A review on the use of sensors to monitor cattle jaw ... · B A S E Biotechnol. Agron. Soc....

BASE Biotechnol. Agron. Soc. Environ.201620(S1),

AreviewontheuseofsensorstomonitorcattlejawmovementsandbehaviorwhengrazingAndriamasinoroLalainaHerinainaAndriamandroso(1,2,3)*,JérômeBindelle(3)*,BenoîtMercatoris(2),FrédéricLebeau(2)(1)UniversitédeLiège-GemblouxAgro-BioTech.TERRA.AgricultureIsLife.PassagedesDéportés,2.BE-5030Gembloux(Belgium).E-mail:[email protected](2)UniversitédeLiège-GemblouxAgro-BioTech.PrecisionAgricultureUnit.PassagedesDéportés,2.BE-5030Gembloux(Belgium).(3)UniversitédeLiège-GemblouxAgro-BioTech.PrecisionLivestockandNutritionUnit.PassagedesDéportés,2.BE-5030Gembloux(Belgium).

*Theseauthorscontributedequallytothispaper.

ReceivedonApril1,2015;acceptedonMay24,2016.

Introduction.PrecisionLivestockFarming(PLF)isspreadingrapidlyinintensivecattlefarms.Itisbasedonthemonitoringofindividualsusingdifferentkindsofsensors.Appliedtograzinganimals,PLFismainlybasedontherecordingofthreeparameters: the location, the posture and themovements of the animal.Until now, several techniques have beenused todiscriminate grazing and ruminating behaviors with accuracies over 90% on average, when compared to observations,providingvaluabletoolstoimprovethemanagementofpastureandgrazinganimals.However,bitesandjawmovementsarestilloverlooked,eventhoughtheyareofutmostimportancetoassesstheanimalgrazingstrategiesforvariouspasturetypesanddevelopfuturetechniquesallowingbetterestimationoftheirintake.Literature. The goal of this review is to explore the possibility of monitoring the individual jaw movements and thedifferentiation of bites in grazing animals. For this purpose, (1) themechanisms of forage intake in cattle are explainedbrieflyinordertounderstandthemovementsperformedbythecow,especiallyduringgrazing,(2)thevarioussensorsthathavebeenproposedtomonitorjawmovementsofruminantssuchasmechanicalsensors(pressuresensors),acousticsensors(microphone)andelectromyographysensorsarecomparedand(3)finallytherelationshipbetweenjawmovements,bitingbehaviorandforageintakeisdiscussed.Conclusions.Thereviewclearlydemonstratedtheabilitiesofmechanical,acousticandelectromyographysensorstoclassifythe difference types of jawmovements.However, it also indicated awide range of accuracies and different observationwindowsrequiredtoreachtheseaccuracieswhencomparedtotheobservedmovement.Thisclassificationpurposecouldleadtoabetterdetectionofmorespecificbehavior,e.g.bitedetection,andtheirexactlocationonpasture.Keywords.Precisionagriculture,livestockmanagement,cattle,sensors,movements,forage.

Synthèse sur l’utilisation de capteurs pour le suivi des mouvements de mâchoire et du comportement de bovins au pâturageIntroduction.L’élevagedeprécisionse répandrapidementauniveaudesexploitationsbovinesde type intensif. Ilutilisedifférents capteurs pour suivre chaque individu présent dans le troupeau. Pour les ruminants au pâturage, le système estbasésurl’enregistrementdetroisparamètres:leurlocalisation,leurpostureetleursmouvements.Lestechniquesactuellespermettentdedétecterlescomportementsdepâturageetderuminationavecuneprécisionmoyennesupérieureouégaleà90%,comparéeauxobservations.Cestechniquespeuventfournirdesoutilsintéressantspouraméliorerlagestiondelaprairieetdesanimaux.Cependant,lacaractérisationdesmouvementsdelamâchoireetdesbouchéesrestesouventnégligée,sachantquecesparamètrespeuventêtretrèsimportantspourévaluerlesstratégiesdepâturagedesanimauxetpourespérerestimerlaquantitédefourrageingérée.Littérature.L’objectifdecette synthèseestdediscuterdes techniquesutiliséespour lacaractérisationet laclassificationdesmouvementsde lamâchoireainsiquedesbouchéeschez les ruminants.Pourcela, (1) lesmécanismesd’ingestiondefourragedesbovinssontd’abordbrièvementexpliqués,ensuite(2)lesdifférentstypesdecapteursutiliséspourdétecterlesmouvementsdelamâchoire,telsquelescapteursdepression,accéléromètre,microphonesetcapteursélectromyographiques,sontdécritsetcomparés,et(3)leséventuelsliensentrelesmouvementsdelamâchoire,labouchéeetlefourrageingérésontdiscutésensebasantsurlesrésultatsderecherchedéjàeffectuésdanscedomaine.

2 Biotechnol. Agron. Soc. Environ. 201620(S1), AndriamandrosoA.L.H.,BindelleJ.,MercatorisB.C.N.etal.

1. INTRODUCTION

Sensors enable the monitoring of many physicalvariables.Their use in Precision Livestock Farming(PLF)hasincreasedrapidlyoverthepastdecadeforresearch purposes and also in on-farm applications.Unlike more traditional livestock managementmethodswhichfocusontheherd,PLFisbasedon:–themonitoring of variables at the individual level(Hostiouetal.,2014)andatanappropriatefrequencywithreliablesensors;

–the development of predictive models describingtheanimal’sresponsestoenvironmentalstimuliforeachmeasuredvariable;

–thecomparisonofthepredictionmodelswithwhatisactuallymeasuredthroughthesensors.

Theultimategoalistosuggestmanagerialoptionsto the farmer. Measuring such variables requirestrade-offs between upstream data acquisition athigh frequency, while preserving battery life andconsideringmemory limitswhichare specific to thesensorsthatareused,anddownstreamoutputaccuracyobtainedusingadequatedatatreatmentmethods.

In this respect, sensors can be used individuallyor in combination to track, detect, classify, manageand possibly control animal movements andbehaviors.Monitoringruminantbehavior isakey tounderstandinghowanimals fulfill their requirementsinpastoralsystemsbygrazingadynamicvegetationto achieve optimal plant production, animal forageintakeandperformances(Carvalho,2013).Traditionalmanagerial tools are limited to adjusting stockingrates and occupation times in rotational grazing,supplementingtheanimalsandcontrollingconcentrateintake at herd or individual level and indirectmonitoringofforageintakethroughmilkproduction,growth performance, or pasture disappearance(Holechek et al., 2011; Carvalho, 2013). PrecisionLivestock Farming opens a wide range of newperspectives in both intensive pasture and extensiverangelandmanagementbyfocusingontheindividualinsteadofthewholeherd.Forexample,Laca(2009)proposed a system inwhich both animal health andplant-animalinteractionsaremonitoredbycombininganimalpositionandbehaviordatatoremotelymanageindividualhealthandfeeding.

Controllingindividualanimalforagingbehavioronpasturemeansthemonitoringofgrazing,rumination

andrestingbehaviors,whichalltogetheroccupy90%to95%ofthedailytimebudget.Duringthelast5to10%, animals are busy displaying social behaviors,walking,drinking,eatingsupplements,etc.thatmightalso provide interesting insights in terms of animalmanagement(Walkeretal.,2008).Focusingonfeedingbehavior,individualmonitoringofgrazinganimalsisbasedontherecordingofthreemainparameters:–thelocationoftheanimal:whereitisinthepaddock,inordertoidentifygrazingstations;

–the posture of the animal, the static elementcomposing a behavior such as the position of theheadortheback;

–themovementof theanimal, thedynamicelementcomposingabehaviorsuchasmovinglegsorjaw.

