基于超限學(xué)習(xí)機(jī)的無(wú)設(shè)備定位方法研究

基于超限學(xué)習(xí)機(jī)的無(wú)設(shè)備定位方法研究基于超限學(xué)習(xí)機(jī)的無(wú)設(shè)備定位方法研究

摘要

無(wú)線定位技術(shù)因其方便快捷、無(wú)需硬件部署、精度高等優(yōu)點(diǎn)而受到廣泛關(guān)注。本文提出一種基于超限學(xué)習(xí)機(jī)的無(wú)設(shè)備定位方法，其中超限學(xué)習(xí)機(jī)被用于實(shí)現(xiàn)非線性函數(shù)映射，通過(guò)收集Wi-Fi信號(hào)強(qiáng)度和位置數(shù)據(jù)作為訓(xùn)練集，并以壓縮感知的方式實(shí)現(xiàn)極限降維，來(lái)進(jìn)行定位。為了進(jìn)一步提升精度，本文引入了局部權(quán)重貢獻(xiàn)方法來(lái)降低信號(hào)強(qiáng)度測(cè)量誤差對(duì)定位結(jié)果的影響。

本文還在室內(nèi)環(huán)境下進(jìn)行了一系列實(shí)驗(yàn)，比較了所提出的方法與傳統(tǒng)的KNN定位算法和基于支持向量機(jī)的定位方法。實(shí)驗(yàn)結(jié)果表明，所提出的無(wú)設(shè)備定位方法具有較高的定位精度和更好的魯棒性。

關(guān)鍵詞：超限學(xué)習(xí)機(jī)；無(wú)設(shè)備定位；Wi-Fi；壓縮感知；局部權(quán)重貢獻(xiàn)。

Abstract

Wirelesspositioningtechnologyhasattractedwidespreadattentionduetoitsconvenience,nohardwaredeployment,andhighaccuracy.Inthispaper,adevice-freepositioningmethodbasedonextremelearningmachine(ELM)isproposed,inwhichtheELMisusedtoachievenon-linearfunctionmapping.Wi-Fisignalstrengthandlocationdataarecollectedastrainingsets,andextremedimensionreductionisachievedbycompressivesensingtoperformpositioning.Inordertofurtherimprovetheaccuracy,thispaperintroducesthelocalweightcontributionmethodtoreducetheimpactofmeasurementerrorsonthepositioningresults.

Inaddition,aseriesofexperimentswerecarriedoutinanindoorenvironmenttocomparetheproposedmethodwiththetraditionalKNNpositioningalgorithmandthesupportvectormachine-basedpositioningmethod.Theexperimentalresultsshowthattheproposeddevice-freepositioningmethodhashigherpositioningaccuracyandbetterrobustness.

Keywords:Extremelearningmachine(ELM);device-freepositioning;Wi-Fi;compressivesensing;localweightcontributionDevice-freepositioninghasbecomeanimportantresearchareaduetoitswiderangeofapplicationssuchassecurity,healthcare,andhomeautomation.Inthisstudy,anoveldevice-freepositioningalgorithmbasedonextremelearningmachine(ELM)andcompressivesensingwasproposed.Theproposedalgorithmutilizesthereceivedsignalstrength(RSS)ofWi-Fisignalstoestimatethepositionofatargetuserwithouttheneedforanyadditionaldevicesorsensors.

TheELMalgorithmwasutilizedtotrainalocalweightcontributionmatrix,whichisusedtodeterminethecontributionofeachsignalstrengthmeasurementtothepositioningresults.CompressivesensingwasusedtoreducethedimensionalityoftheRSSmatrix,thusreducingthecomputationalcomplexityandimprovingtheaccuracyofthealgorithm.

Aseriesofexperimentswereconductedinanindoorenvironmenttoevaluatetheproposeddevice-freepositioningmethod.TheexperimentalresultsshowedthattheproposedmethodoutperformedthetraditionalKNNpositioningalgorithmandthesupportvectormachine-basedpositioningmethodintermsofaccuracyandrobustness.

