Research Article QoE Perceptive Cross-Layer Energy Efficient Method for Mobile Video ...downloads.hindawi.com/journals/ijdsn/2015/980174.pdf · 2015-11-24 · Research Article QoE

Research ArticleQoE Perceptive Cross-Layer Energy Efficient Method forMobile Video Devices

Zhaoming Lu Hongchun Zhang Yawen Chen Hua Shao and Xiangming Wen

Beijing University of Posts and Telecommunications Beijing Key Laboratory of Network System Architecture and ConvergenceBeijing 100876 China

Correspondence should be addressed to Zhaoming Lu lzy 0372163com

Received 21 April 2015 Revised 13 August 2015 Accepted 24 August 2015

Academic Editor Andrei Gurtov

Copyright copy 2015 Zhaoming Lu et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Over the last couple of years video service distribution among smart phones and other mobile video devices is becomingincreasingly popular in sensor networks However the huge energy consumption caused by video encoding and transmitting andthe slowly evolving battery technologies become the major bottlenecks that hinder the development of video streaming servicesHence energy efficient video coding and transmitting solutions are required to be investigated Yet energy consumption reductionof mobile video devices will be accompanied with Quality of Experience (QoE) degradation of video applications Such diversetendency makes it difficult to encode and transmit video streams with less energy as well as better QoE This paper analyzes themajor energy consuming factors in mobile video devices A specific energy consumption model concerning encoding bitrates andtransmitting power level is built Further a noninvasive QoE perceptive model is adopted so that the energy efficiency problembecomes a cross-layer optimization problem Chaos particle swarm optimization is used to solve this cross-layer optimizationproblem with fast convergence By this method energy consumption of mobile video device is minimized with acceptable QoE forvideo users At last Pareto front of energy and QoE is analyzed to certify the performance of our method

1 Introduction

By the end of 2019 mobile data traffic is expected to growat a Compound Average Growth Rate (CAGR) of around45 percent [1] This will result in an increase of around 10times by the end of 2019 Users are consuming more datatraffic per subscription mainly driven by video servicesVideo streaming services such as Netflix HBO and Vimeohave shown very strong uptake in mobile devices New videotechnologies such as High Dynamic Range (HDR) [2] andultra-high definition (UHD) [3] videos will become morecommon With providing truly immersive QoE to the usersthey will bring about more bitrate to wireless networks andmore energy consumption for video devices In additionoperators are increasingly making their own TV servicesavailable as streaming services on mobile networks This isfacilitated by the higher network speeds that come with HighSpeed Packet Access (HSPA) and Long TermEvolution (LTE)development It is increasingly common to share video clipsover mobile networks and smart mobile devices make up an

increasing capacity of video capture processing and viewingThis has contributed to the growth in mobile data trafficCisco Forecast [4] shows that until 2017 70 percent of mobiledata traffic will be video traffic Video applications must bethe primary service in wireless sensor networks Howeverthere are two challenges confronted by the development ofvideo applications in mobile devices

Firstly huge energy consumption of encoding and decod-ing and big data rate in radio interface for video streamingbring great challenges for energy saving of mobile videodevices In future with the development of HDR and UHDvideo devices may require more energy consumption How-ever battery is usually adopted as the energy source ofmobilevideo devices In spite of several decadesrsquo developmentunfortunately the capacity of battery is still the bottleneckof the development of pocket devices while the hardwareof mobile devices is frequently updated In addition formobile video devices high complexity of video processingalgorithms and large bitrate of video streaming data willconsiderably cut down the life of batteries Energy efficient

Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2015 Article ID 980174 13 pageshttpdxdoiorg1011552015980174

2 International Journal of Distributed Sensor Networks

method should be proposed to reduce the energy consump-tion of video streaming and extend the lifetime of batterysignificantly Therefore as there is no significant advancein battery technology energy efficient mechanism for videoencoding and transmitting becomes a crucial issue to beinvestigated

Secondly wireless multimedia sensor networks deeplychange the interactive mode between cyber space andphysical space which have been applied in areas such asmilitary health care traffic control and disaster resistanceMultimedia sensors such as video sensors and image sensorsare engaged in perception processing and transmission notonly for simple data information but also for multimediainformation However dynamic and unpredicted interfer-ence in wireless sensor networks will make a big dent inquality of video services In addition video service is a kindof service with strong content correlation QoS metrics suchas throughput delay and jitter cannot fully reflect the userperceived quality of video service QoE which depicts theapproximate subjective experience of services in wirelesssensor networks by terminal users is increasingly adopted asthe metric for video quality With regard to video services inwireless sensor networks how to dynamically accommodatevideo encoding scheme and manage radio resources so thatQoE of mobile video services could be assured is an issue thaturgently needs to be addressed

Therefore in wireless sensor networks video devicesneed not only to assure the QoE for video applications butalso to deal with the huge energy consumption generated byhigh data rate of video streaming Appropriate strategy profileshould be picked out to guarantee the QoE of video applica-tions while minimizing the energy consumption of mobilevideo devices However there are conflicts between these twoobjectives Reducing energy consumption of video streamingis always in the price of degradation of transmitting powerlevel which will lead to degradation of the QoE for video ser-vices Meanwhile enhancement of QoE is always in the priceof enhancement of transmitting power level which will leadto extra energy consumption of video services Therefore itis necessary to find an appropriate tradeoff between energysaving and QoE so as to reduce the energy consumptioncaused by video streaming with assurance of QoE

In this paper as shown in Figure 1 we propose a QoEperceptive cross-layer energy efficient (QP-CEE) method formobile video devices to dispose of the energy efficiencyproblem and QoE assurance problem together Energy con-sumption of mobile devices is reduced and QoE of videoapplications in mobile video device is enhanced throughdynamic regulation of transmitting power level modulationand coding scheme (MCS) and encoding bitrate of videostreaming By elaborative analysis of the encoding andtransmitting process ofmobile video device a comprehensiveenergy consumption model for video encoding and trans-mitting is established A noninvasive QoE perception modelis adopted in terms of the mean opinion score (MOS) Theproposed QoE perception model is significant in its ownright as the optimization of QoE is crucial for mobile videoservices Energy efficiency problem ismodeled as a joint opti-mization problem including optimization of QoE and energy

Encodingenergy

consumption

Transmittingenergy

consumption

Energyconsumption

model for videodevices

Encodingbitrate

Transmittingpower level

QoE perceptionmodel

Cross-layeroptimization

Suboptimalstrategies

CPSO

MCS

Figure 1 Framework of QoE perceptive cross-layer energy effi-ciency method for video devices

consumption Then chaos particle swarm optimization algo-rithm is adopted to search for optimal power level and videoencoding rate in parallel so that the energy consumption isminimized and QoE of video applications is ensured FinallyPareto front of energy and QoE is analyzed to certify theperformance of the proposed method There are three maincontributions in this paper Firstly encoding bitrate modu-lation and coding scheme and transmitting power level areinvestigated simultaneously by which multilayer (includingapplication layer and physical layer) performance gain isachieved Secondly an energy efficient method based onQoEperception is proposed and both requirements ofmobile userfor video quality and minimization of energy consumptionformobile devices are achievedThirdly chaos particle swarmoptimization is adopted to solve the optimization problem Inone aspect the search process is not apt to converge to localoptimum solution In the other aspect time delay for solvingthe cross-layer optimization problem can be shortened byparallel process which can meet the real-time requirementsfor video services such as video live and video conference

The rest of this paper is arranged as follows In Section 2state-of-the-art research about QoE perception methodenergy saving strategy tradeoff method of QoE and energysaving is surveyed and summarized and the open questionsand possible research directions are analyzed In Section 3the uplink energy efficient scenario for mobile video devicesis described The energy consumption model for mobilevideo devices is proposed in Section 4 In Section 5 jointoptimization problem of QoE and energy consumption isinvestigated and chaos particle swarm optimization methodis adopted to find optimum encoding bitrate and powerlevel Evaluation and analysis for our work are carried out inSection 6 In Section 7 we conclude our work in this paper

2 Related Work

Various energy efficient methods have been proposed toimprove the energy efficiency of wireless networks Spe-cially for video streaming with growing popularity the largeenergy consumption generated by encoding and transmitting

International Journal of Distributed Sensor Networks 3

constantly drives researchers to investigate energy efficienttechniques A lot of work has been done on software andhardware of mobile devices to extend the battery lifetimeFor example low complexity encoding strategies [5ndash7] lowpower embedded video coding algorithm [8] adaptive powercontrol strategies [9ndash11] and joint encoding and hardwareadaptive method [12] are proposed as the possible solutionsHowever performance enhancement resulting from theseresearches cannotmeet the increasing requirements of energysaving by users

Recently researchers believe that cross-layer design playsan important role in energy efficiency methods for multi-media systems [13 14] Joint video encoding and wirelesstransmission control [14 15] is proposed to implement thecross-layer design for mobile video stream applicationsFor mobile video devices energy consumption is primarilycaused by encoding and transmission However in [5ndash15]the authors propose energy efficiency method while there isno system level energy consumption model to support theirmethods Power-rate-distortion model [16 17] is proposedto analyze the energy consumption of video applications byextending traditional R-D analyzed model Researchers fromSimon Fraser University propose a novel energy efficientmethod [18ndash20] through choosing the most suitable numberof encoding layers Unfortunately QoE is not considered asan optimization target for these energy efficient methods

Actually to ensure QoE of emerging video applicationsfor users QoE perception and prediction model have beendeeply investigated recently Amounts of objective videoquality assessment methods are proposed to perceive thequality for video applications [21ndash23] However these distor-tion based methods always calculate video quality throughmeasuring the distortion between source video and receivedvideo and then mapping this distortion to MOS or otherQoE metrics The MOVIE model presented in [24] is afull reference model developed from the spatial-temporalfeatures of the video Full reference video quality predictionmodels are difficult to implement for real-time monitoringdue to their complexity In order to estimate user experiencedvideo quality methods with no reference are also employedMany existing quality metrics usually use bit-error rate thathas low correlation with user perceived video quality The noreference video quality estimation method proposed in [25]estimates video quality degradation with respect to humanperception that is measured as video quality metric scoreswhich require too much processing power that cannot beobtained in handled mobile devices The model presentedin [26] combines video content features with the distortionscaused by the encoder The work in [27] introduces aldquorelative qualityrdquo metric (rPSNR) over IP networks fromnetwork distortions only by developing an approach that iscapable of mapping network statistics for example packetlosses available from simple measurements to the quality ofvideo sequences reconstructed by receivers Some researches[28 29] prove that video quality is affected by parametersassociated with the encoder (eg source bitrate) and thenetwork condition (eg packet loss) Khan et al [30] proposea QoE perception model based on video streaming bycomprehensive analysis of video content encoding bitrate

and transmitting power level However for mobile videodevices it is not practical to merely enhance QoE for videoapplications without any care for huge energy consumptioncaused by video encoding and transmitting