Tracking location on pasture was made possibleby the large dissemination of Global PositioningSystem (GPS) sensors. Global Positioning Systemhasbeensuccessfullyusedtodetectstaticordynamicunitary behaviors differentiated through changes inpathspeeds:foragingorgrazing,restingandwalking(Anderson et al., 2012). Nonetheless, accuracies ofbehavior classification based on GPS sensors, withsampling frequency lower than 10Hz, remain poor,i.e. <80% when compared to visual observationsbasedontimewindowsof5mineach(Schlechtetal.,2004; Larson-Praplan et al., 2015). Posture analysishasbeenmorerecentlydevelopedthroughtheuseofaccelerometersandbasedmainlyonthepositionofthehead:upordown.This information, in combinationwithGPS-baseddata,alloweddiscriminationbetweenseveral kinds of feeding related behaviors forgrazinganimalswithhighaccuracies(>90%).Thoseaccuracieswereobtainedwithashorttimewindowof5to10swhilethedataacquisitionfromtheGPSandtheaccelerometerranbetween4Hzand10Hz(Duttaetal.,2015;Gonzálezetal.,2015).

Finally,monitoringofcattlemovementsismainlyobtained using accelerometers. Through diverseanalysis methods, accelerometers recording data at10Hz could be used to classify behaviors, as donefor example by Mangweth et al. (2012) when theyclassified lame and non-lame cows using a basicstatistical method, reaching an average accuracy of91%.Similarly,Martiskainenet al. (2009)classifiedmultiplebehaviorsusingamachine learningmethodwithaccuraciesrangingfrom29%to86%withsampleswindowedfor10sforallbehaviorclassifications.

Conclusions.Laconclusiondecettesynthèseestquelescapteursmécaniques,acoustiquesetélectromyographiquesontmontréleurcapacitéàclassifierlesdifférentstypesdemouvementsdelamâchoireavecdifférentesprécisionsetdifférentesfenêtresdetempsnécessairespourcetteclassification.Cetteclassificationpourraitmeneràunemeilleuredétectiondecomportementsplusprécistellequeladétectiondesbouchéesetleurlocalisationsurleparcours.Mots-clés.Agriculturedeprécision,conduited’élevage,bovins,capteurs,mouvements,fourrage.

Review:sensorsforcattlejawmovementsmonitoring 3

The online detection and classification of thebehaviors are essential for the development ofremoteandautomaticmonitoringofcattleonpasture.However, other components of animal movements,like jawmovements,arepresentlyoverlooked,whiletheyareofutmostimportancetoassessanimalgrazingstrategieswhengrazingvarioustypesofpasturesandtodevelopnewmethodstobetterestimatetheirintake.Nonetheless,bitemassandsubsequentintakeratearethemostvariablecomponentsofthefeedingbehavior,thusthemostdifficulttopredict.Bitemassandintakerateareinfluencedbyverylargerangewithswardheight,swardbulkdensity,botanicalfamilyandspecies,animalcharacteristicsandmotivationdurationofthepreviousstarvationperiod,grazing system,pasture allowance,concentrate and forage supplementation level, dailytimeofaccesstopasture,hourofday,etc.(Rooketal.,1994;Gibbetal.,1997;Barrettetal.,2001).Thereforerecordingbitemassfromjawmovementsseemsonlya dream today.Bonnet et al. (2015) recently studiedthepossibilityofdoingacontinuousbitemonitoringwith acoustic sensors coupled to direct observationwith trained observers. Combined with preliminaryestimatesofbitemassperformedbythehand-pluckingmethod (Bonnet et al., 2011), they could estimatebite mass with an accuracy ranging between 80%and 94%, in a short-term intake rate (approximately

10min).Although not applicable for PLF purposes,such methods suggest that combining informationprovided by different sensors, for example location,head position and acceleration, and jawmovements,mayhelpovercometheeverlastingchallengeofintakeestimationonpasture.

The main goal of this review is to assess thetechnologiesavailableforthemonitoringofindividualjawmovementsincattleforresearchandPLFusesandto discuss the mechanisms of bite mass constitutionandtheirlinkswithjawmovements.Forthispurpose,wewill:–discussthemechanismsofforageintake,–detail various sensors that have been proposed tomonitorjawmovementsofruminants,

–outline the link between jaw movements, bitingbehavior and forage intake, focusing mainly oncattle.

2. MECHANISM OF FORAGE INTAKE FOR CATTLE

Grazing is a complex combination of variousmovements and activities performed at differenttemporalandspatialscalesasshowninfigure 1.Thesinglebiteistheelementarycomponentofthegrazing

Spatio-temporal scale

1 - 2 s

5 - 100 s

1 - 30 min

1 - 4 h

1 - 4 days to months

Bite area

Path width

Patch

Grazing site

Pasture

Animal intake level

Bite volume

Bite rate

Intake rate

Intake per meal

Daily intake

Bite

Feeding

station

Grazing bout

Grazing event or meal

Pasture or paddock

Figure 1. Spatio-temporal components of grazing behavior (adapted fromGibb, 1996;Gregorini et al., 2006; Carvalho,2013)—Composantes spatio-temporelles du comportement de pâturage (adapté de Gibb, 1996 ; Gregorini et al., 2006 ; Carvalho, 2013).


process (Ungar et al., 2006a; Carvalho, 2013). Itsfrequency ranges from 0.75 to 1.2bites per secondanditssizeismainlydeterminedbythemouthoftheanimalaswellassomevegetationcharacteristicssuchasswardheight,tensilestrengthanddensity(Griffithset al., 2003; Oudshoorn et al., 2013a; Oudshoornetal.,2013b).Severalbitesperformedinarowbyananimalonasinglefeedingstationwithoutinterruptioncomposeagrazingbout(Gibb,1996)thatwillcoverafewsquaremetersandlastbetween10to100seconds(Andriamandrosoetal.,2015).Severalgrazingboutsare performed during each grazing event or meal(Gibb,1996)thatoccurseachdayforseveralminutesto hours during which a significant portion of thepaddockisexplored.Finally,thepaddockisoccupiedfor somedays to severalmonthsandcoversanareathatisusuallyover1ha.Onlythetwohighestlevelsinfigure 1,i.e.grazingeventandpaddock,arediscussedinmostofresearchonthedetectionandclassificationofgrazingbehavior.

A detailed observation of cattle movementsat individual bite level (Figure 2) shows thatforagingrequiresmainly jawandaccessorily tonguemovementsthatcanbebrokendownintofourphases.Firstly, during prehension, cows surround a bunchof grass using their tongue and lips (Frames 1 and6,Figure 2)and take it into theirmouth (Frame11,Figure 2). Then the grass is grabbed between thelower jawand the gum (Frame16,Figure 2) and itis finally cutwith a suddenmovement of the lowerjawusuallyaccompaniedbyamovementofthewholehead to perform the proper defoliation (Frames 21and 25, Figure 2). This abrupt head movement ismarkedbyanupwardthrustof themouthvisiblebythe increase in distance between themouth and thebaseline and is usually considered as the actual bite(Gibb,1996) (Figure 2).Thewhole forage intake isendedbychewingandswallowingtheplantbiomass(Vallentine,2000).

