In e na ional Semina on Vi ual Acous ics VALENCIA, No embe 24-25, 2011
153
PERCEIVED FEASIBILITY OF COMPUTER-GENERATED AURALIZATION
IN CONCERT HALLS
Re e ence PACS: 43.55.Gx , 43.55.Hy, 43.55.Ka
Ga ido, José A.1; Zama eño, Teó ilo2; Gi ón, Sa a2; Ruiz-Padillo, Diego P.1
1Dp o. de Física Aplicada, Uni e sidad de G anada, Facul ad de Ciencias, A da. Fuen e Nue a s/n,
18071-G anada (SPAIN) Tel.: +34 958244161. Fax: +34 958243214.
2Dp o. de Física Aplicada II, Uni e sidad de Se illa, ETS de A qui ec u a, IUACC, A da. de la Reina
Me cedes 2, 41012-Se illa (SPAIN) Tel.: + 34 954556612. Fax: +34 954557892
[email p o ec ed], [email p o ec ed], [email p o ec ed], [email p o ec ed]
ABSTRACT
O e he yea s au aliza ion has become a use ul ool o simula ing and e alua ing he lis ening expe ience
in i ual en i onmen s. Psychoacous ic phenomena, embodied by he human hea ing sys em, highly
de e mine he accu acy o sound- ield ec ea ion ha is equi ed o seemingly easible au aliza ion. In his
scena io, he main aim o his s udy is o assess he sui abili y o compu e -gene a ed oom impulse
esponses when used o au alizing spaces. To his pu pose, simula ed and expe imen ally measu ed
binau al oom impulse esponses (BRIRs) ha e been con ol ed wi h iconic musical exce p s o a ious
ep esen a i e sea s wi hin a medium-sized conce hall. Lis ening es s ha e been he eby conduc ed and
hei esul s will be shown, analyzed and discussed.
RESUMEN
En es os úl imos años, la au alización se ha con e ido en una he amien a ú il pa a la simulación y
e aluación de la expe iencia del oyen e en en o nos i uales. En es e ámbi o, son de especial
impo ancia los aspec os psicológicos y pe cep i os, de i ados del uncionamien o del sis ema audi i o y
del ce eb o humanos, que de e minan la p ecisión que se equie e pa a una ec eación apa en emen e
c eíble del campo sono o. En es e con ex o, el p incipal obje i o de es e es udio es e alua la alidez de
espues as al impulso de la sala gene adas po o denado pa a su u ilización en la au alización de
espacios. Con es e p opósi o, pa a algunas localidades ep esen a i as de una sala de concie os de
amaño medio, se han con olucionado a ias espues as al impulso binau ales (BRIRs), simuladas y
medidas expe imen almen e, con agmen os musicales conocidos. Con los da os ob enidos, se han
lle ado a cabo p uebas de audición, cuyos esul ados se p esen an, analizan y discu en en es a
comunicación.
1. INTRODUCTION
Al hough au aliza ion has been p o ed an impo an ool o mul iple applica ions in ields as
di e se as A chi ec u e, Telecommunica ions, Psychology o Vi ual Reali y, e y li le is s ill
known abou he accu acy o he ou pu audio iles gene a ed by he comme cial compu e
p og ammes ea u ing his abili y ha a e cu en ly a ailable in he ma ke . T ying o e alua e
he quali y o he pe o mance achie ed by some o hese pieces o p edic ion so wa e when
con on ed wi h in si u measu emen s, h ee ound- obin es s ha e been held in his o y
[1][2][3][4]. Howe e , none o hem emb aced au aliza ion wi hin i s scope.
When i comes o au aliza ion, physical accu acy does no cons i u e he only co ne s one o
his kind o so wa e, bu achie ing co ec ness in he lis ening expe ience e en ually c ea ed
In e na ional Semina on Vi ual Acous ics VALENCIA, No embe 24-25, 2011
154
also plays a decisi e pa [5]. The e o e, Psychoacous ics ough o ake o e he ask o d awing
he uzzy line ha de e mines whe he he sound ield c ea ed wi h his so wa e is p ecise
enough o gene a e audi o y e en s deemed co ec by a majo i y o lis ene s.