Inconclusion,thisstudyproposesanoveldevice-freepositioningalgorithmbasedonELMandcompressivesensing,whichcanaccuratelyestimatethepositionofatargetuserusingonlyWi-Fisignals.Themethodhaspotentialforawiderangeofapplications,includinghomeautomation,healthcare,andsecurityTherearesomelimitationsandfuturedirectionsfortheproposeddevice-freepositioningalgorithmbasedonELMandcompressivesensing.First,thealgorithmassumesthattheenvironmentisstaticduringthepositioningprocess.However,inreal-worldscenarios,theenvironmentmaychangedynamicallyovertime,whichcouldaffecttheaccuracyofthealgorithm.Therefore,futureresearchcanfocusondevelopingdynamicalgorithmsthatcanadapttochangingenvironments.

Second,thealgorithmisbasedonWi-Fisignals,whichmaynotbeavailableinallenvironments.Insuchcases,alternativesignals,suchasBluetoothorRFID,couldbeused.Futureresearchcanexplorehowtheproposedalgorithmcouldbeadaptedtoworkwithothertypesofsignals.

Third,theproposedalgorithmrequiresatrainingphasetobuildthedictionarymatrix.Thisprocesscanbetime-consumingandmaynotbefeasibleinsomereal-worldscenarios.Therefore,futureresearchcanfocusondevelopingalgorithmsthatdonotrequireatrainingphase.

Fourth,theproposedalgorithmcurrentlyonlyworksforsingle-userscenarios.Inmulti-userenvironments,interferencebetweenuserscouldaffecttheaccuracyofthealgorithm.Therefore,futureresearchcanexplorehowthealgorithmcouldbeadaptedtoworkinmulti-userscenarios.

Finally,whiletheproposedalgorithmoutperformedtraditionalpositioningalgorithmsintermsofaccuracyandrobustness,thereisstillroomforimprovement.Futureresearchcanfocusondevelopingmoreadvancedalgorithmsthatfurtherimprovetheaccuracyandefficiencyofdevice-freepositioningsystems.

Overall,theproposeddevice-freepositioningalgorithmbasedonELMandcompressivesensinghasthepotentialtorevolutionizeindoorpositioningsystems.Withfurtherdevelopmentandresearch,itcouldenableawiderangeofapplicationsthatbenefitsocietyOnepotentialapplicationofdevice-freepositioningsystemsisinthefieldofhealthcare.Hospitalstaffneedtokeeptrackofpatientsandmedicalequipmentwithinthehospitalenvironment,andaccurateindoorpositioningcanhelptoincreaseefficiencyandreduceerrors.Forexample,adevice-freepositioningsystemcouldbeusedtotrackthemovementofahospitalbedandalertstaffwhenitreachesacertainlocation,suchastheoperatingroom.Itcouldalsobeusedtotrackthelocationofmedicalstaff,ensuringthattheyareinthecorrectareatoprovidetherequiredmedicalcare.

Anotherpotentialapplicationisinthefieldofsecurity.Traditionalsecuritysystemssuchasvideocamerasmaybeineffectiveincertainsituations,suchaswhentheintruderiswearingamaskorifthecamera'sviewisblocked.Adevice-freepositioningsystemcandetectthepresenceofahumanbeingeveniftheyarenotcarryinganyelectronicdevices,enablingsecuritypersonneltoidentifytheintruderandtakeappropriateaction.

Moreover,device-freepositioningsystemscanalsobeusedinenvironmentalmonitoring.Theycandetectandtrackthemovementofwildlifeinnaturalhabitatswithoutdisturbingthem,providingvaluableinformationtoresearchersandconservationists.Theycanalsobeusedtomonitorthemovementofpeopleindisasterzones,enablingfirstresponderstolocatesurvivorsandprovideassistancemoreefficiently.

Finally,device-freepositioningsystemscanbeusedinretailenvironments.Theycanprovidevaluableinsightsintocustomerbehavior,suchashowtheynavigatethestoreandwhichitemsaremostpopular.Thisinformationcanbeusedtoimprovestorelayoutandproductplacement,leadingtoincreasedsalesandcustomersatisfaction.

Inconclusion,device-freepositioningsystemshaveenormouspotentialtoenhance