Rumination jaw movements are more quiet andregular.Theyarecomposedofacyclicprocesswhichbegins with the regurgitation of a rumino-reticularbolusfollowedbysemi-circularjawmovementswitha specific frequency of 1.06± 0.06bites.s-1 duringmasticationandendswiththedeglutition.Deglutitionisdescribedasapausebetweentwoboutsofmasticationwhilethemasticationcyclelastsbetween15sand60s(Andriamandrosoetal.,2014).

In order to detect bites, various techniqueshave been developed to monitor jaw movements ofruminants(Figure 3).Beforetheearly1980’s,toolsforcountingjawmovementswerestrictlymechanical.Jawmovementswererecordedonpaperrollsordisks(e.g.Balch,1958)orcountedviabuilt-inelectricalcircuitsusedasrecorders(e.g.Stobbsetal.,1972).Howeversuchdevicescannotbeproperlyconsideredassensorssinceasensorisadevicethatdetectseventsormeasureschangesinaphysicalpropertysuchaslight,forceandsound in its environmentand transforms them intoausablesignal,usuallyanelectricaloutput, for furtheranalysis (Kenny, 2005). According to a literaturesurvey (Figure 3) pressure and microphone sensorsare the most used sensors for monitoring cattle jawmovements with 35 and 14 references, respectively.Whileonly4referencesmentionaccelerationsensors,their use is rising rapidly.Electromyography sensorsarealsosometimescited(2references),whilethetwolastreferencescompareddifferenttypesofsensors.

3. USE OF SENSORS FOR JAW MOVEMENTS DETECTION AND QUANTIFICATION IN CATTLE

Devicesdedicatedforjawmovementsdetectioncanbeclassifiedinfivegroups(Figure 3):–jawswitches,whereaswitchisactivatedateachjawmovement;

Base

line

Frame 1 Frame 6 Frame 11 Frame 16 Frame 21 Frame 250 s 1 s

Figure 2.Visualizationofcattlemouthmovements(25framespersecondvideo)—Mouvements effectués par la vache lors du pâturage (photographies tirées d’une vidéo à 25 images par seconde).


Figure 3. Principaltoolsusedtocharacterizejawmovementsofcattle:referencesusedinthisfigurecamefromresearchinScopus(www.scopus.com,Elsevier,TheNetherlands,31/07/2015)usingcombinationofthreegroupsofkeywords(totalof57references):a:“bite”OR“chewing”OR“mastication”OR“jawmovements”OR“jaw”;b:“cattle”OR“cows”OR“heifer”OR“heifers”OR“calves”;c:“sensor”OR“electronicdevice”OR“tool”—Principaux outils utilisés pour caractériser les mouvements de mâchoire des vaches. Les références utilisées dans cette figure sont issues de recherches bibliographiques effectuées dans Scopus (www.scopus.com, Elsevier, The Netherlands, 31/07/2015) selon les groupes de mots-clés suivants (total : 57 références) : a : « bite » OR « chewing » OR « mastication » OR « jawmovements » OR « jaw » ; b : « cattle » OR « cows » OR « heifer » OR « heifers » OR « calves » ; c : « sensor » OR « electronicdevice » OR « tool ».

Jaw switches

Pressuresensors

Tools for jawmovementsdetection

Microphones

Accelerometers/Pendulum

Electromyography

Grazing durationNumber of bites

Behavior detection (grazing or not)

Duckworth et al., 1955;Stobbs et al., 1972;Chambers et al., 1981

Balch, 1958Ink record on paper

Electrical impulse

Pressure

Vibration

IGER + GRAZE software

ART-MSR Rumination sensor

Number of bitesBites intensityBite rateBites duration

Detection of chewing activityDetection, duration andnumber of ruminations

Number of bitesEstimation of intakeBehavior detection

Jaw movementsdetectionBite rateBite durationDiscrimination of jawmovements (bite, chew, chew-bites)

Jaw movements detection

Jaw movementsclassification andcountingBehavior detection(eating vs ruminating)

Jaw movements detectionDiscrimination of jawmovements (bite, chew,chew-bites)Behavior detection(eating, ruminating)

Bite rate

Behavior duration(grazing, ruminating, idling)Bite detectionNumber of bitesBite rateBehavior detection

Penning, 1983;Luginbuhl et al., 1987;Beauchemin et al., 1989

Rutter et al., 1997;Rutter, 2000

Nydegger et al., 2010

Ruckebusch et al., 1973;Lawrence et al., 1997

Delagarde et al., 1999;Unger et al., 2006b

Oudshoorn et al., 2013a

Umemura et al., 2009

Rus et al., 2013


–pressuresensors,whereajawmovementcorrespondstoachangeinpressureorlengthinatubearoundthenose;

–microphones, where sound patterns allow jawmovementdetection;

–accelerometers,todetectmovementsoperatedduringajawmovement;

–electromyography, transducing a jawmovement toanelectricalsignalfromthemovementofthemuscle.

Inthefollowingsection,onlythelastfoursensorswillbediscussed,becausejawswitchesarenotactualsensors as previously explained, since the collectedinformationisprinteddirectlyandisnottransformedinto a digital or electrical signal output (e.g. Balch,1958).

3.1. Pressure sensors

Theuseofpressuresensorsinbitemonitoringbeganwith the pioneering works of Penning (1983). Hisinstrument was composed of a halter fitted witha silicon noseband connected to two electrodes.When thenosebandstretched froma jawmovement,it changed the electrical resistance between theelectrodesplacedatbothendsofthetube.Thisinduceda change in voltage which produced a signal whichwas proportional to the extent of the jawmovementandwaving(Harman,2005).Itwasusedsuccessfullytodifferentiategrazingandruminatingbehaviorsandtomeasuretimespentgrazing,ruminatingandidling,achieving a 95%agreementwith visual observationsoveratimewindowof3min(Penning,1983).Severalvariationsof themethodexist, including initially theuseofa rubber tubeorballoonplaced justunder thelowerjaw(Luginbuhletal.,1987)butenhancedlater

as a tube encircling the nose (Penning, 1983;Rutteretal.,1997;Nydeggeretal.,2010)orunder the jaw(Dadoetal.,1993).

Building on this technology, two systems weredevelopedwhichwereusedexclusivelyinresearchandnot applied inPLF (Figure 4).The IGERBehaviourRecorder (Institute of Grassland and EnvironmentalResearch, Okehampton, UK, Rutter et al., 1997)and the ART-MSR pressure sensor (AgroscopeReckenholz-TänikonARTResearchinstitute,ModularSignal RecorderMSR145,MSR Electronics GmbH,Switzerland,Nydeggeretal.,2010)weredesignedforpasture and for stable use respectively.Both devicesareabletomakea24-hourscontinuousdatarecording,with a maximum of 40h for ART-MSR (Nydeggeretal.,2010).TheIGERBehaviourRecorderconsistsofanosebandandanelectronicinterfaceincludingarechargeable battery connected to a computer boardallowingamemorycardtorecord,analyzeandstoredataat20Hz.Ajawmovementisrecognizedasapressurepeakthroughthetransmissionofthemovementtothehalterandthechangeinthetubepressure.Thesoftwareinstalledonthecomputerboard(Rutter,2000)isableto classify jawmovements (bites or chews), identifyjaw movement bouts and determine the behaviorassociated with each bout with defined thresholdsbased on the analysis of the peaks shape. Peaks areconsideredasbiteswhentheyareacombinationofamajorlongpeakfollowedbyasmallersub-peakoranon-symmetricalpeakintheabsenceofthesub-peak(Nadinetal.,2012).Conversely,achewcontainsonlyonepeakofsymmetricalshape(Championetal.,1997).Thedetectionaccuracyreachesanoverallconcordanceof91%with95%foreatingand93%forruminatingwhen compared to the observations on a 5-min timewindow-basiswithamaximalsamplingrateof20Hz

Figure 4.Comparisonofthetwomost-usedpressuresensorstomonitorjawmovementsofcattle—Comparaison des deux capteurs de pression les plus utilisés pour le monitoring des mouvements de la mâchoire des vaches.