There are few researches on joint optimization of QoEand energy Current researches on tradeoff between QoE andenergy consumption in wireless networks focus on theoreti-cal analysis which always have high complexity and cannotbe used in practical scenario Liberal et al propose [31] a lowencoding complexity and transmitting power consumptionmethod by dynamic heuristic programming other than jointoptimization In [32] the authors propose a new metric EE-QoE which denotes QoE improvements per unit powerThismetric is applied into wireless cellular networks with interfer-ence limit However these methods lead to high complexityTherefore for mobile video devices with limited computa-tional capacity a novel practical joint QoE and energy effi-cient method with low complexity is urgent to be proposed

In this paper a QoE perceptive cross-layer energy effi-ciency method for mobile video devices is proposed tojointly optimize the energy consumption andQoE formobilevideo services Comprehensive optimization model of QoEand energy consumption is investigated and rapid cross-layer optimization method based on chaos particle swarmoptimization (CPSO) algorithm is adopted to make thevideo device learn appropriate encoding and transmittingparameters in real time

3 Energy Efficient Scenario forMobile Video Devices

31 Transmission Scenario of Video Services To focus on theenergy efficiency problem caused by video services onmobiledevices we assume that there are only real-time video ser-vices such as gaming video live and video conference run-ning on the mobile video devices for simplicityThis assump-tion is reasonable as in all kinds of mobile services video ser-vices are the main source of energy consumption One obvi-ous characteristic of these services is the requirement of real-time processing Hence mobile devices need to encode anddecode video streaming in real time In this paper appropri-ate resource portfolio in radio uplink is investigated bywhichnot only QoE assurance of video users but also minimizationof mobile devicersquos energy consumption could be achievedFigure 2 shows a typical scenario for real-time video services

32 Major Energy Consuming Factors Previous studies haveshown that [33] there are many impact factors for energyconsumption of video devices including service type mobilenetwork transmitting power level of mobile device screenBluetoothwireless local network andGPS Since energy con-sumption of factors such as screen GPS and Bluetooth couldbe optimized from operation level rather than algorithmlevel in this paper we primarily investigate the two factorsencoding bitrate and transmitting power level of mobiledevices which have specific impact on energy consumptionon mobile devices To explore the relationship in qualitativelevel of these two factors and two objectives associated withenergy efficiency problem such as energy consumption and


Table 1 Samples of encoding bitrate and normalized transmittingpower level

Test sample 1 2 3 4 5 6Encoding bitrate(kbps) 50 70 90 110 130 150

Normalizedtransmittingpower level

033 047 060 073 087 10

Video sequen

ce

Capture Encode Transmit

Figure 2 Scenario for real-time video services

QoE an experimental test is implemented Two differenttypes of mobile devices HTC One and HUAWEI Honorare selected as the test devices Test video clips that comefrom LIVE database [34] with a QCIF resolution are usedin this experiment As shown in Table 1 a set of samples ofencoding bitrate and normalized transmitting power levelare selected as the input arguments and the correspondingnormalized transmitting power consumption is measuredWe assess the MOS value of video clips through subjectivemarking by users according to ITUrsquos specification [35] forvideo quality assessment Then the relationships are figuredout by artificial neural network toolbox in MATLAB Finallyenergy consumption and QoE (measured by MOS) curves ofencoding bitrate and transmitting power level are drawn uprespectively

321 Encoding Bitrate To study how the encoding bitrateinfluences the energy consumption andQoE of video servicesin UMTS we change the bitrate from 50 kbps to 150 kbpsPower level of mobile devices is 20 dBm video stream isencoding by H264 and Bluetooth and GPS are both turnedoff The relationship of QoE and energy consumption toencoding bitrate is depicted in Figure 3

Energy consumption of video devices can be dividedinto encoding energy consumption and transmitting energyconsumption With the growth of encoding bitrate lessencoding energy consumption is required for the same videosequence As the encoding energy consumption decreasesthe encoder will generate more encoded video data whichimpliesmore energy consumption is required for video trans-mission in wireless channels When the growth rate of videotransmitting energy consumption exceeds the decent rate ofvideo encoding energy consumption the whole energy con-sumption will decline with the growth of encoding bitrate Asillustrated in Figure 3 energy consumption of video devicesis first decreasing with the improvement of video encodingbitrate due to the decrease of encoding energy consumption

1

095

09

085

08

075

Nor

mal

ized

ener

gy

50 60 70 80 90 100 110 120 130 140 150

Encoding bitrate (kbps)

43

425

42

415

41

405

MO

S

Figure 3 Impact on energy consumption and QoE by video bitrate(the power level of mobile devices is 20 dBm)

05 055 06 065 07 075 08 085 09 095 105

1

15

Nor

mal

ized

ener

gy

Normalized transmitting power level

423

424

425

MO

S

Normalized energyMOS

Figure 4 Impact on energy consumption and QoE by transmittingpower level (the encoding bitrate of video stream is fixed on100 kbps)

and then increasing with the growth of transmitting energyconsumption while QoE of video applications is increasingall the time

322 Transmitting Power Level To study the impact onenergy consumption and QoE by transmitting power levelwe change the transmitting power of mobile video devicesBluetooth and GPS are turned off and the encoding bitrateof video stream is fixed on 100 kbps The relationships ofQoE and energy consumption to transmitting power level aredepicted in Figure 4 It shows that energy consumption ofmobile video devices is increasing with the improvement oftransmitting power level as well as the QoE of video services

Consequently both the video encoding bitrate and trans-mitting power lever have a close relationship with QoE andenergy consumption of video services By the above quali-tative analysis it is easy to conclude that through dynamicadjustment of transmitting power level and encoding bitrate


appropriate energy consumption and QoE tradeoff could beachieved In the following section we will investigate thespecific quantitative relationship of these two configurablefactors to energy consumption and QoE of video services

4 Energy Consumption Model forMobile Video Devices

For mobile video devices energy consumption model de-notes the energy consumed by encoding and transmitting forvideo streaming As video traffic is delivered throughwirelesschannels encoding bitrate is constrained by the channelcapacity which is dependent on the transmitting power ofmobile devices

41 Energy Consumption Model for Video Devices Energyconsumption in video devices can be divided into two partspower consumption caused by video encoding and powerconsumption caused by video transmitting Let 119875 denote thetotal power consumption ofmobile deviceDefine119875

119904and119875119905as

the encoding power and transmitting power respectively Asonly video services run on the mobile video devices energyconsumption can be calculated by

119875 = 119875119904+ 119875119905 (1)

Obviously 119875119904is in correlation with the complexity of

encoding process which is mainly dependent on encodingbitrate of video streaming The relationship among encodingpower distortion and encoding bitrate can be expressed as[17]

119863(SBR 119875119904) = 1205902

2minus120582SBRsdot119892(119875

119904)

0 le 119875 le 1 (2)

where 1205902 is the variance of encoded frame 120582 is a parameterdenoting the resource utilization of video encoding 119892(119875

119904)

is the inverse function of normalized energy consumptionfunction and 119892(0) = 0 119892(1) = 1 and

119892 (119875119904) = 1198751120574

119904 (3)

For diverse microprocessor with dynamic voltage regula-tion ability the range of 120574 is 1 le 120574 le 3

Generally two different video encoding schemes areinvolved One scheme is that video sequence is consideredto be steady and the optimal encoding power level is deter-mined through statistical averaging method [36] All videoframes are encoded with this constant encoding power levelThe other scheme is that encoder divides video sequence intodifferent video segments [17] and then resource allocationstrategy is optimized in terms of video segments to minimizethe overall power consumption As its adaptive characteristicfor the same video sequence the second scheme consumesless energy than the first one Define 119878 as the input videosequence which is divided into several video segments 119904

119894|

1 le 119894 le 119868 For a certain segment 119878119894 suppose that average

variance and model parameter are 1205902119894and 120582

119894 respectively

SBR119894and 119901

119894denote the encoding bitrate and encoding power

consumption respectively and 119863119894denotes the distortion

which is predicted in accordance with different encodingbitrates On the basis of formula (2) the encoding power of119878119894can be expressed by

119901119894= (

1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

(4)

For video sequence 119878

119875119904= sum

119894

119901119894= sum

119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

(5)

Therefore energy consumption of mobile video devicecan be calculated by

119875 = 119875119904+ 119875119905= sum

119894

119901119894+ 119875119905

= sum

119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

+ 119875119905

(6)

42 Encoding Bitrate Constraint Based on Automatic Modula-tion and Coding The available transmitting bitrate for videostreaming depends on the bit-error rate (BER) which can beexpressed by

SBRlimit = 119877119888sdot (1 minus BER (120574

119899))119871

(7)

where 119871 is the number of bits involved in video packet120574119899denotes Signal to Interference plus Noise Ratio (SINR)

of radio link 119877119888denotes the transmitting rate of wireless

channel in ideal conditionAutomatic modulation and coding (AMC) has different

alternative of constellations of modulation and rates of error-control codes based on the time-varying channel qualitywhich can effectively enhance the throughput of wirelesscommunication systems [37] with AMC the bit-error ratecan be depicted by

BER (120574119899) =

119886119898

119890120574119899times119887119898 (8)

where 119898 denotes the mode index of modulation and codingscheme Mobile devices are capable of dynamically choosingthe modulation and coding schemes and adjusting the trans-mission rates according to the measured channel conditionindicated by BER Given a particular modulation schemeBER is uniquely determined by the SINR experienced by thereceiver of the link Coefficients 119886

119898and 119887

119898are shown in

Table 2 In addition SINR 120574119899is dependent on video trans-

mitting power level 119901119905

120574119899=119901119905sdot1003816100381610038161003816ℎ119899

1003816100381610038161003816

2

1198730119861119873

(9)

where ℎ119899is the channel gain of channel 119899 119873

0is the spectral

density of white Gaussian noise 119861 is overall bandwidth 119873is the available number of channels According to the above


Table 2 Modulation mapping model

AMC mode (119898) 119898 = 1 119898 = 2 119898 = 3 119898 = 4 119898 = 5 119898 = 6

Modulation BPSK QPSK QPSK 16-QAM 16-QAM 64-QAMCoding rate (119888

119898) 12 12 34 916 34 34

119877119898(bitssym) 050 100 150 225 300 450

119886119898

11369 03351 02197 02081 01936 01887119887119898

75556 32543 15244 06250 03484 00871

formulas available transmission rate sustained by wirelesschannel can be expressed by


119899))119871

= 119877119888sdot (1 minus

119886119898

119890120574119899times119887119898)

119871

= 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(10)

Source bitrate of video streaming should satisfy thefollowing constraint

SBR119894le SBRlimit (11)

That is

SBR119894le 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(12)

5 QoE Perceptive Cross-Layer EnergyEfficient Method

To ensure QoE of video services in mobile devices a non-invasive video QoE perception model based on parametersassociated with encoder and wireless channel is adoptedto perceive video quality Through combining this QoEperception model and the energy consumption model formobile video devices a QoE perceptive cross-layer energyefficient model can be established Finally CPSO algorithmis adopted to solve this cross-layer optimization problem

