWORKS

a wireless link and in particular over a single hop in

a wireless ad hoc network. Selection of P

, P

values for this specific video encoding was 1.3, 1.1

and 0.9 respectively. In order to model various

instances of network degradation, we have

performed a series of experiments with various

scenarios, with both stationary and mobile nodes:

 Scenario 1: Two nodes, both stationary

 Scenario 2: Two nodes, one stationary, one

moving away

 Scenario 3: Two nodes, one stationary, one

moving closer and then moving away

 Scenario 4: Two nodes, one stationary, one

moving closer

 Scenario 5: Two nodes, one stationary, one

moving away and then stops moving

In all scenarios, the nodes communicate wirelessly

using 802.11 MAC protocol and the distributed

coordination function (DCF) from the Carnegie

Mellon University. Propagation model used was

two-ray ground reflection model.

Table 2: Stationary nodes.

Triple

cross-layer

power mgmt

Binary

power

mgmt

Without

power

mgmt

PSNR average 37.8 37.6 37.1

Energy

Consumption

0.051 W 0.046 W 0.046 W

MOS Excellent (5)

Excellent

(5)

Excellent

(5)

In this scenario both nodes are stationary, so

power requirements do not vary. Nevertheless,

power management mechanisms offer a better PSNR

with slightly increase in transmission power.

This time, the proposed mechanism displays a

noticeable performance advantage over the approach

Table 3: One node moving away.

Triple cross-

layer power

mgmt

Binary

power

mgmt

Without

power

mgmt

PSNR

average

35.3 34.8 30.2

Energy

Consumption

0.049 W 0.047 W 0.047 W

MOS Good(4) Good(4) Fair (3)

without any mechanism. We observe that it actually

achieves Good Mean Opinion Score while the value

for the same scenario without any power

management mechanism in fair.

Table 4: One node moving closer and then away.

Triple cross-

layer power

mgmt

Binary

power

mgmt

Without

power

mgmt

PSNR

average

36.2 36.1 33.3

Energy

Consumption

0.050 W 0.048 W 0.048 W

MOS Good(4) Good(4) Good (4)

The same applies to this scenario, where the

power management mechanisms significantly

improve received video quality as shown by the

PSNR values. Power increase is non-existent or very

small in both cases. The reason is that both

mechanisms are capable to adapt to the changing

distances between the nodes and tweak the power

levels accordingly.

Table 5: One node moving closer.

Triple

cross-layer

power

mgmt

Binary

power

mgmt

Without power

mgmt

PSNR

average

38.8 37.9 34.6

Energy

Consumption

0.049 W 0.046 W 0.046 W

MOS

Excellent

(5)

Excellent

(5)

Good (4)

When a node is moving closer it is natural to

achieve a better PSNR value in all methods. By also

using rapid adjustment of power even better results

occur, whereas power consumption again stays

relatively low.

In this case where the node stops after moving,

the power management mechanisms adjust

themselves to be as power saving as possible

without making a reduction to the quality of video

image transmitted by giving greater weight to the

most important frames.

DCNET 2011 - International Conference on Data Communication Networking

Table 6: One node moving away and then stop moving.

Triple

cross-layer

power

mgmt

Binary power

mgmt

Without

power

mgmt

PSNR

average

37.2 36.8 31.8

Energy

Consumption

0.050 W 0.047 W 0.047 W

MOS

Excellent

(5)

Good(4) Good (4)

The results from all scenarios demonstrate that in all

cases the proposed mechanism significantly

outperforms the default behaviour (without any

power management mechanism) as it achieves

higher video quality reception, with only slight

increases of average power levels. The following

figure summarizes the results of the experiments in

terms of the ratio PSNR/power which gives us an

estimation of how well the trade-off between power

consumption and video quality is balanced.

We can see that the proposed mechanism

achieves a significantly improved trade-off, which

means that the mobile nodes may gain in either

quality or power consumption or both, compared to

the original approach that does not utilize the cross-

layer information.

600

620

640

660

680

700

720

740

760

780

800

Moving aw ay Moving closer

and then

aw ay

Moving clos er Moving aw ay

and then stop

PSNR/pow er

Proposed mechanism

Original behav ior

Figure 5: PSNR/power ratio.

5 CONCLUSIONS AND

FUTURE WORK

In this paper we have proposed an advanced power

management cross-layer mechanism for power

management in wireless TFRC transmission, which

significantly improves both the objective quality of

the transmitted video, and makes more optimal

usage of available power utilizing information from

three different layers of the TCP/IP stack. In this

paper we have seen that minor tweaks to the

algorithm can achieve both goals and can be fine-

tuned depending on the specific requirements of

each particular situation. Most of the presented

approaches have their strong and weak points,

depending on the specific type of movement

performed by the nodes.