Jaw movementsclassification

Behavior detection


Behavior detection

IGER Behaviour Recorder (Rutter et al., 1997; Rutter, 2000)

ART-MSR(Nydegger et al., 2010)

Pressure sensors for jawmovements monitoring

Designed for pasture

Designed for stable

Average accuracy: 91%

Average accuracy: 100%

BiteChew

EatingRuminating

Other

EatingGrazingRuminatingIdling


(Rutter et al., 1997). Those accuracies decreased forcattle grazing a tropical pasture and confronted withmore heterogeneous grazing conditions, as it wasdifficult to clearly differentiate biting fromother jawmovements(Nadinetal.,2012).

Several studies (12 references) used the IGERBehaviourRecordertoinvestigatetheeffectofdifferentfactorssuchasthetimeofgrazing(Abrahamseetal.,2009), the sward height (Gibb et al., 1999; Fonsecaetal.,2013),thephysiologicalstateoftheanimal(Gibbetal.,1999)orthemilkingfrequency(O’Driscolletal.,2010) on grazing behavior or on grazing intake.Thelatest study was done by Fonseca et al. (2013) whousedgrazingbitesdetectedandcountedby theIGERBehaviourRecordertoestimatebitemassandbiteratefordeterminingtheeffectofswardheightandlevelofherbagedepletiononthesebitefeatures.

In theART-MSR system (Nydegger et al., 2010),thetubeencirclingthemouthisfilledwithoilandthesensor records the pressure change when the jaw ismoving.Whenthecowopenshermouth,an increaseinpressurealterstheelectricalresistance,resultinginasignalinthesensorcorrespondingtoonechew(Braunetal.,2014).Apart from the jawmovements featuresas described for the IGER Behaviour Recorder, itis also possible to estimate the feed intake from thenumber of chews and duration of eating time withreasonable correlation coefficients (R²= 0.6 to 0.9)(Pahletal.,2016).Accuraciesindetectingeatingandruminatingactivitiesapproached100%,whilecountingjawmovementsperformedduringeachbehaviorgavedisagreement rates of 12% and 0.24% respectivelycompared to observations performed over 5min(Nydeggeretal.,2010).Thus,thistoolismoreaccuratewhen counting ruminating jawmovements, probablyduetotheregularpatternofthisbehaviorcomparedtoeating.

Toconclude,pressuresensorsderivetheirpowerindetectingjawmovementsbyidentifyingpatternsthataredifferentduringeatingandrumination.Thesilicontubeencirclesthenoseasanormalhalter,soitdoesnotalterthenormalbehaviorof the cow.Misclassifications inthismethodusuallyarisefrompracticalconsiderations.Variation in the tighteningof thehalteron individualanimals can generate different pressure values,modifying discrimination thresholds. The analysis oftheoutputwavesignalbasedonthepeakdetectioniscompromisedifthehalterismountedtootightlyortooloosely(Nydeggeretal.,2010).Moreover,automatedtransmissionofdataisnotyetdeveloped.

3.2. Acoustic monitoring of jaw movements using microphones

Theminiaturizationandaccessibilityofdifferentkindsof sensors increased the use of microphones for the

detectionandcharacterizationofcattlejawmovements.Inamicrophone,soundsaregoingthroughaflexiblediaphragmandcausevibrations.Theoutputelectricalsignalisproportionaltotheintensityofthesevibrationsaswellastheirfrequencies.

Microphones used for recording jaw sounds of agrazingruminantcanbeusedtodiscriminatebitesorchews.Thisallowstheclassificationbetweengrazingorruminatingbehaviortobeachievedovertimewitha succession of bites or chews (Navon et al., 2013;Benvenuttietal.,2015).

Acoustic analysis allows differentiation of threetypesofjawmovements:chew,biteandchew-bite.Bitereferstoarippingsoundwhilechewreferstoagrindingsound, easing the differentiation between these twotypesofjawmovements.Chew-bitecorrespondstoanintermediatebetweenchewandbitesound,i.e.duringthisjawmovementtheherbagealreadyinthemouthischewedandsimultaneouslyanewmouthfulofherbageissevered;thesetwomovementsareperformedwithinasinglejawmovement(Ungaretal.,2006b).

Methodsusingsoundrecordingsforjawmovementsclassification differ according to the location of themicrophone,theprocessingoftheacousticsignalandtheaimofclassification(Table 1).Somemethodsareonlyabletodetectjawmovements.Forexample,basedon10minofgrazingsessionrecordedbyacamera,thesimpledetectionof jawmovementsusing amachinelearningtechniquereachesanaccuracyof94%whencomparedtotheauralanalysisofsoundsbyatrainedoperator (Navon et al., 2013).Themachine learningalgorithm uses four properties of the signal pattern:the shape to determine jaw movements interval,the intensity of each jaw movement represented bya peak in the time domain, the duration and theirintegration in a sequence of behavior (Navon et al.,2013).Claphametal.(2011)usedsimilarparameters(frequency, intensity, duration and time betweenevents) calculated during sound segments of 1 to5mintodetectandanalyzebites,reachinganoverallbehavior classification accuracy of 94%. Using adiscriminant function, bite and chew could also bedifferentiated with an accuracy of 94% (Claphametal.,2011).Thediscriminantanalysisisbasedhereonthreeparametersdeterminedfromthesignalpatternofthesoundproducedduringbitingandchewinginsidea1kHzsoundwindow:peakfrequency,peakintensity,average intensity and their duration (Laca et al.,2000).Finally,detectionandclassificationofthethreetypes of jawmovements (bite, chew and chew-bite)are possible using the Hidden Markov model. Thismodelestimatessequencesofbitesorchewsorchew-bites, called hidden states, which are not observabledirectly, using their acoustic spectrum characteristicsi.e. theenergyproduced, indecibels,byeach sound.Usingdifferentframelengths(20to80milliseconds)

8 Biotechnol. Agron. Soc. Environ. 201620(S1), AndriamandrosoA.L.H.,BindelleJ.,MercatorisB.C.N.etal.Ta

ble

1. M

ethodsforjawm

ovem

entsclassificationbasedonacousticrecordings,typeofjawm

ovem

entsdifferentiated(B:b

ites;C:chews;CB:chew-bites)and

classificationaccuracy(C

A,%

)orerrorrate(E

R,%

)overrequiredsamplingtim

ewindow—

Les

mét

hode

s de

cla

ssifi

catio

n de

s m

ouve

men

ts d

e la

mâc

hoire

bas

ées

sur

les

capt

eurs

aco

ustiq

ues,

les

type

s de

mou

vem

ents

de

la m

âcho

ire e

nreg

istr

és (

B :

bouc

hées

; C

: m

astic

atio

n ;

CB

: m

astic

atio

n-bo

uché

es)

et la

pré

cisi

on d

es

clas

sific

atio

ns (C

A, %

) ou

le ta

ux d

’err

eur (

ER, %

) sel

on la

dur

ée d

e la

fenê

tre d

’éch

antil

lonn

age.