51 Noninvasive QoE Perception Model for Video StreamingFor mobile video services mean opinion score (MOS) isusually used as metric for QoE measurement The MOSvalues range from 1 to 5 representing video quality fromldquobadrdquo to ldquoexcellentrdquo MOS is a subjective evaluation from theperspective of user experience which can be approximatedby some objective perception models In order to restrain thehigh complexity of traditional distortion based QoE percep-tion models a content-based noninvasive QoE perceptionmodel [30] for video services is presented which can beexpressed by

MOS =120572 + 120573 lowast ln (SBR) + CT lowast (120574 + 120575 lowast ln (SBR))

1 + 120578 lowast (BER) + 120590 (BER)2 lowastMBL (13)

Here CT represents the specific content type of videostreaming which depends on the spatial complexity andthe temporal activity of the depicted visual signal and candetermine the efficiency of the coding procedure Cluster

Table 3 Coefficients of QoE perception model

120572 120573 120574 120575 120578 120590

39560 00919 minus58497 09844 01028 minus00236

analysis tools based on multivariate statistics are adopted toextract a combination of temporal and spatial features ofvideo sequences by which video content is classified intodifferent groups MBL denotes mean burst length of videosequence which depends on BER For random uniform errorchannel model MBL is equal to 1 Values of other coefficientsof the model are listed in Table 3 Then the QoE perceptionmodel is established

The basic idea in this model is to move away from theuse of individual network parameters such as packet lossprobability or delay towards perceptual-based approach inorder to achieve the best possible perceived video qualityThis model takes into account quality degradation causedby the wireless channel and the encoder by combinationof parameters associated with the encoder and the wirelesschannel for different types of content

52 Cross-Layer Energy Efficiency Problem On the basis ofthe proposed energy consumption model in Section 4 tominimize the system power consumption as well as ensuringvideo userrsquos perceived quality the optimization problem canbe formulated by

min 119875 = min119891 (SBR 119875119905MCS)

= min(sum119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

+ 119875119905)

st SBR119894le 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

MOS ge MOSth

(14)

where 119875 is the total power consumption for a certain mobilevideo device MOS is user perceived quality of the videoservices and MOSth denotes the acceptable video qualitythreshold MCS is modulation and coding scheme Theobjective function requires a joint optimization of QoE andenergy consumption According to test in Section 3 thereare conflicts between QoE of video services and energy con-sumption of mobile video devices The target is to establishthe objective function and find the optimum solution toequalize the QoE and power consumption In this paperas video bitrate is a parameter in application layer while


each layerGive feedback to

solutionFind out optimal

modelconsumptionBuild energyFeedback

optimizerCross-layer

Parameter selection

Feedback

Parameter selection

level)(transmitting power

coding scheme)(modulation and

Physical layer

bitrate)(video encodingApplication layer

Figure 5 QoE perceptive cross-layer energy efficiency framework

modulation and coding scheme and transmitting power leverare parameters in physical layer a cross-layer optimizationproblem could be established

Figure 5 shows the QoE perceptive cross-layer energyefficiency framework In this framework these cross-layerparameters are adaptively regulated to figure out the opti-mum strategy profile so as to minimize the power consump-tion of mobile video devices and ensure the quality of videoservices As a result energy consumption is reduced and theoperational lifetime of the device is lengthened

53 Chaos Particle Swarm Optimization Algorithm Particleswarm optimization method was originally presented byEberhart and Kennedy in 1995 [38] inspired by socialbehavior of bird flocking in the process of migration andclustering Particle swarm optimization method does notrequire a centralized control node and each individual hasonly limited intelligence Through a number of informationinteractions of individuals swarm can put up strong intel-ligence However due to the inherent inertness of particlesoptimization process is often prone to converging to localoptimal solution Value of parameters and convergence ofoptimization are the key factors influencing the performanceand efficiency of algorithms In order to guarantee the globalconvergence of particle swarm optimization chaos theory isintroduced in this paper

Chaos refers to irregular or chaotic motion generatedby some nonlinear systems and the kinetics law of thesesystems can determine the unique evolution of the systemstatus over time from previous experience system The basicidea of chaotic particle swarm optimization algorithmmainlyembodies that the chaotic sequence is used to improve thediversity of population and ergodicity of particle searcheswithout changing the randomness nature of original particleswarm optimization Many neighborhood points of localoptimal solution by chaotic sequence in the iterative processwhich can help inert particles escape from local extremepointand search for the global optimal solution as soon as possible

In this chaotic particle swarm optimization algorithmaccording to formula (14) a constrained optimization prob-lem is formulated In order to find a feasible optimal

solution for this constrained cross-layer optimization prob-lem modification [39] of the objective function is imple-mented to convert the original constrained optimizationproblem to an unconstrained problem Firstly the constraintcondition is replaced by the following two inequalities

SBR119894minus 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

le 0

MOSth minusMOS le 0

(15)

Then define

ℎ (SBR 119875119905MCS) = max (SBR

119894minus 119877119888

sdot (1 minus 119886119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

MOSth

minusMOS)

(16)

Finally the cross-layer optimization problem without con-straint can be represented as

119891 (SBR 119875119905MCS) = ℎ (SBR 119875

119905MCS)

ℎ (SBR 119875119905MCS) gt 0

119891 (SBR 119875119905MCS) = 119886 tan [119891 (SBR 119875

119905MCS)] minus 120587

2

others

(17)

Therefore chaotic particle swarm optimization algorithmcould be used to solve this unconstrained cross-layer energyefficiency problem and the power consumptions of encodingand transmitting areminimized under condition of satisfyingthe MOS requirements for each video segment

The objective function 119891(SBR 119875119905MCS) is defined as

fitness function and optimized with respect to the vector119883 =

(SBR 119875119905MCS) Hence the abovementioned optimization

problems can be mathematically formulated by

min119883

119891 (SBR 119875119905MCS) 119883 isin 119863 sube 119877

3

(18)

where119863 denotes a 3-dimensional search spaceWe consider a swarm consisting of 119898 particles 119909

1 1199092

119909119898 As optimummodulation and coding scheme could be

achieved by automatic modulation and coding methodologyeach particle 119909

119894is in fact a 2-dimensional vector

119909119894= [SBR 119875

119905] isin 1198772

(19)

where 119894 isin [1 2 119898] In order to avoid premature con-vergence to local optima Logistic map is used to generatechaotic sequence

119911119895+1

= 120583119911119895(1 minus 119911

119895) 119895 = 1 2 119904 (20)

where 120583 is the control parameter The chaotic sequence isdistributed in [0 1] The iterative Logistic map can generatea chaotic sequence 119911

1 1199112 119911

119904through assigning an initial


value 1199111 Assume that the solution space is [Γlower Γupper]

where these two vectors are depicted by

Γlower

= [SBRlower 119875

lower119905

]

Γupper

= [SBRupper 119875

upper119905

]

(21)

so that the sequence is mapped to the solution space by thefollowing formula

119909119894= Γ

lower+ (Γ

upperminus Γ

lower) sdot 119911119894 119894 = 1 2 119898 (22)

Then in order to find an optimal solution each particle 119909119894

evolves in the search space119863 based on the following positionupdating algorithm

119909119896+1

119894= 119909119896

119894+ V119896+1119894

V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896

swarm minus 119909119896

119894)

(23)

where 1198881is the cognitive scaling factor 119888

2is the social scaling

factor and 119908 is a positive parameter called inertia weightwhich is introduced to ensure convergence of the particlesrsquosearch process

119908 = 119908min +119908max minus 119908min

119896 (24)

where119908min and119908max are the upper limit and lower limit of119908When119908 is dramatically attenuated the convergence of searchprocess will accelerate The random numbers 120585119896

119894and 120577

119896

119894are

uniformly distributed in [0 1] and represent the stochasticbehaviors of the CPSO 119901119887119890119904119905119896

119894 which denotes the previously

obtained position of the 119894th particle is defined as

119901119887119890119904119905119896

119894= argmin

119909119895

119894

119891 (119909119895

119894) 0 le 119895 le 119896 (25)

On the other hand 119892119887119890119904119905119896swarm which denotes the bestposition in the entire search space at the current iteration 119896is defined as

119892119887119890119904119905119896

swarm = argmin119909119896

119894

119891 (119909119896

119894) forall119894 (26)

Hence the term 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus119909119896

119894) is associatedwith cognition

since it takes into account only the best position of the par-ticlersquos own experience while 119888

2120577119896

119894(119892119887119890119904119905

119896


119894) represents

the social interaction of all particlesParticles will search iteratively in terms of fitness in the

search space until the user-defined termination criterionsare satisfied Initial particles may be selected from chaoticsequences and initial velocity could be randomly generatedThe CPSO method is presented in Algorithm 1

Algorithm 1 (CPSO algorithm)

(1) Set 119896 = 1 and set maximum iterations 119896max 119908 =

119908max

(2) Initialize particle 1199091119894and its corresponding updating

velocity V1119894(119894 = 1 2 119898) and randomly generate a

2-dimensional vector [11991111 11991112] in the range of [0 1]

According to formula (20) 119904 chaotic vectors could beobtained and map these vectors to the solution space[Γ

lower Γ

upper] by formula (22) Then select 119898 vectors

with better fitness(3) Consider 119901119887119890119904119905

1

119894= 119909

1

119894 119892119887119890119904119905

1

swarm =

argmin1199091

119894

119891(1199091

119894) forall119894

(4) While (119896 le 119896max)(5) Update velocity of particle

V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(27)

(6) Update position of particle 119909119896+1119894

= 119909119896

119894+ V119896+1119894

(7) Calculate the fitness value 119891(119909119894)

(8) Consider 119896 = 119896 + 1 119908 = 119908min + (119908max minus 119908min)119896

(9) Consider 119901119887119890119904119905119896119894= argmin

119909119895

119894

119891(119909119895

119894) 0 le 119895 le 119896

(10) Consider 119892119887119890119904119905119896swarm = argmin119909119896

119894

119891(119909119896

119894) forall119894

(11) Optimize the global position by chaotic sequenceGenerate chaotic vector sequence 119892

1 1198922 119892

119904in

[0 1] and map it to the solution space [Γlower Γupper]

(12) Consider 119892119887119890119904119905119896119895= 119892119887119890119904119905

119896

swarm + 119903(119892119895minus 05) 119895 =

1 2 119904 119903 being the search radius

(13) If (119891(119892119887119890119904119905119896119895) lt 119891(119892119887119890119904119905

119896

swarm) 119895 = 1 2 119904) updatethe best position 119892119887119890119904119905119896swarm = 119892119887119890119904119905

119896

119895

(14) End if(15) End while

6 Simulation Results and Analysis

In this section we evaluate the performance through MAT-LAB In order to conveniently validate the proposed opti-mization algorithm three video sequences (Suzie carphoneand football) that come fromLIVE database [34] with aQCIF(176 times 144) frame size are chosen and the algorithm alsoapplies to other high-resolution videos Let video encodingrate range from 50 kbps to 150 kbps with a 2 kbps step Thewireless channel is set as the frequency selective multipathchannel that consists of six independent Rayleigh pathswith an exponentially fading profile In addition the videostream is transmitted over UDP to guarantee the real timeIn the following the performance of the proposed QP-CEEmethod the tradeoff of QoE and energy consumption andthe quantitative convergence are analyzed