The proposed cross-layer mechanism could be

further improved in a wide range of ways. Firstly,

we could estimate power consumption by taking into

account both power consumption for the

computational complexity of encoding and the

power consumption for the transmission.

Furthermore, by using the capabilities of H.264 one

can change video quality dynamically so that there

can be adaptation of the transmission rate according

to the available bandwidth. Finally, the latest and

most promising mechanism for wireless

transmission of H.264 video is SVC (Scalable Video

Coding). SVC (Schwarz, 2006) enables the

transmission and decoding of partial bit streams to

provide video services with lower temporal or

spatial resolutions or reduced fidelity while retaining

a reconstruction quality that is high relative to the

rate of the partial bit streams. Hence, SVC provides

functionalities such as graceful degradation in lossy

transmission environments as well as bit rate,

format, and power adaptation, so another step could

be to use our proposed power management

mechanism, while exploiting features the structure

of H.264 video.

REFERENCES

Anastasi G., Conti M., Gregori E., Passarella A., 2004. “A

performance study of power-saving polices for Wi-Fi

hotspots”, Computer Networks: The International

Journal of Computer and Telecommunications

Networking. Volume 45 , Issue 3, pp. 295-318.

Bouras C., Gkamas A., Kapoulas V., Papapanagiotou V.,

Stamos K., Zaoudis G., 2011. “Power Management

Adaptation Techniques for Video Transmission over

TFRC”, International Journal of Network

Management (IJNM), Wiley InterScience.

Chen H., Huang C.-W., 2004. “Power management

modeling and optimal policy for IEEE 802.11 WLAN

systems”, IEEE Vehicular Technology Conference.

Haukass T., 2007. “Rate Adaptive Video Streaming over

Wireless Networks Explained & Explored”, MSc

thesis, Norwegian University of Science and

Technology, Department of Telematics.

Jones C. E., Sivalingam K. M., Agrawal P., Chen J. C.,

2001. “A Survey of Energy Efficient Network

Protocols for Wireless Networks”, Wireless Networks.

Volume 7, Issue 4. pp. 343-358.

IEEE 802.11 PSM Standard. Power Management for

Wireless Networks. Section 11.11.2

UTILIZING VIDEO ENCODING FOR POWER MANAGEMENT OVER WIRELESS NETWORKS

ITU-R 2002. Recommendations BT.500-11. Methodology

for the subjective assessment of the quality of

television pictures.

Klaue J., Rathke B., Wolisz A, 2003. EvalVid: a

framework for video transmission and quality

evaluation. In Proceedings of the 13th International

Conference on Modeling, Techniques and Tools for

Computer Performance Evaluation, Urbana, IL.

Lie A., Klaue J., 2008. “Evalvid-RA: Trace Driven

Simulation of Rate Adaptive MPEG-4 VBR Video”,

Multimedia Systems, Springer Berlin / Heidelberg,

Volume 14, Number 1, pp. 33-50.

Molta D., 2005. “Wi-Fi and the need for more power”,

Network Computing.

Schwarz H., Marpe D., Wiegand T., 2006. Overview of

the Scalable Video Coding Extension of the

H.264/AVC Standard Paper templates. In

TEMPLATE’06, 1st International Conference on

Template Production. SciTePress.

Simunic T., 2005. “Power Saving Techniques for Wireless

LANs”, Conference on Design, Automation and Test

in Europe - Volume 3. pp. 96-97.

Srivastava M. 2005. “Cross-Layer Design: A Survey and

the Road Ahead”, IEEE Communications Magazine.

Stockhammer T., Hannuksela M. M., 2005. “H.264/AVC

video for wireless transmission,” IEEE Wireless

Communications, vol. 12, no. 4, pp. 6–13.

Tsaoussidis V., Badr H., 2000. “TCP-Probing: Towards an

Error Control Schema with Energy and Throughput

Performance Gains” 8th IEEE Conference on Network

Protocols, Japan.

Van der Schaar M, Sai Shankar N., 2005. Cross-layer

wireless multimedia transmission: challenges,

principles and new paradigms. IEEE Wireless

Communications; 12(4): 50–58.

Zhang C., Tsaoussidis V., 2001. “TCP Real: Improving

Real-time Capabilities of TCP over Heterogeneous

Networks” 11th IEEE/ACM NOSSDAV, NY.

DCNET 2011 - International Conference on Data Communication Networking