Type

of

mic

roph

one

Loca

tion

Aud

io si

gnal

pr

oces

sing

befo

re

clas

sifica

tion

Jaw

mov

emen

tscl

assifi

catio

n m

etho

dC

ondi

tions

Acc

urac

y Ti

me

win

dow

R

efer

ence

ShureVPwireless

microphone(Shure

Incorporated,

Niles,USA

)

Forehead

None

Quadraticdiscriminant

functionusing3setsof

variable

Seta

ria

lute

scen

spasture

(tall

vssh

ortturves),

4Herefordcross-bred

steers

ForC

andB,C

A=94%

notreported

Lacaetal.,

2000

ShureVPwireless

microphone(Shure

Incorporated,

Niles,IL,U

SA)

Forehead

Transformation

intofrequency

domain(FFT

)+

normalization+

extractionof6

features

Discriminantanalysis

Loliu

m p

eren

neand

Trifo

lium

repe

nspasture,

6cows

Totalnum

berofpeaks

(CA=67%

)Totalnum

berofpeaks+

frequencybandwidth+

frequencyskew

ness

(CA=81.8%

)notreported

Ungaretal.,

2007

Logisticregression

CA=86.7%

Neuralnetworks

B:C

A=25-63%

C:C

A=65-80%

CB:C

A=70-90%

Lavalier

microphone

(SennheiserM

E2-US,Wedem

ark,

Germany)

Mouth

Highpassfilterw

ith

cutofffrequencyof

600H

z

SIGNALsoftw

are:de-

tectbitesw

ithfrequency,

intensity,durationand

timebetweenthose

events

Fest

uca

arun

dina

cea,

Dac

tylis

glo

mer

ata,

Poa

prat

ensi

sand

Trifo

lium

repe

ns

pasture,Angussteers

B:C

A=95%

Autom

atedbitecounts:

ER=9.1%

1-5min

Clapham

etal.,

2011

Nadymicrophone

(Nady151VR,

NadySystem

s,Oakland,U

SA)

Forehead

None

1-featureextraction

usingHam

mingwindow,

2-HiddenMarkovModel

Fest

uca

arun

dina

cea

and

Med

icag

o sa

tiva

(tall

vssh

ort),2Holstein

cows

B:C

A=76%

-90%

C:C

A=88%

-99%

CB:C

A=61%

-94%

13min

Miloneetal.,

2012

ShureVPwireless

microphone(Shure

Incorporated,

Niles,USA

)

Forehead

None

Machinelearningfrom

tim

edomainfeatures:

shape,sequence,

intensity,duration

Loliu

m p

eren

neand

Trifo

lium

repe

nspasture,

12Holstein-Friesian

cows

Jaw-movem

ents

identification:CA=94%

(tolerance=0.2s)

10min

Navonetal.,

2013


correct classification of those three jaw movementtypesrangedbetween61%and99%andisinfluencedby the pasture type and grass height (Milone et al.,2012).Usingdiscriminantanalysis,logisticregressionand neural networks as classification methodologiesyielded67%to82%,87%and25%to90%ofcorrectclassifications respectively, while the time windowusedinthecalculationswasnotreported(Ungaretal.,2007).

In addition, Nadin et al. (2012) showed nosignificant differences between visual observationsand microphone-based detection methods in studiesdemonstratingtheperformanceofacousticsensorstomonitorgrazing jawmovementsof cattle.Moreover,sincegrazingcorrespondstoasuccessionofbitesandchews with accessorily chew-bites and ruminationcorrespondstoasuccessionofchews,itisalsopossibleto differentiate those behaviors using microphones(Navonetal.,2013).

To summarize, microphone-based methods reachgoodaccuracyforjawmovementsdetection.Inaddition,theyareabletodifferentiatethreedifferentkindsofjawmovements:twovisiblemovementscorrespondingtochewandbite,andoneintermediatedifficulttodetectvisually (chew-bite). However, outdoor applicationsare disturbed by environmental noises, so extendingsoundrecordingandinterpretationtechniquesacquiredunderidealexperimentalconditionstoanon-farmleveltoolrecordingandanalyzingsoundsautomaticallyforPLFapplicationsstillrequiressignificantdevelopment.

3.3. Acceleration sensors

An accelerometer is an electronic sensortransforming physical acceleration from motion orgravity into waveform voltage signal output. It canmeasure both static acceleration due to gravity, thelow-frequencycomponentoftheaccelerationandthedynamicaccelerationduetomovementsimprintedbytheanimal(Almeidaetal.,2013;Brownetal.,2013).

Despitethedepthofliteraturesurveyed,onlyfourreferences used this type of sensor to identify andclassifyjawmovements,amongwhichthreedescribeddifferentmethodsofclassification.Instables,Tanietal.(2013)coupleda1-axisaccelerometertoamicrophoneto classify cattle chewing activities by matching1minute segment waveform patterns to observedeating and ruminating behaviors. Intake chewingactivitieswerehighlydistinguishedat90%, reaching99%whenthesensorwasattachedtothecow’shorn.

On pasture, a 3-axis accelerometer was usedby Oudshoorn et al. (2013b) to record cow bites.A visualization of recorded signals from the threeindividualorthogonalaxes(x,y,z)wasdonefirst, inorder to determinewhich onematched bestwith theobserved bites. To determine each bite, a series of

thresholds were tested to determine the peak whichhadthebestcorrelationtotheobservation.Theaveragecorrelationcoefficientwas0.65indicatingthedifficultytocountbitesusinganaccelerometerthisway.Finally,Umemura et al. (2009)modified a pedometer into apendulumunder the lower jaw tomonitor cattle jawmovements.Thedatacouldbewirelesslydownloadedfromthesensorwhichhadalifespanofoneyear.Thissystem was able to count jaw movements with anaccuracyof90%comparedwithmanualcountsover10minsegments.Thisbitecountwascorrelatedwithacoefficientof0.7topasturedisappearanceestimatedindirectlybyarisingplatemeter.

Accelerometer sensors thus provided interestingoptionstoautomaticallycountcattlejawmovements.As for soundsensors, interferencemaybepresent inthesignalrecordedbythesensor.Bitesaretheresultofjawandheadmovementswhilechewimprintsmostlyjaw movements. The sensitivity of accelerometerscould provide undesirable signals during recordingsessionsduetoearmovementsorsuddenheadturnstodrivefliesorotherinsectsaway.Thus,apre-processingofthesignalisprobablyrequiredtoisolatethesignalrelative to the jaw movements in order to considerthismethodforactualPLFusesonfarms.Intermsofmaterial,theuseofaninertialmeasurementunit(IMU)whichisacombinationofaccelerometers,gyroscope,magnetometerandlocationsensorsmightofferarealadvantage knowing that all these variables can berecordedsimultaneouslyatahighsamplingfrequency(100Hz).Forexample,Andriamandrosoet al. (2015)used smartphone IMUs to count thenumberof bitesthroughfrequencypatternof1-axisaccelerationdata.Asforpressuresensors,iftheaccelerometerismountedinahalter,thetighteningalsoplaysanimportantroleinthetransmissionofthemovementtothesensor.