61 Performance Analysis of QP-CEE Method According tothe ITU-T criteria [40]MOS are divided into five gradesThehigher value denotes the better perceptive video quality In


0 10 20 30 40 50 60 70 80 90 100055

06

065

07

075

08

085

09

095

1 Suzie

Frame

Nor

mal

ized

pow

er

QP-CEE modelMax SBRMax Pt

(a) Suzie

0 10 20 30 40 50 60 70 80 90 100075

08

085

09

095

1 Carphone

Frame

Nor

mal

ized

pow

er


(b) Carphone

Nor

mal

ized

pow

er

0 10 20 30 40 50 60 70 80 90 10004

045

05

055

06

065

07

075

08

085

09 Football

Frame


(c) Football

Figure 6 Energy consumption under different energy efficient methods

this paper we adopt 35 as the threshold of acceptable videoquality for users which is

MOSth = 35 (28)

Therefore the QoE constraint for QP-CEE method couldbe denoted by

MOS ge MOSth = 35 (29)

To demonstrate the performance of the proposed QP-CEE method we compare it with two traditional distortionbased energy efficientmethods max SBRmethod andmax PtmethodMax SBRmethod adoptsmaximumencoding bitrateand adaptive transmitting power level while max Pt methodemploys maximum transmitting power level and adaptiveencoding bitrate

The normalized energy consumption and MOS of testvideo sequences using differentmethods are shown in Figures6 and 7 respectively As the alternative energy consumption


0 10 20 30 40 50 60 70 80 90 10035

36

37

38

39

4

41

42

43

44

45 Suzie

Frame

MO

S


(a) Suzie

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Carphone

Frame

MO

S


(b) Carphone

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Football

Frame

MO

S


(c) Football

Figure 7 MOS value under different energy efficient methods

levels and encoding bitrate are discrete values rather thancontinuous values the suboptimal video encoding bitrate andtransmitting power level could be figured out by our proposedQoE perceptive cross-layer energy efficiency method

Figure 6 shows the minimum energy consumption ofvideo frames Here the energy consumption is normalizedby the maximum energy consumption In Figure 6 wecompared the energy consumption by different strategiesThrough calculating the minimum energy consumption ofvideo encoding and transmitting in mobile video devices

the proposed QP-CEE method can effectively reduce energyconsumption and extend the battery lifetime

Figure 7 depicts the MOS values of video frames usingdifferent methods As Figure 7 shows on the basis of non-invasive QoE perception model of video streaming theproposed cross-layer energy efficient method guarantees theQoE of mobile video services

The average energy consumption andMOSofThree videoclips are shown in Table 4 As different test video clips varyin video content and structure the average values of energy


Table 4 Average energy consumption and MOS

Video 119901 mosSuzie 0652 389Carphone 0771 371Football 0551 353

consumption and MOS achieved by energy efficient methodare distinct from each other Obviously our proposed QP-CEE method reduces the energy consumption of mobilevideo devices while maintaining the perceptive quality ofvideo applications surpassing the threshold

62 The QoE-Energy Consumption Tradeoff We have eval-uated the performance of the proposed method with agiven MOS value through the above analysis In order toanalyze the global performance with dynamic MOS valuesof our proposed cross-layer energy efficient method tradeoffbetween energy consumption and QoE in mobile videodevices is studied in this subsection Based on the proposedenergy consumption model and QoE perception modelPareto fronts of energy consumption andQoE are figured out

(119875MOS)|optimal


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(30)

Figure 8 depicts the Pareto front curve of QoE and energyconsumption of three video sequences respectively It showsthat with the increase of energy consumption MOS valueincreases as well For videos with different content type thesame power consumption results in different MOS valuesSince more complex scene and motion lead to more energyconsumption in encoding and transmitting to obtain thesame perceptive quality video sequences with more complexscene and motion will consume more energy Hence forthe same energy consumption video sequence with complexscene and increased movement in content will show lessMOS

We compared theMOS-energy tradeoff curves of the pro-posed QP-CEE method with power-rate-distortion (P-R-D)[17] method in Figure 9 P-R-D method is an energy efficientmethod developed on the basis of P-R-Dmodel which is reg-ularly used as the video energy efficient methods In Figure 9average energy consumption and MOS of these three videosequences are calculated to depict the tradeoff curves It is nothard to learn that for the same energy consumption theMOSvalue of the proposed QP-CEE method is higher than P-R-Dmethod For the same MOS value the energy consumptionof the proposed QP-CEE method is less than P-R-D methodTherefore Figure 9 shows that the energy efficient methodsfrom the viewpoint of QoE lead to better performance gainsin mobile video devices compared with those methods fromthe viewpoint of video distortion Meanwhile these twocurves are increasing closer with the increase of energyconsumption It means that when the target MOS valuesgradually reach the available upper limit the possible energyefficient gains of QP-CEE method are diminishing

065 07 075 08 085 09 095 132

34

36

38

4

42

44

46

Normalized energy

MO

S

SuzieCarphoneFootball

Figure 8 Pareto front of QoE and energy consumption for Suziecarphone and football

065 07 075 08 085 09 095 132

33

34

35

36

37

38

39

4

Normalized energy

MO

S

QP-CEE modelP-R-D model

Figure 9 Pareto front of P-R-Dmodel and QP-CEEmodel in videofor carphone

63 Quantitative Convergence Analysis Since our proposedQP-CEEmethod is designed to operate onmobile devices forreal-time video streaming requirements for convergence rateare very strict Hence the search process of particles shouldsteadily converge to a suboptimal solution rapidly

Chaotic particle swarmoptimization algorithmhas excel-lent convergence properties Figure 10 shows the convergenceof CPSO in our evaluation and for various types of testvideo sequences it can steadily converge to a suboptimalsolution after about 48ndash58 times iteration with small vari-ance As nowadays mobile video devices are increasinglysmart and intelligent these iteration calculations could be


0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

Frame

Con

verg

ence


Figure 10 Convergence of chaotic particle swarm optimizationalgorithm

accomplished quickly Therefore the real-time requirementscould be achieved

7 Conclusion

We proposed a QoE perceptive cross-layer energy efficientmethod formobile video devices in wireless sensor networksInfluences on energy consumption andQoE of video servicesby video encoding bitrate and transmitting power level areanalyzed respectively The energy consumption model andQoE perceptivemodel formobile video devices are proposedby which a cross-layer optimization problem is formulatedand chaos particle swarm optimization is adopted to solvethis optimization problem with fast convergence propertyIn addition the energy consumption minimization problemrestrained by QoE and the tradeoff problem between QoEand energy consumption are evaluated and analyzed in thispaper which demonstrates the performance of the proposedenergy efficient method By the proposed QP-CEE methodappropriate tradeoff between QoE and energy consumptionis figured out and the search process of particles steadilyconverges to a suboptimal solution rapidly which can satisfythe requirements of real-time video applications

We propose a joint optimization framework includingmodel construction and optimization problem solving in thispaper which can be applied to different types of networksand different resolutions of video services more than UMTSand QCIF Hence in future work QoE perception modelscorresponding to other network types such as 4G5G andvideo resolutions such as CIFVGAWVGAUHD may beapplied to this framework

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work is supported by the National Natural ScienceFoundation of China (61271179) Beijing Municipal Sci-ence and Technology Commission Research Fund Project(D151100000115002) and the Fundamental Research Fundsfor the Central Universities (2014ZD03-01)

References

[1] Ericsson ldquoEricsson Mobility Reportrdquo 2013[2] F Banterle A Artusi K Debattista andA ChalmersAdvanced

High Dynamic Range Imaging Theory and Practice A K PetersCRC Press Natrick Mass USA 2011

[3] Y Ye Y He and X Xiu ldquoManipulating ultra-high definitionvideo trafficrdquo IEEE MultiMedia vol 22 no 3 pp 73ndash81 2015

[4] CISCO Visual Networking Index (VNI) Forecast CISCO 2013[5] S Sharangi R Krishnamurti andMHefeeda ldquoEnergy-efficient

multicasting of scalable video streams over WiMAX networksrdquoIEEETransactions onMultimedia vol 13 no 1 pp 102ndash115 2011

[6] E Belyaev K Egiazarian and M Gabbouj ldquoA low-complexitybit-plane entropy coding and rate control for 3-D DWT basedvideo codingrdquo IEEE Transactions on Multimedia vol 15 no 8pp 1786ndash1799 2013

[7] X Li M Wien and J-R Ohm ldquoRate-complexity-distortionoptimization for hybrid video codingrdquo IEEE Transactions onCircuits and Systems for Video Technology vol 21 no 7 pp 957ndash970 2011

[8] J Kim and C-M Kyung ldquoA lossless embedded compressionusing significant bit truncation for hd video codingrdquo IEEETransactions on Circuits and Systems for Video Technology vol20 no 6 pp 848ndash860 2010

[9] Z Sun X Chen and Z He ldquoAdaptive critic design for energyminimization of portable video communication devicesrdquo IEEETransactions on Circuits and Systems for Video Technology vol20 no 1 pp 27ndash37 2010

[10] M F Sabir A C Bovik and RW Heath ldquoUnequal power allo-cation for JPEG transmission overMIMO systemsrdquo IEEETrans-actions on Image Processing vol 19 no 2 pp 410ndash421 2010

[11] X Lu Y Wang and E Erkip ldquoPower efficient H263 videotransmission over wireless channelsrdquo in Proceedings of theInternational Conference on Image Processing (ICIP rsquo02) pp533ndash536 Rochester NY USA September 2002

[12] D G Sachs S V Adve and D L Jones ldquoCross-layer adaptivevideo coding to reduce energy on general-purpose processorsrdquoin Proceedings of the International Conference on Image Process-ing (ICIP rsquo03) pp 109ndash112 Barcelona Spain September 2003

[13] C-M Chen C-W Lin and Y-C Chen ldquoCross-layer packetretry limit adaptation for video transport over wireless LANsrdquoIEEE Transactions on Circuits and Systems for Video Technologyvol 20 no 11 pp 1448ndash1461 2010

[14] C Li J Zou H Xiong and C W Chen ldquoJoint codingroutingoptimization for distributed video sources in wireless visualsensor networksrdquo IEEE Transactions on Circuits and Systems forVideo Technology vol 21 no 2 pp 141ndash155 2011

[15] DNguyen T Tran T Nguyen and B Bose ldquoWireless broadcastusing network codingrdquo IEEE Transactions onVehicular Technol-ogy vol 58 no 2 pp 914ndash925 2009

[16] Z He Y Liang L Chen I Ahmad andDWu ldquoPower-rate-dis-tortion analysis forwireless video communication under energy


constraintsrdquo IEEE Transactions on Circuits and Systems forVideo Technology vol 15 no 5 pp 645ndash658 2005

[17] Z He W Cheng and X Chen ldquoEnergy minimization of port-able video communication devices based on power-rate-distor-tion optimizationrdquo IEEE Transactions on Circuits and Systemsfor Video Technology vol 18 no 5 pp 596ndash607 2008

[18] C-H Hsu and M Hefeeda ldquoMulti-layer video broadcastingwith low channel switching delaysrdquo in Proceedings of the 17thInternational Packet Video Workshop (PV rsquo09) pp 1ndash10 IEEESeattle Wash USA May 2009

