alarm sounds based on human factors
TRANSCRIPT
AlarmsoundsbasedonHumanFactors
NatashaBarrett
HFCWEBseminar,oktober 2021
AlarmsoundsforEquinor's Hywind Tampen controlroom.
Challengesandsolutionsarerelevanttodifferentsituations.
Background:
Problemswithtoday'scommonalarms:
Thesounds:• Unpleasantaurally• Unpleasantexperientially• Canbetooloud• Canbetoostartling• Couldbemoreinformative
Theresult:• Silencethealarmassoonaspossible
• Disturbancetocommunication
• Workfatigue
• (Mutingthecompletealarmsystem!)
• (Addingpersonalsounds!)
Whatshouldalarmsbelike?• Indicatewhatresponseisrequiredwithoutthenegativeeffectsofcurrentalarmssounds.
• Beeasytounderstand.
• Unique
• Bedesignedinkeepingwiththecontrol/notificationsystem(whichinturnshouldconsiderhumanlimitations).
Whatelsecouldalarmsdo?• Indicatewhichconsoleissoundinginamulti-consoleenvironment
• Providemoreinformationaboutthetypeorclassofalarm.
Newalarmsounds:Function|Sound-design|Acoustics|Workenvironment|Perception
Pastwork:Patterson(1982)
MostofthepastworkinalarmsoundbuildsonideaspublishedbyRoyPattersonin1982:
• Aburstofsoundwithitsownsetoffrequenciesintherange150–3000HzatadBlevel.
• Arepeatedpatternofsound-pausedurations
• Differentlevelsofpriorityareindicatedbychangesinthesound-pausepattern
• Subsequentvariationsonthisideahaveaddedpitch/notepatternstothesound-pausepattern.
Thesesolutionsneglectsomeobviousproblems:Soundswerechosento'cutthrough'noisyenvironments:
Backgroundnoisehasdecreased.
Thesuggestedfrequenciesarelocatedinourmostsensitivehearingrange.
• Toavoidconfusinglevelsofurgency,fastnotesequencesareneeded.Thesecreatestress.
• Development:notesequencesareshowntocreateconfusionratherthanaddclarity.
• Alackofdesigncompoundsnegativefactorsconcerninguninterestingsounds.
Pastwork:Alarmsand'earcons'• Soundsthatmayrangefromrepresentational(resemblingreal-worldsounds)to
abstract(puretones,buzzersorinstrumentalsounds).
• Theideaisthatthereissomeidentifiablecontentthatovercomestheproblemsassociatedwithlearningthemeaningofsingletones.
• Experimentalworkonrepresentationalsounds:lessappropriateintheworkplace.• Toostrange• Experiencedasannoying• Spectrally(frequencycontent)notclearenough
SpecificationfromEquinor• ’Alarm’sounddesigntosignaldifferencelevelsofpriority• ‘Earcons’sounddesignasindicationsounds.
‘Alarms’and'earcons'?
Theideaof‘earcons’isobsoleteifweinsteadassumethatouralarmsovercomestheproblemsassociated
withlearningthemeaningofsingletones.
Alarmsandnotifications
Basicdesignconsiderations
Environment:• Backgroundnoisesand
othersounds.• Peopleinteraction.• Peoplelocation.• Acoustics.• Loudspeakerlocations.
Perceptionandaction:• Audibility.• Identity.• Affect.• Repetition.• Redundancy.• Interaction(silencing
alarm).
Designprocess
Moveawayfrom:
• Extremedifferencesinpatternspeedreflectingurgency.
• Attackingthemostsensitivefrequencyareaofourhearing.
• Simple/boringnotesequences.
Insteadask:
• Whatkindofsoundsweareusedtohearing?
• Whatkindofsoundswecanacceptasalarmsounds?
• Howwouldthedifferentsoundsfittogetherasaset?
• Howtooptimisethevolumeandspectrumonsitebysimplecalibration?
• Urgency/priorityasrelativeratherthanabsolute(i.e.onlyarchetypessuchasfirealarmbellsandthelikeshouldbeabsolute).
Choiceofsoundtypes
• Acousticsourcesandtheirtransformations(identity/uniqueness).• Synthesisedsources(identity/genre).• Pitchrangesandspectraldistribution.
Transformation
Digitalsignalprocessingacoustic synthetic
Morphology,identityandinterestAmplitude
Spectrum
attack
decay
sustainrelease at
tack Therest..
Firstsetofsounds• Aselectionofacousticsoundstransformedwithdigitalsignalprocessing.
• Aselectionofsynthesisedsoundstransformedwithdigitalsignalprocessing.
• Importanceofcontrast(evenifitwasexpectedthatsoundsoundswouldappeartoostrangeformostlisteners).
examples
Testingstage1:privatetestingonmyself• Environmentalemulation
• Loudspeakerthatwillbeusedinthecontrolroom
• Combinations/permutations
• Repetitionandpatterns
Result:shortlist
Testingstage2:Testset-up• Testingwithreallistenersinaworkenvironment:four/fivepeopleondifferentshifts.
• Howto'rate'thealarms?statisticalapproachese.g.pairwisecomparisonsareproblematicwithoutalargenumberoflistenersandspecialBayesianapproachestotestfor‘self-contradictions’.
• Asklistenerstoexplainwhytheylikedonesoundmorethananotherusingtheirownwords.
• UseadBmetertoensureanapproximatelycorrectlisteninglevel.
ResultsListenercomments:
• Likeordislikeofwhatalistenerassociatedthesoundwith.• Suggestionofthesound'simpliedpriority.• Descriptionofhowstressful/relaxedthesoundfelt.• Descriptionofpartsofthesoundthatwerelikedordisliked(andwhatitmadethe
listenerfeel).• Whetherthesoundwassimplypleasanttohear,irritating,orunpleasant.• Howwellthesoundcontrastedorfittedwiththeothersoundsintheworkplace.
Results• Less‘real-world’soundsweremoresuccessful.• Thefollowingsimpleradjustmentschangedtheresults:• Listeningvolume:morphologyallowedreducedvolume.• EQ(spectralbrightness):tunedforthespeakersandacoustic.• Repetitionspeedimplyingurgencyanditsconnectiontostress.
Secondtestset• Reducedlistofsounds.
• Newresults:somecontradictiontothefirstresultsforthesamesoundpresentedinanewwaytothesamelisteners.
• Sound-pausesequenceswherenotonlyabouturgency,butabout‘sustainability’.
• Smallchangesindurationandmorphologywereimportant:i.e.playthedurationthatisneededtocommunicatethemessage.
Finalsetoffoursounds
Futurepotential
Sharedcontrolroomswithuniquealarms.
Acontrolstationaudiofingerprint.
Alarmscouldsoundsimultaneouslyandnotleadtoconfusionorstress.