3.4. System based on electromyography

While the noseband pressure sensor quantifieschanges in tube pressure and translates it into anelectricalimpulse,electromyographicsensorsquantifytheelectricalpotentialofmasticatorymusclesduringcontractions (Rus et al., 2013). Two electrodes arefixed on a halter and measure electrical signalsoccurringduringajawmovementwithacontractionoftheMassetermuscle.Thissensor,coupledtoa3-axisaccelerometer and awireless transmission of data inreal time, constitutes the DairyCheck sensor (Rusetal.,2013).Thissystemisabletodetectruminatingand feeding behavior on the basis of the regularityand irregularity of signal pattern respectively. TheDairyChecksystemyieldedanoverallconcordanceof87%compared to visual observations over 1minute,when detecting feeding time and rumination time(Bücheletal.,2014).


Finally, when the different above-mentionedtechniques are compared, noseband pressure sensorsandmicrophonesarebestabletodetectjawmovementswithhighaccuracy(over94%)anddifferentiatebitingandchewingjawmovementswithfairaccuracy(61%to 95% for microphone) (Table 2). Accelerometerscan identify jaw movements with less than 90% ofaccuracy but discrimination of the different typesof jaw movements is not mentioned yet. Authorsusing theelectromyographymethoddidnotgiveanyinformation about the accuracy of the detection ofspecificjawmovements.

4. ESTIMATION OF GRAZING INTAKE THROUGH BITES QUANTIFICATION

Formany years, variousmethods have been used toquantifyforageintakeofgrazingherbivores,includingthemeasurementofpasturebiomassbeforeandaftergrazing, changes in animal bodyweight, digestivemarkers,orfecalnear-infraredreflectancespectroscopy(reviewed by Decruyenaere et al., 2009) (Figure 5).

Amongthesetechniques,animalbehaviorcanalsobeusedasameasurementofgrazingintakebycombininggrazingduration,biting rateandbitemass.Knowingthatabiteistheelementaryandindivisibleunitofthewholegrazingprocess,thistechniquestressestheneedtoproperlyquantifybites,toestimatetheintakealongwiththemassofeachindividualbiteortheiraveragemass.

Thismethodisbasedonthedeterminationofbitemass(quantity, ingrams,ofgrasstakenineachbite)andbite rate (numberofbitesperminute)asper thefollowingformula(Vallentine,2000):

Forageintake(g.day-1)=Bitemass(g.bite-1)xBiterate(bite.min-1)xGrazingactivityduration(min.day-1)

Toefficientlyusethisformula,accuratedetectionsofgrazingbehaviorandindividualbitesareessential.As already mentioned, several sensors are able toquantify these parameters with various accuracies.Oudshoornetal.(2013a)correlatedintakeforgrazingcattlefromgrazingtimeestimatedbyanaccelerometerand bite frequency with a prediction precision of

Table 2.Comparisonofdifferenttypesofsensorstodetectandclassifyjawmovements—Comparaison des différents types de capteurs utilisés pour détecter et classifier les mouvements de la mâchoire.Type of sensor Jaw movements

detection accuracy1 (%)Jaw movementsclassification (accuracy1in%)

Required sampling time window (min)

References

Nosebandpressuresensors

91-95 BiteChew 5min Rutteretal,1997;

Nydeggeretal.,2010Microphones 94-95 Bite(76-95)

Chew(88-94)Chew-bite(61-94)

1-13minUngaretal.,2007;Claphametal.,2011;Miloneetal.,2012

Accelerometers 65-90 (datanotprovided) 10min Oudshoornetal.,2013a1:Comparisonwithvisualobservations—comparaison par observations visuelles.

Figure 5.Listoftechniquesforgrazingintakemeasurement(fromDecruyenaereetal.,2009)—Liste des techniques utilisées pour la mesure de l’ingestion lors du pâturage (selon Decruyenaere et al., 2009).

Method of difference: before and after grazingCutting method: grazing simulationLive weight difference

Indigestible plant compound

Markers Chemical marker

Radio technique: forage digestibility + fecal outputAnimal behavior: grazing time, biting rate, bite massEmpirical technique: modelsAnalysis of plant residues in digestive tractNIRS

Direct measurements

Indirect measurements

Measurement of grazing intake


less than 1.4kg of dry matter per cow per day. Inthis experiment, grass intake was initially measuredusing an indigestible marker and the difference innet energy balance between energies offered by thegrassandrequiredforanimalneeds.Thispapershowsthat prediction of intake from grazing behavior andbites counts is still beyond reach due to the lack ofaccuratebitemassestimation.Indeed,beyondbiterate,forage intake of grazing animals depends on pasturecharacteristics (sward height and bulk density) asexpressed by the bitemass formula (Carvalho et al.,2015):

Bite mass (g.bite-1) = Bite area (cm²) x Bite depth(cm)xBulkdensity(g.m-3)

Inthisformula,thebulkdensityandthebiteareaarecalculatedfromempiricalmodelsusing,swardheightandrelativebitedepth(swardheight/2),tillerdensityandforagebiomassperareaunit,and thesizeof thedentalarcade.Suchpredictivemodelsyieldacceptablebitemassestimatesinshort-termexperimentswithveryhomogenousvegetationcharacteristics(Carvalhoetal.,2015),butfailinmorecomplexvegetationunits.Untilnow,thebestmethodavailableremainshand-plucking.Itsimulatesabitebymimickinggrassprehensionbyhand and bite mass estimation accuracies can be ashighas95%forcowsandgoatswithtrainedoperators.Thisaccuracycorrespondedtothecorrelationbetweentheamountofgrasstakenbytheanimalperbiteandthose pluckedmanually (Bonnet et al., 2011).Whilethismethodseemsuseful tocalibratebitemassesforintakemeasurementsforresearchpurposes,itistime-consumingandnosensorscanperformasimilartask,henceitseemsuselessfromaPLFperspective.

Therefore,monitoringintakeingrazingruminantsusingPLFapproachesisstilloutofreach.Nonetheless,from a research perspective, the bite mass formulaprovides interesting directions for future research onthequantificationofbitemassusingacombinationofpromisingtechnologiessuchas:–accuratemonitoringofthepasturewithforexampledistancemetersand/ortime-of-flightcameras,

–animal positioning with wifi triangulation orcentimeter-accurateGPS,

–rigid body attitude estimation from accelerometerdatatoreconstructheadmovement,

–jaw monitoring using sensor-based technologiesdescribedinthispaper.

5. CONCLUSIONS

Mostsensorsmentionedinthisreviewwereprimarilydesigned for research. Although some sensors suchas accelerometers (e.g. SensOor®, CowManager,

Utrecht, The Netherlands) are already used forbehaviorclassificationinfarmsituations,thereuseasonfarmtoolsforjawmovementsmonitoringofgrazinganimals still requires significant hard- and softwaredevelopments, especially regarding the automationprocessandreal-timedataacquisition,aswellaseaseof installation and use. For example, whether basedonmechanical(pressureoracceleration),electrical,oracousticsignals,mostsensorsrequiretheuseofahalter,andthewayitismountedisextremelyimportantintherecordingofjawmovements.Apre-processingofthesignalmayalsoberequiredtoeliminateexistingnoisesaroundtheanimalorduringthemovement.Combiningdifferent sensors, for example accelerometers tomicrophones,maybeasolutionforabettermonitoringof bites. Dedicated signal processing also requiressignificantdevelopment.Forexample,usingfrequencydomain signalprocessingapproachesonaccelerationdata might provide useful progress. Accuraciesmentionedinthisdocumentwereobtainedusingshorttimewindowsanddifferentcalculationmethodswhichcould affect the percentage values. Longer time stepshouldbeconsidered tocoveroneorseveraldays inordertoshowifthedetectionaccuracywouldchangesignificantlyornot.Finally,PLFrequiresthesystemtoberobustandadaptabletoawiderangeofsituations.Most techniques presented here were applied understrict controlled conditions for research and theirimplementationinthefarmswouldalsorequiresomeability for auto-calibration of the device or tools toovercome differences in individual physiologicalstates,morphologies or grazing conditions accordingtotheseasonandpasture.