[19] C H Hsu and M Hefeeda ldquoVideo broadcasting to heteroge-neous mobile devicesrdquo in Networking 2009 vol 5550 pp 600ndash613 Springer Berlin Germany 2009

[20] C-H Hsu and M Hefeeda ldquoFlexible broadcasting of scalablevideo streams to heterogeneous mobile devicesrdquo IEEE Transac-tions on Mobile Computing vol 10 no 3 pp 406ndash418 2011

[21] H Sohn H Yoo W De Neve C S Kim and Y M Ro ldquoFull-reference video qualitymetric for fully scalable andmobile SVCcontentrdquo IEEE Transactions on Broadcasting vol 56 no 3 pp269ndash280 2010

[22] S Chikkerur V Sundaram M Reisslein and L J KaramldquoObjective video quality assessment methods a classificationreview and performance comparisonrdquo IEEE Transactions onBroadcasting vol 57 no 2 pp 165ndash182 2011

[23] MNarwariaW Lin andA Liu ldquoLow-complexity video qualityassessment using temporal quality variationsrdquo IEEE Trans-actions on Multimedia vol 14 no 3 pp 525ndash535 2012

[24] K Seshadrinathan and A C Bovik ldquoMotion tuned spatio-tem-poral quality assessment of natural videosrdquo IEEE Transactionson Image Processing vol 19 no 2 pp 335ndash350 2010

[25] S Paulikas ldquoEstimation of video quality of H264AVC videostreamingrdquo in Proceedings of the IEEE International Conferenceon Computer As a Tool (EUROCON rsquo13) pp 694ndash700 IEEEZagreb Croatia July 2013

[26] K-C Yang C C Guest K El-Maleh and P K Das ldquoPerceptualtemporal quality metric for compressed videordquo IEEE Transac-tions on Multimedia vol 9 no 7 pp 1528ndash1535 2007

[27] S Tao J Apostolopoulos andRGuerin ldquoReal-timemonitoringof video quality in IP networksrdquo IEEEACM Transactions onNetworking vol 16 no 5 pp 1052ndash1065 2008

[28] A K Moorthy K Seshadrinathan R Soundararajan and A CBovik ldquoWireless video quality assessment a study of subjectivescores and objective algorithmsrdquo IEEE Transactions on Circuitsand Systems for Video Technology vol 20 no 4 pp 587ndash5992010

[29] K Seshadrinathan R Soundararajan A C Bovik and L KCormack ldquoStudy of subjective and objective quality assessmentof videordquo IEEE Transactions on Image Processing vol 19 no 6pp 1427ndash1441 2010

[30] A Khan L Sun and E Ifeachor ldquoQoE predictionmodel and itsapplication in video quality adaptation over UMTS networksrdquoIEEE Transactions on Multimedia vol 14 no 2 pp 431ndash4422012

[31] F Liberal I Taboada and J-O Fajardo ldquoDealing with energy-QoE trade-offs in mobile videordquo Journal of Computer Networksand Communications vol 2013 Article ID 412491 12 pages2013

[32] X Zhang J Zhang Y Huang and W Wang ldquoOn the study offundamental trade-offs between QoE and energy efficiency inwireless networksrdquo European Transactions on Telecommunica-tions vol 24 no 3 pp 259ndash265 2013

[33] Y Lang Research of WCDMA handset power consumptionmechanism and prediction model based on neural network[MS thesis] Department of Electronic Engineering BeijingTechnology and Business University Beijing China 2010

[34] A K Moorthy K Seshadrinathan R Soundararajan and A CBovik LIVEWireless Video Quality Assessment Database 2009httpliveeceutexaseduresearchqualitylive wireless videohtml

[35] ITU ldquoMethodology for the subjective assessment of videoquality in multimedia applicationsrdquo ITU-RecommendationBT1788 ITU Geneva Switzerland 2007

[36] I Taboada J O Fajardo and F Liberal ldquoQoE and energy-awareness for multi-layer video broadcastingrdquo in Proceedings ofthe 12th IEEE International Conference onMultimedia and Expo(ICME rsquo11) pp 1ndash6 IEEE Barcelona Spain July 2011

[37] H Luo D Wu S Ci H Sharif and H Tang ldquoTFRC-basedrate control for real-time video streaming over wireless multi-hop mesh networksrdquo in Proceedings of the IEEE InternationalConference on Communications (ICC rsquo09) June 2009

[38] R C Eberhart and J Kennedy ldquoA new optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium on Micro Machine and Human Science (MHS rsquo95) pp 39ndash43 IEEE Nagoya Japan October 1995

[39] T-H Kim I Maruta and T Sugie ldquoA simple and efficientconstrained particle swarm optimization and its application toengineering design problemsrdquo Proceedings of the Institution ofMechanical Engineers Part C Journal ofMechanical EngineeringScience vol 224 no 2 pp 389ndash400 2010

[40] International Telecommunication Union ldquoSubjective videoquality assessment methods for multimedia applicationsrdquo Rec-ommendation ITU-T P910 International TelecommunicationUnion Geneva Switzerland 2008

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



method should be proposed to reduce the energy consump-tion of video streaming and extend the lifetime of batterysignificantly Therefore as there is no significant advancein battery technology energy efficient mechanism for videoencoding and transmitting becomes a crucial issue to beinvestigated

Secondly wireless multimedia sensor networks deeplychange the interactive mode between cyber space andphysical space which have been applied in areas such asmilitary health care traffic control and disaster resistanceMultimedia sensors such as video sensors and image sensorsare engaged in perception processing and transmission notonly for simple data information but also for multimediainformation However dynamic and unpredicted interfer-ence in wireless sensor networks will make a big dent inquality of video services In addition video service is a kindof service with strong content correlation QoS metrics suchas throughput delay and jitter cannot fully reflect the userperceived quality of video service QoE which depicts theapproximate subjective experience of services in wirelesssensor networks by terminal users is increasingly adopted asthe metric for video quality With regard to video services inwireless sensor networks how to dynamically accommodatevideo encoding scheme and manage radio resources so thatQoE of mobile video services could be assured is an issue thaturgently needs to be addressed

Therefore in wireless sensor networks video devicesneed not only to assure the QoE for video applications butalso to deal with the huge energy consumption generated byhigh data rate of video streaming Appropriate strategy profileshould be picked out to guarantee the QoE of video applica-tions while minimizing the energy consumption of mobilevideo devices However there are conflicts between these twoobjectives Reducing energy consumption of video streamingis always in the price of degradation of transmitting powerlevel which will lead to degradation of the QoE for video ser-vices Meanwhile enhancement of QoE is always in the priceof enhancement of transmitting power level which will leadto extra energy consumption of video services Therefore itis necessary to find an appropriate tradeoff between energysaving and QoE so as to reduce the energy consumptioncaused by video streaming with assurance of QoE

In this paper as shown in Figure 1 we propose a QoEperceptive cross-layer energy efficient (QP-CEE) method formobile video devices to dispose of the energy efficiencyproblem and QoE assurance problem together Energy con-sumption of mobile devices is reduced and QoE of videoapplications in mobile video device is enhanced throughdynamic regulation of transmitting power level modulationand coding scheme (MCS) and encoding bitrate of videostreaming By elaborative analysis of the encoding andtransmitting process ofmobile video device a comprehensiveenergy consumption model for video encoding and trans-mitting is established A noninvasive QoE perception modelis adopted in terms of the mean opinion score (MOS) Theproposed QoE perception model is significant in its ownright as the optimization of QoE is crucial for mobile videoservices Energy efficiency problem ismodeled as a joint opti-mization problem including optimization of QoE and energy

Encodingenergy

consumption

Transmittingenergy

consumption

Energyconsumption

model for videodevices

Encodingbitrate

Transmittingpower level

QoE perceptionmodel

Cross-layeroptimization

Suboptimalstrategies

CPSO

MCS

Figure 1 Framework of QoE perceptive cross-layer energy effi-ciency method for video devices

consumption Then chaos particle swarm optimization algo-rithm is adopted to search for optimal power level and videoencoding rate in parallel so that the energy consumption isminimized and QoE of video applications is ensured FinallyPareto front of energy and QoE is analyzed to certify theperformance of the proposed method There are three maincontributions in this paper Firstly encoding bitrate modu-lation and coding scheme and transmitting power level areinvestigated simultaneously by which multilayer (includingapplication layer and physical layer) performance gain isachieved Secondly an energy efficient method based onQoEperception is proposed and both requirements ofmobile userfor video quality and minimization of energy consumptionformobile devices are achievedThirdly chaos particle swarmoptimization is adopted to solve the optimization problem Inone aspect the search process is not apt to converge to localoptimum solution In the other aspect time delay for solvingthe cross-layer optimization problem can be shortened byparallel process which can meet the real-time requirementsfor video services such as video live and video conference

The rest of this paper is arranged as follows In Section 2state-of-the-art research about QoE perception methodenergy saving strategy tradeoff method of QoE and energysaving is surveyed and summarized and the open questionsand possible research directions are analyzed In Section 3the uplink energy efficient scenario for mobile video devicesis described The energy consumption model for mobilevideo devices is proposed in Section 4 In Section 5 jointoptimization problem of QoE and energy consumption isinvestigated and chaos particle swarm optimization methodis adopted to find optimum encoding bitrate and powerlevel Evaluation and analysis for our work are carried out inSection 6 In Section 7 we conclude our work in this paper

2 Related Work

Various energy efficient methods have been proposed toimprove the energy efficiency of wireless networks Spe-cially for video streaming with growing popularity the largeenergy consumption generated by encoding and transmitting















033 047 060 073 087 10

Video sequen

ce






1

095

09

085

08

075

Nor

mal

ized

ener

gy

50 60 70 80 90 100 110 120 130 140 150


43

425

42

415

41

405

MO

S


05 055 06 065 07 075 08 085 09 095 105

1

15

Nor

mal

ized

ener

gy


423

424

425

MO

S











119904and119875119905as


119875 = 119875119904+ 119875119905 (1)



119863(SBR 119875119904) = 1205902

2minus120582SBRsdot119892(119875

119904)

0 le 119875 le 1 (2)


119904)


119892 (119875119904) = 1198751120574

119904 (3)



119894|



119894 respectively

SBR119894and 119901




119901119894= (

1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

(4)


119875119904= sum

119894

119901119894= sum

119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

(5)


119875 = 119875119904+ 119875119905= sum

119894

119901119894+ 119875119905

= sum

119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

+ 119875119905

(6)



119899))119871

(7)





BER (120574119899) =

119886119898

119890120574119899times119887119898 (8)


119898and 119887

119898are shown in



120574119899=119901119905sdot1003816100381610038161003816ℎ119899

1003816100381610038161003816

2

1198730119861119873

(9)


0is the spectral




AMC mode (119898) 119898 = 1 119898 = 2 119898 = 3 119898 = 4 119898 = 5 119898 = 6