In o de o shed some ligh on he easibili y o compu e -aided au aliza ion, a h ee-dimension
geome ic model o a 1,311-sea conce hall, whose acous ic pe o mance had al eady been
alida ed o he mos common pa ame e s, was simula ed o ob ain impulse esponses a
a ious loca ions sp ead o e he audience a ea. Expe imen al measu emen s p e iously done
a analogous posi ions in he ac ual oom had al eady ga he ed equi alen da a. Finally,
au alized audio iles co esponding o impulse esponses om bo h sou ces ha e been
p oduced and subjec ed o compa ison wi h he assessmen o a panel o un ained indi iduals,
who will ul ima ely look o di e ences be ween s imuli h oughou lis ening es s.
Likewise, au aliza ions coming om simula ing he conce hall unde changing acous ical
condi ions because o e u bishmen wo ks ha e been judged agains hose a ained in he
o me s a e. In his case, an e en ual s a emen will go in he necessi y o he unde aking.
2. MANUEL DE FALLA AUDITORIUM
Fi s inaugu a ed on June 10, 1978, Manuel de Falla Audi o ium in G anada (Spain) could be
p oudly coun ed amongs a selec ed g oup o Spanish conce halls wi h ema kable acous ic
quali ies. The audi o ium is loca ed a La Sabika hill beside he ca men once inhabi ed by he
Gadi an compose a e whom he edi ice was named, and wi hin he a ea o in luence o La
Alhamb a, a wo ld-wide admi ed landma k. Se illian a chi ec José Ma ía de Pa edes along wi h
enowned Ge man acous ician Lo ha C eme accoun espec i ely o he a chi ec u al design
and acous ical consul ancy owa ds he execu ed bluep in s o he building6.
As illus a ed by he pho og aph in Figu e 1(a), he oom ea u es a o und bila e al symme y on
whose axis he s age is disposed a nea ly wo- hi ds o he way, p o iding an uncommon
a angemen o he audience in o wo main a eas, A and B –labelling sea s in on o and
behind he s age in ha o de -. One o he main ad an ages de i ed om his unusual plan lays
on he adap abili y o he enue o spec acles o a ious kinds. Changeable capaci y and
olume allow o se e al con igu a ions and, he eo , a iable acous ics.
As shown in he g ound plan o he audi o ium displayed in Figu e 1(b), h ee sea s wi hin he
oom ha e been ca e ully chosen o ca y ou his wo k. I mus be said ha ecei e loca ions
close o ei he he sou ce o la ge e lec ing su aces ha e been uled ou since expe imen al
measu emen s a such in ica e posi ions usually yield de ec i e esul s. To p ope ly sample he
oom, wo o hem -no ed A-1 and A-2- ha e been placed in on o he s age whe eas only one
–namely B- has been loca ed in he backs age.
In ecen imes, e u bishmen wo ks in ol ing oo eno a ion, hea ing, en ila ion and ai
condi ioning sys em emodelling, es i u ion o de e io a ed inishes and eplacemen o
Figu e 11. Manuel de Falla Audi o ium, in G anada (Spain): (a) gene al iew o he oom a e he ecen comple ion o
e u bishmen wo ks –on he le -, and (b) g ound plan showing he placemen o he sou ce (in blue) and h ee ecei e s
(in ed) conside ed in his s udy –on he igh -. A chi ec : J. M. Ga cía de Pa edes.
In e na ional Semina on Vi ual Acous ics VALENCIA, No embe 24-25, 2011
155
audience sea ing ha e been unde aken in he g and conce hall. A e hei comple ion, he
Manuel de Falla Audi o ium eopened o he gene al public in Sep embe , 2010.
3. EXPERIMENTAL MEASUREMENTS
Ex ensi e on-si e measu emen s we e pe o med in he audi o ium unde emp y- oom and
clea ed-s age condi ions, in a endance o wha es ablished in he ISO 3382-1 [7] in e na ional
s anda d. Tempe a u e and humidi y we e moni o ed du ing he measu emen campaign wi h a
digi al he mo-hyg ome e . Tempe a u e anged om 19.6 °C o 20 °C, whe eas ela i e
humidi y exhibi ed alues wi hin he 49% o 50% in e al.
Binau al oom impulse esponses (BRIRs) o each ecei e loca ion ha e been ob ained by
means o 30-second sine sweep signals co e ing he en i e audible equency ange, gene a ed
and subsequen ly analyzed in WinMLS 2004, h ough he Digig am VX Pocke 2 soundca d.
The audio signal so-p oduced was sen – ia an INTER-1000 ampli ie - o a 01-dB S ell AVM
DO12 dodecahed al loudspeake –omnidi ec ional sound sou ce- placed ons age a a cen al
posi ion, 1.5 me es high abo e he s age le el.