Acknowledgements

ThisreviewisinvolvedinaresearchfundedbytheCAREAgricultureIsLife(TERRAResearchUnit,GemblouxAgro-BioTech,UniversitédeLiège).The authors specially acknowledge Jeffrey Peter Corfield(Corfield Consultants, Australia) for corrections andsuggestionshegavetoourreview.

Bibliography

AbrahamseP.A.,TammingaS.&DijkstraJ.,2009.Effectofdailymovementofdairycattletofreshgrassinmorningor afternoon on intake, grazing behaviour, rumenfermentation andmilk production. J. Agric. Sci., 147,721-730.

AlmeidaP.R. et al., 2013. Testing a 3-axis accelerometeracoustic transmitter (AccelTag) on the Lusitaniantoadfish.J. Exp. Mar. Biol. Ecol.,449,230-238.

AndersonD.M.etal.,2012.Characterisingthespatialandtemporalactivitiesoffree-rangingcowsfromGPSdata.Rangeland J.,34,149-161.


AndriamandrosoA.L.H., LebeauF. & BindelleJ., 2014.Accurate monitoring of the rumination behaviour ofcattle using IMU signals from a mobile device. In:HopkinsA.etal.,2014.Proceedings of the 25th General meeting of the European Grassland Federation, EGF at 50: the future of European grasslands, 7-11 September 2014, Aberystwyth, Wales.

AndriamandrosoA.L.H., LebeauF. & BindelleJ., 2015.Changes in biting characteristics recorded using theinertial measurement unit of a smartphone reflectdifferences in sward attributes. In: GuarinoM. &BerckmansD., 2015.Proceedings of the 7th European conference on Precision Livestock Farming, Precision Livestock Farming ’15, 15-18 September 2015, Milan, Italy,283-289.

BalchC.C.,1958.Observationsontheactofeatingincattle.Brit. J. Nutr.,12(3),330-345.

BarrettP.D.,LaidlawA.S.,MayneC.S.&ChristieH.,2001.Pattern of herbage intake rate and bite dimensions ofrotationallygrazeddairycowsasswardheightdeclines.Grass Forage Sci.,56,362-373.

BeaucheminK.A., ZelinS., GennerD. & Buchanan-SmithJ.G.,1989.Anautomaticsystemforquantificationofeatingandruminatingactivitiesofdairycattlehousedinstalls.J. Dairy Sci.,72,2746-2759.

BenvenuttiM.A.etal.,2015.Defoliationpatternsandtheirimplicationsfor themanagementofvegetativetropicalpastures to control intake and diet quality by cattle.Grass Forage Sci.,doi:10.1111/gfs.12186

BonnetO.J.F. et al., 2011. Is hand plucking an accuratemethodofestimatingbitemassandinstantaneousintakeofgrazingherbivores?Rangeland Ecol. Manage.,64(4),366-374.

BonnetO.J.F. et al., 2015. Continuous bite monitoring: amethodtoassesstheforagingdynamicsofherbivoresinnaturalgrazingconditions.Anim. Prod. Sci.,55,339-349.

BraunU., TschonerT. & HässigM., 2014. Evaluation ofeating and rumination behaviour using a nosebandpressure sensor in cows during the peripartum period.BMC Vet. Res.,10,195.

BrownD.D. et al., 2013. Observing the unwatchablethroughaccelerationloggingofanimalbehavior.Anim. Biotelem.,1,1-16.

BüchelS.&SundrumA.,2014.Technicalnote:evaluationofanewsystemformeasuringfeedingbehaviorofdairycows.Comput. Electron. Agric.,108,12-16.

CarvalhoP.C.deF.,2013.HarryStobbsmemoriallecture:can grazing behavior support innovations in grasslandmanagement?Trop. Grasslands,1,137-155.

CarvalhoP.C. de F. et al., 2015.Can animal performancebepredictedfromshort-termgrazingprocesses?Anim. Prod. Sci.,55,319-327.

ChambersA.R.M., HodgsonJ. & MilneJ.A., 1981. Thedevelopment and use of equipment for the automaticrecording of ingestive behaviour in sheep and cattle.Grass Forage Sci.,36,97-105.

ChampionR.A., RutterS.M. & OrrR.J., 1997.Distinguishing bites and chews in recordings of thegrazingjawmovementsofcattle.In:Proceedings of the 5th British Grassland Society, BGS Research Meeting, September 1997, Seale Hayne, United Kingdom, 171-172.

ClaphamW.M., FeddersJ.M., BeemanK. & NeelJ.P.S.,2011.Acousticmonitoringsystemtoquantifyingestivebehavior of free-grazing cattle. Comput. Electron. Agric.,76(1),96-104.

DadoR.G. & AllenM.S., 1993. Continuous computeracquisitionoffeedandwaterintakes,chewing,reticularmotility, and ruminal pH of cattle. J. Anim. Sci., 76,1589-1600.

DecruyenaereV.,BuldgenA.&StilmantD.,2009.Factorsaffecting intake by grazing ruminants and relatedquantification methods: a review. Biotechnol. Agron. Soc. Environ.,13(4),559-573.

DelagardeR., CaudalJ.P. & PeyraudJ.L., 1999.Development of an automatic bitemeter for grazingcattle.Ann. Zootec.,48,329-339.

DuckworthJ.E.&ShirlawD.W.,1955.Thedevelopmentofanapparatustorecordthejawmovementsofcattle.Br. J. Anim. Behav.,3(2),56-60.

DuttaR. et al., 2015. Dynamic cattle behaviouralclassification using supervised ensemble classifiers.Comput. Electron. Agric.,111,18-28.

FonsecaL.etal.,2013.Effectofswardsurfaceheightandlevel of herbage depletion on bite features of cattlegrazingSorghumbicolorswards.J. Anim. Sci.,91,4357-4365.

GibbM.J., 1996. Animal grazing/intake terminology anddefinitions. In: Proceedings of pasture ecology and animal intake workshop for concerted action AIR3-CT93-0947, 24-25 September 1996, Dublin, Ireland,20-35.

GibbM.J., HuckleC.A., NuthallR. & RookA.J., 1997.Effect of sward surface height on intake and grazingbehaviour by lactating Holstein Friesian cows. Grass Forage Sci.,52,309-321

GibbM.J.,HuckleC.A.,NuthallR.&RookA.J.,1999.Theeffectofphysiologicalstate(lactatingordry)andswardsurfaceheightongrazingbehaviourandintakebydairycows.Appl. Anim. Behav.Sci.,63(4),269-287.

GregoriniP., TammingaS. & GunterS.A., 2006. Review:behavioranddailygrazingpatternsofcattle.Prof. Anim. Sci.,22,201-209.