119898) 12 12 34 916 34 34

119877119898(bitssym) 050 100 150 225 300 450

119886119898

11369 03351 02197 02081 01936 01887119887119898

75556 32543 15244 06250 03484 00871



119899))119871

= 119877119888sdot (1 minus

119886119898

119890120574119899times119887119898)

119871

= 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(10)



That is

SBR119894le 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(12)








120572 120573 120574 120575 120578 120590

39560 00919 minus58497 09844 01028 minus00236




min 119875 = min119891 (SBR 119875119905MCS)

= min(sum119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

+ 119875119905)


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

MOS ge MOSth

(14)







Parameter selection

Feedback

Parameter selection



Physical layer










119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

le 0

MOSth minusMOS le 0

(15)

Then define

ℎ (SBR 119875119905MCS) = max (SBR

119894minus 119877119888


)(1198730119861119873))times119887

119898)

119871

MOSth

minusMOS)

(16)


119891 (SBR 119875119905MCS) = ℎ (SBR 119875

119905MCS)

ℎ (SBR 119875119905MCS) gt 0

119891 (SBR 119875119905MCS) = 119886 tan [119891 (SBR 119875

119905MCS)] minus 120587

2

others

(17)






min119883

119891 (SBR 119875119905MCS) 119883 isin 119863 sube 119877

3

(18)


1 1199092




119909119894= [SBR 119875

119905] isin 1198772

(19)


119911119895+1

= 120583119911119895(1 minus 119911

119895) 119895 = 1 2 119904 (20)


1 1199112 119911





Γlower

= [SBRlower 119875

lower119905

]

Γupper

= [SBRupper 119875

upper119905

]

(21)


119909119894= Γ

lower+ (Γ

upperminus Γ

lower) sdot 119911119894 119894 = 1 2 119898 (22)



119909119896+1

119894= 119909119896

119894+ V119896+1119894

V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(23)




119908 = 119908min +119908max minus 119908min

119896 (24)


119894and 120577

119896

119894are




119901119887119890119904119905119896

119894= argmin

119909119895

119894

119891 (119909119895

119894) 0 le 119895 le 119896 (25)


119892119887119890119904119905119896


119894

119891 (119909119896

119894) forall119894 (26)

Hence the term 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus119909119896



2120577119896

119894(119892119887119890119904119905

119896


119894) represents





119908max





lower Γ



1

119894= 119909

1

119894 119892119887119890119904119905

1

swarm =

argmin1199091

119894

119891(1199091

119894) forall119894


V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(27)


= 119909119896

119894+ V119896+1119894



(9) Consider 119901119887119890119904119905119896119894= argmin

119909119895

119894

119891(119909119895

119894) 0 le 119895 le 119896


119894

119891(119909119896

119894) forall119894


1 1198922 119892

119904in


(12) Consider 119892119887119890119904119905119896119895= 119892119887119890119904119905

119896

swarm + 119903(119892119895minus 05) 119895 =


(13) If (119891(119892119887119890119904119905119896119895) lt 119891(119892119887119890119904119905

119896


119896

119895






0 10 20 30 40 50 60 70 80 90 100055

06

065

07

075

08

085

09

095

1 Suzie

Frame

Nor

mal

ized

pow

er


(a) Suzie

0 10 20 30 40 50 60 70 80 90 100075

08

085

09

095

1 Carphone

Frame

Nor

mal

ized

pow

er


(b) Carphone

Nor

mal

ized

pow

er

0 10 20 30 40 50 60 70 80 90 10004

045

05

055

06

065

07

075

08

085

09 Football

Frame


(c) Football



MOSth = 35 (28)






0 10 20 30 40 50 60 70 80 90 10035

36

37

38

39

4

41

42

43

44

45 Suzie

Frame

MO

S


(a) Suzie

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Carphone

Frame

MO

S


(b) Carphone

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Football

Frame

MO

S


(c) Football












(119875MOS)|optimal


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(30)



065 07 075 08 085 09 095 132

34

36

38

4

42

44

46

Normalized energy

MO

S



065 07 075 08 085 09 095 132

33

34

35

36

37

38

39

4

Normalized energy

MO

S






0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

Frame

Con

verg

ence




7 Conclusion





Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation























033 047 060 073 087 10

Video sequen

ce






1

095

09

085

08

075

Nor

mal

ized

ener

gy

50 60 70 80 90 100 110 120 130 140 150


43

425

42

415

41

405

MO

S


05 055 06 065 07 075 08 085 09 095 105

1

15

Nor

mal

ized

ener

gy


423

424

425

MO

S











119904and119875119905as


119875 = 119875119904+ 119875119905 (1)



119863(SBR 119875119904) = 1205902

2minus120582SBRsdot119892(119875

119904)

0 le 119875 le 1 (2)


119904)


119892 (119875119904) = 1198751120574

119904 (3)



119894|



119894 respectively

SBR119894and 119901




119901119894= (

1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

(4)


119875119904= sum

119894

119901119894= sum

119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

(5)


119875 = 119875119904+ 119875119905= sum

119894

119901119894+ 119875119905

= sum

119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

+ 119875119905

(6)



119899))119871

(7)





BER (120574119899) =

119886119898

119890120574119899times119887119898 (8)


119898and 119887

119898are shown in



120574119899=119901119905sdot1003816100381610038161003816ℎ119899

1003816100381610038161003816

2

1198730119861119873

(9)


0is the spectral




AMC mode (119898) 119898 = 1 119898 = 2 119898 = 3 119898 = 4 119898 = 5 119898 = 6


119898) 12 12 34 916 34 34

119877119898(bitssym) 050 100 150 225 300 450

119886119898

11369 03351 02197 02081 01936 01887119887119898

75556 32543 15244 06250 03484 00871



119899))119871

= 119877119888sdot (1 minus

119886119898

119890120574119899times119887119898)

119871

= 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(10)



That is

SBR119894le 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(12)








120572 120573 120574 120575 120578 120590

39560 00919 minus58497 09844 01028 minus00236




min 119875 = min119891 (SBR 119875119905MCS)

= min(sum119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

+ 119875119905)


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

MOS ge MOSth

(14)







Parameter selection

Feedback

Parameter selection



Physical layer










119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

le 0

MOSth minusMOS le 0

(15)

Then define

ℎ (SBR 119875119905MCS) = max (SBR

119894minus 119877119888


)(1198730119861119873))times119887

119898)

119871

MOSth

minusMOS)

(16)


119891 (SBR 119875119905MCS) = ℎ (SBR 119875

119905MCS)

ℎ (SBR 119875119905MCS) gt 0

119891 (SBR 119875119905MCS) = 119886 tan [119891 (SBR 119875

119905MCS)] minus 120587

2

others

(17)






min119883

119891 (SBR 119875119905MCS) 119883 isin 119863 sube 119877

3

(18)


1 1199092




119909119894= [SBR 119875

119905] isin 1198772

(19)


119911119895+1

= 120583119911119895(1 minus 119911

119895) 119895 = 1 2 119904 (20)


1 1199112 119911





Γlower

= [SBRlower 119875

lower119905

]

Γupper

= [SBRupper 119875

upper119905

]

(21)


119909119894= Γ

lower+ (Γ

upperminus Γ

lower) sdot 119911119894 119894 = 1 2 119898 (22)



119909119896+1

119894= 119909119896

119894+ V119896+1119894

V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(23)




119908 = 119908min +119908max minus 119908min

119896 (24)


119894and 120577

119896

119894are




119901119887119890119904119905119896

119894= argmin

119909119895

119894

119891 (119909119895

119894) 0 le 119895 le 119896 (25)


119892119887119890119904119905119896


119894

119891 (119909119896

119894) forall119894 (26)

Hence the term 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus119909119896



2120577119896

119894(119892119887119890119904119905

119896


119894) represents





119908max





lower Γ



1

119894= 119909

1

119894 119892119887119890119904119905

1

swarm =

argmin1199091

119894

119891(1199091

119894) forall119894


V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(27)


= 119909119896

119894+ V119896+1119894



(9) Consider 119901119887119890119904119905119896119894= argmin

119909119895

119894

119891(119909119895

119894) 0 le 119895 le 119896


119894

119891(119909119896

119894) forall119894


1 1198922 119892

119904in


(12) Consider 119892119887119890119904119905119896119895= 119892119887119890119904119905

119896

swarm + 119903(119892119895minus 05) 119895 =


(13) If (119891(119892119887119890119904119905119896119895) lt 119891(119892119887119890119904119905

119896


119896

119895






0 10 20 30 40 50 60 70 80 90 100055

06

065

07

075

08

085

09

095

1 Suzie

Frame

Nor

mal

ized

pow

er


(a) Suzie

0 10 20 30 40 50 60 70 80 90 100075

08

085

09

095

1 Carphone

Frame

Nor

mal

ized

pow

er


(b) Carphone

Nor

mal

ized

pow

er

0 10 20 30 40 50 60 70 80 90 10004

045

05

055

06

065

07

075

08

085

09 Football

Frame


(c) Football



MOSth = 35 (28)






0 10 20 30 40 50 60 70 80 90 10035

36

37

38

39

4

41

42

43

44

45 Suzie

Frame

MO

S


(a) Suzie

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Carphone

Frame

MO

S


(b) Carphone

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Football

Frame

MO

S


(c) Football












(119875MOS)|optimal


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(30)



065 07 075 08 085 09 095 132

34

36

38

4

42

44

46

Normalized energy

MO

S



065 07 075 08 085 09 095 132

33

34

35

36

37

38

39

4

Normalized energy

MO

S






0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

Frame

Con

verg

ence




7 Conclusion





Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













033 047 060 073 087 10

Video sequen

ce






1

095

09

085

08

075

Nor

mal

ized

ener

gy

50 60 70 80 90 100 110 120 130 140 150


43

425

42

415

41

405

MO

S


05 055 06 065 07 075 08 085 09 095 105

1

15

Nor

mal

ized

ener

gy


423

424

425

MO

S











119904and119875119905as


119875 = 119875119904+ 119875119905 (1)



119863(SBR 119875119904) = 1205902

2minus120582SBRsdot119892(119875

119904)

0 le 119875 le 1 (2)


119904)


119892 (119875119904) = 1198751120574

119904 (3)



119894|



119894 respectively

SBR119894and 119901




119901119894= (

1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

(4)


119875119904= sum

119894

119901119894= sum

119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

(5)


119875 = 119875119904+ 119875119905= sum

119894

119901119894+ 119875119905

= sum

119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

+ 119875119905

(6)



119899))119871

(7)





BER (120574119899) =

119886119898

119890120574119899times119887119898 (8)


119898and 119887

119898are shown in



120574119899=119901119905sdot1003816100381610038161003816ℎ119899

1003816100381610038161003816

2

1198730119861119873

(9)


0is the spectral




AMC mode (119898) 119898 = 1 119898 = 2 119898 = 3 119898 = 4 119898 = 5 119898 = 6


119898) 12 12 34 916 34 34

119877119898(bitssym) 050 100 150 225 300 450

119886119898

11369 03351 02197 02081 01936 01887119887119898

75556 32543 15244 06250 03484 00871



119899))119871

= 119877119888sdot (1 minus

119886119898

119890120574119899times119887119898)

119871

= 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(10)