Head Acous ics head and o so simula o Type HSU III (Code 1323) has been used o
eco ding BRIRs a he h ee sea s o choice, alloca ing bo h en ances o he ea canals a a
1.2-me e heigh abo e he loo . Monau al oom impulse esponses (RIRs) ha e been
measu ed as well, allowing o he calcula ion o widely-known acous ic pa ame e s. Rega ding
he p e iously-desc ibed se -up, a couple o addi ional mic ophones –namely, B&K 4190 ½”
omnidi ec ional and Audio-Technica AT4050/CM5 mul i-pa e n condense mic ophones- wi h
sui able 01dB-S ell p e-ampli ie s ha e been u ilized ins ead o he head and o so simula o o
his pu pose.
4. COMPUTER SIMULATION
CATT TUCT 1.0 was he piece o oom
acous ics p edic ion so wa e chosen o
ca y ou he simula ions whose esul s a e
he eby p esen ed. TUCT –ac onym o The
Uni e sal Cone T ace - was ecen ly bo n
as a CATT-Acous ic-dependen ba e
calcula ion engine capable o pe o ming
acous ic simula ion and au aliza ion o
i ual en i onmen s. Howe e , CATT-
Acous ic keeps on being necessa y o build
up he h ee-dimension geome ical model
o he oom and la e de ine he acous ic
ea u es o enclosing su aces.
Fo he ime being, h ee dis inc algo i hms
ha e been implemen ed in TUCT o he
gene a ion o echog ams and impulse esponses o any gi en sou ce- ecei e combina ion
speci ied in he digi al model [8]. Di e ences among hemsel es mainly a ec he way di use
e lec ions a e handled. The h ee o hem deal wi h di ec sound and 1s o de specula
e lec ions de e minis ically. So do hey wi h 1s o de di use e lec ion by means o he so-
called andom spli -up, a b and-new echnique made g adually ex ensi e o highe o de s o
di use e lec ion in algo i hms 2 and 3. I is o ha eason ha un- o- un a iabili y adi ionally
associa ed o compu e -aided acous ic simula ion due o he usage o Lambe ’s law [9]
accoun ing o sca e ing phenomena has been signi ican ly educed. TUCT’s algo i hm 2 has
been he one o choice o a aining he da a u ilized in his wo k.
I mus be no iced ha TUCT p edic ions o he mos common acous ic pa ame e s a e based
on he p e ious calcula ion o ei he echog ams o impulse esponses, which gi es ise o
sligh ly di e en esul s coming om ene gy (E) o p essu e (h) analyses espec i ely. Since he
Figu e 12. Geome ical model o he acous ic simula ion o
he Manuel de Falla Audi o ium. Audience a ea is shaded in
g ey colou .
In e na ional Semina on Vi ual Acous ics VALENCIA, No embe 24-25, 2011
156
ou come o he la e ollows a p ocedu e ha is iden ical o ha ins i u ed by he in e na ional
s anda d ISO 3382-1 [7] o expe imen al measu emen s –successi e oc a e equency-band
il e ing, squa ing and p essu e- unc ion ime in eg a ion-, a good ag eemen o in si u measu ed
desc ip o s and hei IR-based simula ed coun e pa s is o be expec ed, p o iding as well clea
e idence o simila i y be ween eco ded and p edic ed impulse esponses. Addi ionally, BRIRs
a he speci ied loca ions can be gene a ed and expo ed in o sui able o ma s making i
possible hei la e usage o au aliza ion.
In o de o simula e he acous ical beha iou o he Manuel de Falla Audi o ium, a h ee-
dimension geome ic model o he oom –illus a ed in Figu e 2- has been implemen ed, in
acco dance o he guidelines supplied by he de elope s o p edic ion so wa e [10] and s a ed
in he li e a u e [11]. Acous ic cha ac e is ics o he ma e ials in he model we e i s es ima ed
and de ined wi h sca e ing and abso p ion coe icien s [11][12], which ha e been subsequen ly
ine- uned by means o an i e a i e calib a ion p ocess based on achie ing a ma ch in he
compa ison be ween measu ed and simula ed e e be a ion imes a a ious inne loca ions
[13]. Finally, he model esul ing om such a p ocedu e has been alida ed o he es o
con en ionally-accep ed oom acous ics pa ame e s in e ms o jus no iceable di e ences
(JNDs) as p oposed by Bo k [3][4].
