Wearable Sensors for Reliable Fall Detection

Proceedings of the 2005 IEEE

Engineering in Medicine and Biology 27th Annual ConferenceShanghai, China, September 1-4, 2005

WearableSensorsforReliableFallDetection

JayChen ,KarricKwong ,DennisChang ,JerryLuk§,RuzenaBajcsy¶

DepartmentofElectricalEngineering

andComputerSciences

UniversityofCalifornia,Berkeley

Berkeley,CA94720

¶ {bajcsy,dhchang}@eecs.berkeley.edu,{jayzchen, karrickk}@berkeley.edu,§jerryluk@cal.berkeley.edu

Abstract—Unintentionalfallsareacommoncauseofsevere

injuryintheelderlypopulation.Byintroducingsmall,non-invasivesensormotesinconjunctionwithawirelessnetwork,http://doc.xuehai.netingasmalldevicewornonthewaistandanetworkof xedmotesinthehomeenvironment,wecandetecttheoccurrenceofafallandthelocationofthevictim.Low-costandlow-powerMEMSaccelerometersareusedtodetectthefallwhileRFsignalstrengthisusedtolocatetheperson.

falldetectiontechnology,localizationcandetectwheretheincidentoccurredandrequesttherelevantservices.

II.FALLDETECTIONANDLOCALIZATION

A.PreviousWorkonFallDetection

Accelerometryhasbeenusedinvariousstudiesandappli-cationstoobjectivelymonitorarangeofhumanmovement,forexampletomeasuremetabolicenergyexpenditure,phys-icalactivitylevels,balanceandposturalsway,gait,andtodetectfalls[4].Withrespecttofalldetection,therehasbeenrelativelylittleworkpublished.Accordingto[4],thebasicapproachofusingaccelerometrytodetectthefallwas rstpublishedby[5],[6].Inthisapproach,achangeinbodyorientationfromuprighttolyingthatoccursimmediatelyafteralargenegativeaccelerationindicatesafall.Thesetwoconditionshavebeenincorporatedintofalldetectionalgorithmsusingaccelerometers[7],[8].

Reference[9]presentsafalldetectorwornonthewristthatincorporatesamulti-stagefalldetectionalgorithm.The rstconditionisthedetectionofahighvelocitytowardstheground.Nextanimpactneedstobedetectedwithin3seconds.Afterimpact,theactivityisobservedfor60seconds,andifatleast40secondsofinactivityarerecorded,analarmisactivated.Theresultswerepositiveinthesensethatnofalsealarmsweregiven,butalsodisappointingsincealargepercentageofbackwardsandsidewaysfallswerenotdetected.

Reference[6]documentsthedesignofthecommerciallyavailableTunstallfalldetectorthatusesapatentedtwo-stagedetectionalgorithm.Thedetectorwakesupfromthesleepstatewhenastrongimpactisdetected.Thenasecondsensorestimatesthewearer’sorientationandifhe/sheisinalyingstateforasettimeperiod,analarmisraised.Variouslocationsforthedevicewereconsideredanditwasdeterminedthatthewaistwastheoptimumlocationthatsuitedthewearerandallowedreliablemeasurementofimpacts.

III.SYSTEMDESIGN

OurapplicationutilizesTinyOSandMica2Dotmotesde-velopedatUCBerkeleyasaresearchplatformforlow-powerwirelesssensornetworks[10],[11].TheMica2DotmoteisequippedwiththeAtmelATmega128Lmicrocontroller,4KBofRAManda433MHzradiocapableofdatatransfer

I.INTRODUCTION

A.Background

Fallingcanbeafrequentanddangerouseventfortheelderlypopulation.Itisestimatedthatoverathirdofadultsages65yearsandolderfalleachyear[1],makingittheleadingcauseofnonfatalinjuryforthatagegroup.

Amongolderpersons,55percentoffallinjuriesoccurinsidethehome.Anadditional23percentoccuroutside,butnearthehome[2].Traditionally,placingseniorsinnursinghomesorothercarecentershasmitigatedthedangersoftheelderlyfalling.However,withtheadventofwirelessad-hocnetworksandlow-powermotetechnology,wecannowapproachtheproblemfromadifferentperspective.B.Motivation

ThegoaloftheIvyProject[3]istoprovideaninfrastruc-tureofnetworkedsensorsthatsupportsmultipleapplicationssimultaneously.Thesensornetwork,likeivy,wouldspreadthroughouttheenvironment,whetheritisanof cespaceorhome,linkingleaves(motes)totheroot(basestation).Ingeneral,themotescanbedividedintotwotypes:1)Fixed/infrastructuremotes,forexampleattachedalongsidethewallsandcorridors,and2)mobilemotes,whosegeo-graphicalpositioncanchangeovertime.

Inourapplication,thisnetworkedinfrastructureisusedtodetectwhenapersonhassustainedafallandrelaythisinformationacrosssomemediumsuchthatimmediateandappropriateactioncanbetaken.ThisiscurrentlythefocusoftheIvyProjectandmotivatestheexperimentsdiscussedinthispaper.

Inaddition,http://doc.xuehai.netbinedwiththe

第1页下一页