That is

SBR119894le 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(12)








120572 120573 120574 120575 120578 120590

39560 00919 minus58497 09844 01028 minus00236




min 119875 = min119891 (SBR 119875119905MCS)

= min(sum119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

+ 119875119905)


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

MOS ge MOSth

(14)







Parameter selection

Feedback

Parameter selection



Physical layer










119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

le 0

MOSth minusMOS le 0

(15)

Then define

ℎ (SBR 119875119905MCS) = max (SBR

119894minus 119877119888


)(1198730119861119873))times119887

119898)

119871

MOSth

minusMOS)

(16)


119891 (SBR 119875119905MCS) = ℎ (SBR 119875

119905MCS)

ℎ (SBR 119875119905MCS) gt 0

119891 (SBR 119875119905MCS) = 119886 tan [119891 (SBR 119875

119905MCS)] minus 120587

2

others

(17)






min119883

119891 (SBR 119875119905MCS) 119883 isin 119863 sube 119877

3

(18)


1 1199092




119909119894= [SBR 119875

119905] isin 1198772

(19)


119911119895+1

= 120583119911119895(1 minus 119911

119895) 119895 = 1 2 119904 (20)


1 1199112 119911





Γlower

= [SBRlower 119875

lower119905

]

Γupper

= [SBRupper 119875

upper119905

]

(21)


119909119894= Γ

lower+ (Γ

upperminus Γ

lower) sdot 119911119894 119894 = 1 2 119898 (22)



119909119896+1

119894= 119909119896

119894+ V119896+1119894

V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(23)




119908 = 119908min +119908max minus 119908min

119896 (24)


119894and 120577

119896

119894are




119901119887119890119904119905119896

119894= argmin

119909119895

119894

119891 (119909119895

119894) 0 le 119895 le 119896 (25)


119892119887119890119904119905119896


119894

119891 (119909119896

119894) forall119894 (26)

Hence the term 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus119909119896



2120577119896

119894(119892119887119890119904119905

119896


119894) represents





119908max





lower Γ



1

119894= 119909

1

119894 119892119887119890119904119905

1

swarm =

argmin1199091

119894

119891(1199091

119894) forall119894


V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(27)


= 119909119896

119894+ V119896+1119894



(9) Consider 119901119887119890119904119905119896119894= argmin

119909119895

119894

119891(119909119895

119894) 0 le 119895 le 119896


119894

119891(119909119896

119894) forall119894


1 1198922 119892

119904in


(12) Consider 119892119887119890119904119905119896119895= 119892119887119890119904119905

119896

swarm + 119903(119892119895minus 05) 119895 =


(13) If (119891(119892119887119890119904119905119896119895) lt 119891(119892119887119890119904119905

119896


119896

119895






0 10 20 30 40 50 60 70 80 90 100055

06

065

07

075

08

085

09

095

1 Suzie

Frame

Nor

mal

ized

pow

er


(a) Suzie

0 10 20 30 40 50 60 70 80 90 100075

08

085

09

095

1 Carphone

Frame

Nor

mal

ized

pow

er


(b) Carphone

Nor

mal

ized

pow

er

0 10 20 30 40 50 60 70 80 90 10004

045

05

055

06

065

07

075

08

085

09 Football

Frame


(c) Football



MOSth = 35 (28)






0 10 20 30 40 50 60 70 80 90 10035

36

37

38

39

4

41

42

43

44

45 Suzie

Frame

MO

S


(a) Suzie

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Carphone

Frame

MO

S


(b) Carphone

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Football

Frame

MO

S


(c) Football












(119875MOS)|optimal


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(30)



065 07 075 08 085 09 095 132

34

36

38

4

42

44

46

Normalized energy

MO

S



065 07 075 08 085 09 095 132

33

34

35

36

37

38

39

4

Normalized energy

MO

S






0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

Frame

Con

verg

ence




7 Conclusion





Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation














119904and119875119905as


119875 = 119875119904+ 119875119905 (1)



119863(SBR 119875119904) = 1205902

2minus120582SBRsdot119892(119875

119904)

0 le 119875 le 1 (2)


119904)


119892 (119875119904) = 1198751120574

119904 (3)



119894|



119894 respectively

SBR119894and 119901




119901119894= (

1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

(4)


119875119904= sum

119894

119901119894= sum

119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

(5)


119875 = 119875119904+ 119875119905= sum

119894

119901119894+ 119875119905

= sum

119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

+ 119875119905

(6)



119899))119871

(7)





BER (120574119899) =

119886119898

119890120574119899times119887119898 (8)


119898and 119887

119898are shown in



120574119899=119901119905sdot1003816100381610038161003816ℎ119899

1003816100381610038161003816

2

1198730119861119873

(9)


0is the spectral




AMC mode (119898) 119898 = 1 119898 = 2 119898 = 3 119898 = 4 119898 = 5 119898 = 6


119898) 12 12 34 916 34 34

119877119898(bitssym) 050 100 150 225 300 450

119886119898

11369 03351 02197 02081 01936 01887119887119898

75556 32543 15244 06250 03484 00871



119899))119871

= 119877119888sdot (1 minus

119886119898

119890120574119899times119887119898)

119871

= 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(10)



That is

SBR119894le 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(12)








120572 120573 120574 120575 120578 120590

39560 00919 minus58497 09844 01028 minus00236




min 119875 = min119891 (SBR 119875119905MCS)

= min(sum119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

+ 119875119905)


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

MOS ge MOSth

(14)







Parameter selection

Feedback

Parameter selection



Physical layer










119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

le 0

MOSth minusMOS le 0

(15)

Then define

ℎ (SBR 119875119905MCS) = max (SBR

119894minus 119877119888


)(1198730119861119873))times119887

119898)

119871

MOSth

minusMOS)

(16)


119891 (SBR 119875119905MCS) = ℎ (SBR 119875

119905MCS)

ℎ (SBR 119875119905MCS) gt 0

119891 (SBR 119875119905MCS) = 119886 tan [119891 (SBR 119875

119905MCS)] minus 120587

2

others

(17)






min119883

119891 (SBR 119875119905MCS) 119883 isin 119863 sube 119877

3

(18)


1 1199092




119909119894= [SBR 119875

119905] isin 1198772

(19)


119911119895+1

= 120583119911119895(1 minus 119911

119895) 119895 = 1 2 119904 (20)


1 1199112 119911





Γlower

= [SBRlower 119875

lower119905

]

Γupper

= [SBRupper 119875

upper119905

]

(21)


119909119894= Γ

lower+ (Γ

upperminus Γ

lower) sdot 119911119894 119894 = 1 2 119898 (22)



119909119896+1

119894= 119909119896

119894+ V119896+1119894

V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(23)




119908 = 119908min +119908max minus 119908min

119896 (24)


119894and 120577

119896

119894are




119901119887119890119904119905119896

119894= argmin

119909119895

119894

119891 (119909119895

119894) 0 le 119895 le 119896 (25)


119892119887119890119904119905119896


119894

119891 (119909119896

119894) forall119894 (26)

Hence the term 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus119909119896



2120577119896

119894(119892119887119890119904119905

119896


119894) represents





119908max





lower Γ



1

119894= 119909

1

119894 119892119887119890119904119905

1

swarm =

argmin1199091

119894

119891(1199091

119894) forall119894


V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(27)


= 119909119896

119894+ V119896+1119894



(9) Consider 119901119887119890119904119905119896119894= argmin

119909119895

119894

119891(119909119895

119894) 0 le 119895 le 119896


119894

119891(119909119896

119894) forall119894


1 1198922 119892

119904in


(12) Consider 119892119887119890119904119905119896119895= 119892119887119890119904119905

119896

swarm + 119903(119892119895minus 05) 119895 =


(13) If (119891(119892119887119890119904119905119896119895) lt 119891(119892119887119890119904119905

119896


119896

119895






0 10 20 30 40 50 60 70 80 90 100055

06

065

07

075

08

085

09

095

1 Suzie

Frame

Nor

mal

ized

pow

er


(a) Suzie

0 10 20 30 40 50 60 70 80 90 100075

08

085

09

095

1 Carphone

Frame

Nor

mal

ized

pow

er


(b) Carphone

Nor

mal

ized

pow

er

0 10 20 30 40 50 60 70 80 90 10004

045

05

055

06

065

07

075

08

085

09 Football

Frame


(c) Football



MOSth = 35 (28)






0 10 20 30 40 50 60 70 80 90 10035

36

37

38

39

4

41

42

43

44

45 Suzie

Frame

MO

S


(a) Suzie

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Carphone

Frame

MO

S


(b) Carphone

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Football

Frame

MO

S


(c) Football












(119875MOS)|optimal


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(30)



065 07 075 08 085 09 095 132

34

36

38

4

42

44

46

Normalized energy

MO

S



065 07 075 08 085 09 095 132

33

34

35

36

37

38

39

4

Normalized energy

MO

S






0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

Frame

Con

verg

ence




7 Conclusion





Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











AMC mode (119898) 119898 = 1 119898 = 2 119898 = 3 119898 = 4 119898 = 5 119898 = 6


119898) 12 12 34 916 34 34

119877119898(bitssym) 050 100 150 225 300 450

119886119898

11369 03351 02197 02081 01936 01887119887119898

75556 32543 15244 06250 03484 00871



119899))119871

= 119877119888sdot (1 minus

119886119898

119890120574119899times119887119898)

119871

= 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(10)



That is

SBR119894le 119877119888sdot (1 minus 119886

119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(12)








120572 120573 120574 120575 120578 120590

39560 00919 minus58497 09844 01028 minus00236




min 119875 = min119891 (SBR 119875119905MCS)

= min(sum119894

(1

120582119894sdot SBR119894

log2

1205902

119894

119863119894

)

120574

+ 119875119905)


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

MOS ge MOSth

(14)







Parameter selection

Feedback

Parameter selection



Physical layer










119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

le 0

MOSth minusMOS le 0

(15)

Then define

ℎ (SBR 119875119905MCS) = max (SBR

119894minus 119877119888


)(1198730119861119873))times119887

119898)

119871

MOSth

minusMOS)

(16)


119891 (SBR 119875119905MCS) = ℎ (SBR 119875

119905MCS)

ℎ (SBR 119875119905MCS) gt 0

119891 (SBR 119875119905MCS) = 119886 tan [119891 (SBR 119875

119905MCS)] minus 120587

2

others

(17)






min119883

119891 (SBR 119875119905MCS) 119883 isin 119863 sube 119877

3

(18)


1 1199092




119909119894= [SBR 119875

119905] isin 1198772

(19)


119911119895+1

= 120583119911119895(1 minus 119911

119895) 119895 = 1 2 119904 (20)


1 1199112 119911





Γlower

= [SBRlower 119875

lower119905

]

Γupper

= [SBRupper 119875

upper119905

]

(21)


119909119894= Γ

lower+ (Γ

upperminus Γ

lower) sdot 119911119894 119894 = 1 2 119898 (22)



119909119896+1

119894= 119909119896

119894+ V119896+1119894

V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(23)




119908 = 119908min +119908max minus 119908min

119896 (24)