Mo eo e , s a ing om his o me model, ano he one has been de eloped in an a emp o
mimic he changed acous ic ea u es wi hin he conce hall due o a ecen ly inished
e u bishmen . Bo h models ha e been acous ically simula ed, main desc ip o s calcula ed, and
BRIRs p oduced.
As o he ini ial p edic ion se ings adop ed in he compu e so wa e, 26,611,576 cones –in
applica ion o he o mula sugges ed by Vo lände [11]- ha e been aced. T unca ion ime o
he calcula ion o impulse esponses and echog ams was se a 3 seconds coun ed om he
a i al o di ec sound – hus, beyond expe imen ally-measu ed e e be a ion imes, as
ecommended by some au ho s [10][11].
5. AURALIZATION
Au aliza ion is he echnique ha allows o he c ea ion o sound iles sui able o ep oduc ion
om me e nume ical da a ob ained by simula ion, measu e o syn hesis [11]. Physically
speaking, such a ask equi es he de ini ion o he so-called binau al oom impulse esponse,
h( ), empo al unc ion based on a h ee-dimension sound p opaga ion model whose calcula ion
in ol es he compu a ional echnique p e iously explained. BRIRs can be also depic ed as
ac ual il e s, ( ), inco po a ing addi ional componen s, o wi : sou ce di ec i i y, head- ela ed
ans e unc ion (HRTF) and exac equaliza ion o he sound- ep oduc ion sys em o choice.
The con olu ion o inpu sound signals eco ded in anechoic en i onmen s, s( ), wi h il e
impulse esponses, ( ), o p oduce ou pu audio iles as hea d by he lis ene s a any gi en
loca ion, g( ), is he basis o signal p ocessing o au aliza ion, and can be ma hema ically
w i en as he ollowing linea - ime in a ian sys em:
!"#$ %&!'#(!")'#*'
+
,+
A li le s and-alone piece o so wa e called Mul iVol e WCP included in CATT-Acous ic
package has been used o ca ying ou he au aliza ions o be la e on p esen ed o he
assesso s. Mul iVol e is in ended as a lexible mul ichannel con ol e o mul iple applica ions
whose pu pose is o con ol e inpu WAV- iles in o ou pu WAV- iles [14]. A mono o binau al
con igu a ion –one o wo channels-, making use o p e iously measu ed and simula ed ini e
binau al impulse esponse il e s, has been aken up in his wo k.
Th ee sho , d y, solois classical-music samples di e ing om one ano he in hy hmical
pa e n, equency spec um and, abo e all, imb e ha e been chosen as inpu signal o
au aliza ion:
In e na ional Semina on Vi ual Acous ics VALENCIA, No embe 24-25, 2011
157
C. Webe : Theme; o cello [15].
F. Chopin: P elude, Op. 28, No. 12 in G-sha p mino ; o piano [16].
B uckne : Symphony No. 8 in C mino , I. Alleg o mode a o; o imbales [17].
I mus be no ed ha , in o de o ha e au aliza ions om impulse esponse il e s o di e se
na u e sui able o compa ison, he gain in dB applied o each o he h ee inpu audio iles in
ei he case was he esul o a p io calib a ion p ocess aimed o yield a ma ch in e ms o
a e age loudness moni o ed a he closes loca ion o he sound sou ce.
6. LISTENING TESTS
Twen y-se en people ha e been subjec ed o lis ening es s in o de o disc imina e whe he
he e a e senso y di e ences be ween audio s imuli. A iangle- es s uc u e [18] has been
deemed well i ed o de e mine bo h whe he :
A signi ican di e ence exis s be ween au aliza ions ob ained om measu ed and
simula ed BRIRs in he Manuel de Falla Audi o ium.
Such a di e ence can be s a ed be ween au aliza ions om simula ed BRIRs be o e
and a e he comple ion o some e u bishmen wo ks wi hin he conce hall.
To his pu pose, a panel o un ained assesso s consis ing o Spanish uni e si y s uden s wi h
no mal-hea ing be ween ages eigh een o wen y-eigh was ga he ed. The campaign o lis ening
es s was ca ied ou in a quie , windowless oom, ee om bias. In his en i onmen , acous ical
s imuli we e adminis e ed o he panellis s ia an ASUS Xona Essence ST in e nal sound ca d
and Sennheise HD 600 o e -ea headphones. A pic u e o one o he indi iduals unde going
he lis ening es s imme se in his se up is shown o be e illus a ion in Figu e 3(a).
The es s ha e been designed o igu e ou answe s o wo ques ions: Is he e a di e ence
be ween he pai o s imuli unde compa ison? And we e ha he case, wha is he na u e o his
di e ence?