GriffithsW.M. & GordonI.J., 2003. Sward structuralresistanceandbitingeffortingrazingruminants.Anim. Res.,52,145-160.

HarmanG., 2005. Pressure sensors. In: WilsonJ.S.,ed. Sensor technology handbook. Amsterdam,TheNetherlands:Elsevier,411-456.

HolechekJ.L., PieperR.D. & HerbelC.H., 2011. Range management. Principles and practices. 6th ed. Boston,USA:PrenticeHall.


HostiouN. et al., 2014. L’élevage de précision: quellesconséquencespourletravaildeséleveurs?INRA Prod. Anim.,27(2),113-122.

KennyT., 2005. Sensor fundamentals. In: WilsonJ.S.,ed. Sensor technology handbook. Amsterdam,TheNetherlands:Elsevier,1-20.

LacaE.A.,2009.Precisionlivestockproduction: toolsandconcepts.Rev. Bras. Zootec.,38,123-132.

LacaE.A.&DeVriesM.F.W.,2000.Acousticmeasurementofintakeandgrazingbehaviourofcattle.Grass Forage Sci.,55,97-104.

Larson-PraplanS., GeorgeM.R., BuckhouseJ.C. &LacaE.A.,2015.Spatialandtemporaldomainsofscaleofgrazingcattle.Anim. Prod. Sci.,55,284-297.

LawrenceP.R. & BeckerK., 1997. The use of vibrationanalysis and telemetry to measure bite frequencyand intensity in free-ranging horned ruminants. In:Proceedings of XVIII IGC Conference, 1997, Winnepeg, Manitoba, Canada,Session5,3-4.

LuginbuhlJ.M.,PondK.R.,RussJ.C.&BurnsJ.C.,1987.Asimpleelectronicdeviceandcomputerinterfacesystemformonitoring chewingbehavior of stall-fed ruminantanimals.J. Dairy Sci.,70,1307-1312.

MangwethG. et al., 2012. Lameness detection in cowsby accelerometric measurement of motion at walk.Berl. Munch. Tierarztl. Wochenschr., 125(9-10), 386-396.

MartiskainenP. et al., 2009. Cow behaviour patternrecognition using a three-dimensional accelerometerand support vectormachines.Appl. Anim. Behav. Sci.,119(1-2),32-38.

MiloneD.H.etal.,2012.Automaticrecognitionofingestivesounds of cattle based on hidden Markov models.Comput. Electron. Agric.,87,51-55.

NadinL.B.etal.,2012.Comparisonofmethodstoquantifythenumberofbites incalvesgrazingwinteroatswithdifferentswardheights.Appl. Anim. Behav. Sci.,139(1-2),50-57.

NavonS., MizrachA., HetzroniA. & UngarE.D., 2013.Automaticrecognitionofjawmovementsinfree-rangingcattle, goats and sheep, using acoustic monitoring.Biosyst. Eng.,114(4),474-483.

NydeggerF., GygaxL. & EgliW., 2010. Automaticmeasurement of rumination and feeding activityusing a pressure sensor. In: International Conference on Agricultural Engineering-AgEng 2010: towards environmental technologies, 6-8 September 2010, Clermont-Ferrand, France.Cemagref.

O’DriscollK., O’BrienB., GleesonD. & BoyleL., 2010.Milking frequency and nutritional level affect grazingbehaviour of dairy cows: a case study. Appl. Anim. Behav. Sci.,122(2-4),77-83.

OudshoornF.W.etal.,2013a.Estimationofgrassintakeonpasturefordairycowsusingtightlyandlooselymounteddi- and tri-axial accelerometers combined with bitecount.Comput. Electron. Agric.,99,227-235.

OudshoornF.W. & JorgensenO., 2013b. Registrationof cow bites based on three-axis accelerometer data.In: BerckmansD. Proceedings of the 6th European Conference on Precision Livestock Farming, Precision Livestock Farming ’13, 10-12 September 2013, Leuven, Belgium, 771-777.

PahlC.etal.,2016.Suitabilityoffeedingandchewingtimeforestimationoffeed intake indairycows.Animal, inpress.

PenningP.D., 1983. A technique to record automaticallysome aspects of grazing and ruminating behaviour insheep.Grass Forage Sci.,38(2),86-96.

RookA.J., HuckleC.A. & PenningP.D., 1994. Effectsof sward height and concentrate supplementation onthe ingestive behaviour of spring-calving dairy cowsgrazing grass-clover swards. Appl. Anim. Behav. Sci.,40(2),101-112.

RuckebuschY.,BuenoL.&LatourA.,1973.Undispositifsimple et autonome d’enregistrement de l’activitéalimentairechezlesbovinsaupâturage.Ann. Rech. Vet.,4,627-636.

RusM.A., WobschallA., StormS. & KaufmannO.,2013. DairyCheck – a sensor system for monitoringand analysis of the chewing activity of dairy cows.Landtechnik,68(6),395-398.

RutterS.M.,2000.Graze:aprogramtoanalyzerecordingsofthejawmovementsofruminants.Behav. Res. Meth. Instrum.Comput.,32(1),86-92.

RutterS.M., ChampionR.A. & PenningP.D., 1997. Anautomaticsystemtorecordforagingbehaviour infree-rangingruminants.Appl. Anim. Behav. Sci.,54,185-195.

SchlechtE.,HülsebuschC.,MahlerF.&BeckerK.,2004.The use of differentially corrected global positioningsystem tomonitor activities of cattle at pasture.Appl. Anim. Behav. Sci.,85(3-4),185-202.

StobbsH.T.&CowperL.J.,1972.Automaticmeasurementofthejawmovementsofdairycowsduringgrazingandrumination.Trop. Grasslands,6(2),107-112.

TaniY.,YokotaY.,YayotaM.&OhtaniS.,2013.Automaticrecognitionandclassificationofcattlechewingactivityby an acoustic monitoring method with a single-axisacceleration sensor.Comput. Electron. Agric., 92, 54-65.

UmemuraK.,WanakaT.&UenoT.,2009.Technicalnote:estimationoffeedintakewhilegrazingusingawirelesssystem requiring no halter. J. Dairy Sci., 92(3), 996-1000.

UngarE.D. et al., 2006a. The implications of compoundchew–bitejawmovementsforbiterateingrazingcattle.Appl. Anim. Behav. Sci.,98(3-4),183-195.

UngarE.D. & RutterS.M., 2006b. Classifying cattle jawmovements:comparingIGERBehaviourRecorderandacoustic techniques.Appl. Anim. Behav. Sci., 98(1-2),11-27.

UngarE.D., BlankmanJ. & MizrachA., 2007. Theclassification of herbivore jaw movements using


acoustic analysis. In: CoxS., 2007. Proceedings of the 3rd European Conference on Precision Livestock Farming, Precision Livestock Farming ’07, 3-7 June 2007, Skiathos, Greece,79-85.

VallentineJ.F., 2000. Grazing management. Amsterdam,TheNetherlands:Elsevier.

WalkerS.L. et al., 2008.Lameness, activity time-budgets,and estrus expression in dairy cattle. J. Dairy Sci.,91(12),4552-4559.

(64ref.)

A review on the use of sensors to monitor cattle jaw ... · B A S E Biotechnol. Agron. Soc....

Documents

Transcript of A review on the use of sensors to monitor cattle jaw ... · B A S E Biotechnol. Agron. Soc....