119894and 120577

119896

119894are




119901119887119890119904119905119896

119894= argmin

119909119895

119894

119891 (119909119895

119894) 0 le 119895 le 119896 (25)


119892119887119890119904119905119896


119894

119891 (119909119896

119894) forall119894 (26)

Hence the term 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus119909119896



2120577119896

119894(119892119887119890119904119905

119896


119894) represents





119908max





lower Γ



1

119894= 119909

1

119894 119892119887119890119904119905

1

swarm =

argmin1199091

119894

119891(1199091

119894) forall119894


V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(27)


= 119909119896

119894+ V119896+1119894



(9) Consider 119901119887119890119904119905119896119894= argmin

119909119895

119894

119891(119909119895

119894) 0 le 119895 le 119896


119894

119891(119909119896

119894) forall119894


1 1198922 119892

119904in


(12) Consider 119892119887119890119904119905119896119895= 119892119887119890119904119905

119896

swarm + 119903(119892119895minus 05) 119895 =


(13) If (119891(119892119887119890119904119905119896119895) lt 119891(119892119887119890119904119905

119896


119896

119895






0 10 20 30 40 50 60 70 80 90 100055

06

065

07

075

08

085

09

095

1 Suzie

Frame

Nor

mal

ized

pow

er


(a) Suzie

0 10 20 30 40 50 60 70 80 90 100075

08

085

09

095

1 Carphone

Frame

Nor

mal

ized

pow

er


(b) Carphone

Nor

mal

ized

pow

er

0 10 20 30 40 50 60 70 80 90 10004

045

05

055

06

065

07

075

08

085

09 Football

Frame


(c) Football



MOSth = 35 (28)






0 10 20 30 40 50 60 70 80 90 10035

36

37

38

39

4

41

42

43

44

45 Suzie

Frame

MO

S


(a) Suzie

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Carphone

Frame

MO

S


(b) Carphone

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Football

Frame

MO

S


(c) Football












(119875MOS)|optimal


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(30)



065 07 075 08 085 09 095 132

34

36

38

4

42

44

46

Normalized energy

MO

S



065 07 075 08 085 09 095 132

33

34

35

36

37

38

39

4

Normalized energy

MO

S






0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

Frame

Con

verg

ence




7 Conclusion





Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation














Parameter selection

Feedback

Parameter selection



Physical layer










119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

le 0

MOSth minusMOS le 0

(15)

Then define

ℎ (SBR 119875119905MCS) = max (SBR

119894minus 119877119888


)(1198730119861119873))times119887

119898)

119871

MOSth

minusMOS)

(16)


119891 (SBR 119875119905MCS) = ℎ (SBR 119875

119905MCS)

ℎ (SBR 119875119905MCS) gt 0

119891 (SBR 119875119905MCS) = 119886 tan [119891 (SBR 119875

119905MCS)] minus 120587

2

others

(17)






min119883

119891 (SBR 119875119905MCS) 119883 isin 119863 sube 119877

3

(18)


1 1199092




119909119894= [SBR 119875

119905] isin 1198772

(19)


119911119895+1

= 120583119911119895(1 minus 119911

119895) 119895 = 1 2 119904 (20)


1 1199112 119911





Γlower

= [SBRlower 119875

lower119905

]

Γupper

= [SBRupper 119875

upper119905

]

(21)


119909119894= Γ

lower+ (Γ

upperminus Γ

lower) sdot 119911119894 119894 = 1 2 119898 (22)



119909119896+1

119894= 119909119896

119894+ V119896+1119894

V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(23)




119908 = 119908min +119908max minus 119908min

119896 (24)


119894and 120577

119896

119894are




119901119887119890119904119905119896

119894= argmin

119909119895

119894

119891 (119909119895

119894) 0 le 119895 le 119896 (25)


119892119887119890119904119905119896


119894

119891 (119909119896

119894) forall119894 (26)

Hence the term 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus119909119896



2120577119896

119894(119892119887119890119904119905

119896


119894) represents





119908max





lower Γ



1

119894= 119909

1

119894 119892119887119890119904119905

1

swarm =

argmin1199091

119894

119891(1199091

119894) forall119894


V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(27)


= 119909119896

119894+ V119896+1119894



(9) Consider 119901119887119890119904119905119896119894= argmin

119909119895

119894

119891(119909119895

119894) 0 le 119895 le 119896


119894

119891(119909119896

119894) forall119894


1 1198922 119892

119904in


(12) Consider 119892119887119890119904119905119896119895= 119892119887119890119904119905

119896

swarm + 119903(119892119895minus 05) 119895 =


(13) If (119891(119892119887119890119904119905119896119895) lt 119891(119892119887119890119904119905

119896


119896

119895






0 10 20 30 40 50 60 70 80 90 100055

06

065

07

075

08

085

09

095

1 Suzie

Frame

Nor

mal

ized

pow

er


(a) Suzie

0 10 20 30 40 50 60 70 80 90 100075

08

085

09

095

1 Carphone

Frame

Nor

mal

ized

pow

er


(b) Carphone

Nor

mal

ized

pow

er

0 10 20 30 40 50 60 70 80 90 10004

045

05

055

06

065

07

075

08

085

09 Football

Frame


(c) Football



MOSth = 35 (28)






0 10 20 30 40 50 60 70 80 90 10035

36

37

38

39

4

41

42

43

44

45 Suzie

Frame

MO

S


(a) Suzie

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Carphone

Frame

MO

S


(b) Carphone

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Football

Frame

MO

S


(c) Football












(119875MOS)|optimal


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(30)



065 07 075 08 085 09 095 132

34

36

38

4

42

44

46

Normalized energy

MO

S



065 07 075 08 085 09 095 132

33

34

35

36

37

38

39

4

Normalized energy

MO

S






0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

Frame

Con

verg

ence




7 Conclusion





Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












Γlower

= [SBRlower 119875

lower119905

]

Γupper

= [SBRupper 119875

upper119905

]

(21)


119909119894= Γ

lower+ (Γ

upperminus Γ

lower) sdot 119911119894 119894 = 1 2 119898 (22)



119909119896+1

119894= 119909119896

119894+ V119896+1119894

V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(23)




119908 = 119908min +119908max minus 119908min

119896 (24)


119894and 120577

119896

119894are




119901119887119890119904119905119896

119894= argmin

119909119895

119894

119891 (119909119895

119894) 0 le 119895 le 119896 (25)


119892119887119890119904119905119896


119894

119891 (119909119896

119894) forall119894 (26)

Hence the term 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus119909119896



2120577119896

119894(119892119887119890119904119905

119896


119894) represents





119908max





lower Γ



1

119894= 119909

1

119894 119892119887119890119904119905

1

swarm =

argmin1199091

119894

119891(1199091

119894) forall119894


V119896+1119894

= 119908V119896119894+ 1198881120585119896

119894(119901119887119890119904119905

119896

119894minus 119909119896

119894)

+ 1198882120577119896

119894(119892119887119890119904119905

119896


119894)

(27)


= 119909119896

119894+ V119896+1119894



(9) Consider 119901119887119890119904119905119896119894= argmin

119909119895

119894

119891(119909119895

119894) 0 le 119895 le 119896


119894

119891(119909119896

119894) forall119894


1 1198922 119892

119904in


(12) Consider 119892119887119890119904119905119896119895= 119892119887119890119904119905

119896

swarm + 119903(119892119895minus 05) 119895 =


(13) If (119891(119892119887119890119904119905119896119895) lt 119891(119892119887119890119904119905

119896


119896

119895






0 10 20 30 40 50 60 70 80 90 100055

06

065

07

075

08

085

09

095

1 Suzie

Frame

Nor

mal

ized

pow

er


(a) Suzie

0 10 20 30 40 50 60 70 80 90 100075

08

085

09

095

1 Carphone

Frame

Nor

mal

ized

pow

er


(b) Carphone

Nor

mal

ized

pow

er

0 10 20 30 40 50 60 70 80 90 10004

045

05

055

06

065

07

075

08

085

09 Football

Frame


(c) Football



MOSth = 35 (28)






0 10 20 30 40 50 60 70 80 90 10035

36

37

38

39

4

41

42

43

44

45 Suzie

Frame

MO

S


(a) Suzie

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Carphone

Frame

MO

S


(b) Carphone

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Football

Frame

MO

S


(c) Football












(119875MOS)|optimal


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(30)



065 07 075 08 085 09 095 132

34

36

38

4

42

44

46

Normalized energy

MO

S



065 07 075 08 085 09 095 132

33

34

35

36

37

38

39

4

Normalized energy

MO

S






0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

Frame

Con

verg

ence




7 Conclusion





Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










0 10 20 30 40 50 60 70 80 90 100055

06

065

07

075

08

085

09

095

1 Suzie

Frame

Nor

mal

ized

pow

er


(a) Suzie

0 10 20 30 40 50 60 70 80 90 100075

08

085

09

095

1 Carphone

Frame

Nor

mal

ized

pow

er


(b) Carphone

Nor

mal

ized

pow

er

0 10 20 30 40 50 60 70 80 90 10004

045

05

055

06

065

07

075

08

085

09 Football

Frame


(c) Football



MOSth = 35 (28)






0 10 20 30 40 50 60 70 80 90 10035

36

37

38

39

4

41

42

43

44

45 Suzie

Frame

MO

S


(a) Suzie

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Carphone

Frame

MO

S


(b) Carphone

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Football

Frame

MO

S


(c) Football












(119875MOS)|optimal


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(30)



065 07 075 08 085 09 095 132

34

36

38

4

42

44

46

Normalized energy

MO

S



065 07 075 08 085 09 095 132

33

34

35

36

37

38

39

4

Normalized energy

MO

S






0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

Frame

Con

verg

ence




7 Conclusion





Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










0 10 20 30 40 50 60 70 80 90 10035

36

37

38

39

4

41

42

43

44

45 Suzie

Frame

MO

S


(a) Suzie

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Carphone

Frame

MO

S


(b) Carphone

30 40 50 60 70 80 90 100 110 120 13035

36

37

38

39

4

41

42

43

44

45 Football

Frame

MO

S


(c) Football












(119875MOS)|optimal


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(30)



065 07 075 08 085 09 095 132

34

36

38

4

42

44

46

Normalized energy

MO

S



065 07 075 08 085 09 095 132

33

34

35

36

37

38

39

4

Normalized energy

MO

S






0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

Frame

Con

verg

ence




7 Conclusion





Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation














(119875MOS)|optimal


119898sdot 119890((minus119875119905sdot|ℎ119899|2

)(1198730119861119873))times119887

119898)

119871

(30)



065 07 075 08 085 09 095 132

34

36

38

4

42

44

46

Normalized energy

MO

S



065 07 075 08 085 09 095 132

33

34

35

36

37

38

39

4

Normalized energy

MO

S






0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

Frame

Con

verg

ence




7 Conclusion





Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

Frame

Con

verg

ence




7 Conclusion





Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation





































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









Documents

Research Article QoE Perceptive Cross-Layer Energy Efficient Method for Mobile Video ...downloads.hindawi.com/journals/ijdsn/2015/980174.pdf · 2015-11-24 · Research Article QoE