To ind ou , assesso s we e p esen ed wi h h ee o hem and old ha wo a e iden ical and one
is di e en . Acco ding o a iangle- es design, six possible o de s o sample p esen a ion ha e
been conside ed: AAB, ABA, BAA, BBA, BAB, and ABB.
The use o naï e assesso s made i ad iso y o use a o ced-choice mode in he expe imen s.
Panellis s we e asked o ca e ully lis en o he samples in he o de o supply and single he odd
one ou . A e hea ing each se o h ee s imuli o he i s ime, assesso s we e no pe mi ed o
e-lis en o he samples. To do so, panellis s could use all a ailable in o ma ion wi hou any
es ain s.
In addi ion o his, hey ha e been ques ioned abou he na u e o he di e ence no iced. To
acili a e o a panel o un ained indi iduals o answe o such an inqui y, ou key acous ical
ea u es –colo a ion, spaciousness, e e be ance and loudness- we e sugges ed. Ha ing been
easily de ined o e e y assesso a he beginning o he expe imen al session, om none o all
o hem could be ma ked in he pop-up window displayed in Figu e 3(b), which u ned up a e
he p esen a ion o each se .
Figu e 13. On he le , one o he olun a y assesso s unde going he es s is being subjec ed o he expe imen s (a).
On he igh , he sho ques ionnai e -in Spanish- o be answe ed in a window en i onmen a e lis ening o e e y ain
o s imuli is shown (b).

In e na ional Semina on Vi ual Acous ics VALENCIA, No embe 24-25, 2011
158
Up o eigh een so-con o med se s o h ee s imuli ha e been adminis e ed in andom
a angemen o e e y assesso , ega ding h ee ecei e loca ions, h ee musical exce p s, and
wo dis inc expe imen s wi h he same s uc u e. P io o he commencemen o he ac ual
lis ening es s, a ial un wi h wo di e en samples was p esen ed wice o he panellis s, as o
p o ide a b ie explana ion on he so o a ia ions ha migh be encoun e ed be ween s imuli
h oughou he expe imen s.
7. RESULTS AND DISCUSSION
Acco ding o plan, he campaign o lis ening es s aimed o unea h signi ican di e ences
be ween he au alized samples in he cases s udied has been conduc ed. The o al numbe o
esponses co ec ly iden i ying he une en s imulus has been coun ed, and he collec ed da a
ega ding each expe imen has been summa ized in he e ical ba cha s shown in Figu e 4.
Fo a panel con o med o wen y-se en assesso s and a a disc e iona y signi icance le el o
5%, he minimum numbe o co ec answe s equi ed be o e concluding a signi ican di e ence
om ei he expe imen is 14 –ma ked by he ho izon al line plo ed wi h dashed ace in he
diag ams-. A ha le el o signi icance, he amoun o igh esponses mus hence exceed his
c i ical alue o a posi i e s a emen .
Fo he sake o cla i y, he analysis o he esul s o he wo expe imen s p esen ed in his wo k
will be epo ed sepa a ely.
7.1. Measu ed-BRIRs Au aliza ion e sus Simula ed-BRIRs Au aliza ion.
In iew o he da a coming om he iangle es s o he i s expe imen , i can be s a ed ha a
signi ican di e ence exis s be ween au aliza ions om ield-measu ed BRIRs and hei
simula ed equi alen s in he conce hall unde in es iga ion. Only one ou o nine possible
combina ions o ecei e placemen and musical exce p does no yield such a conclusion, o
wi , piano sample a he sea loca ed behind he s age.
Ne e heless, i is also wo h poin ing ou ha he di e ence be ween he pai s o audio signals
unde compa ison u ned ou o be less no iceable in au aliza ions o he piano anechoic
eco ding han in he es o assessed cases. By con as , using a pe cussion sample –i.e.
imbales- seems o ha e made i he easies o he assesso s o ell he di e ence be ween
au alized iles om measu ed IRs and hei simula ed coun e pa s.
Looking a acous ic spec og ams o he o iginal e e be a ion- ee musical exce p s o be
au alized illus a ed in Figu e 5, i is o no wonde he s iking dependence ound be ween he
deg ee o di icul y encoun e ed by he assesso s when making hei judgmen s and he na u e
o he musical sample ha was being adminis e ed in his i s expe imen . F equency spec a
Figu e 14. Numbe o co ec iden i ica ions o he odd sample collec ed om he wo expe imen s conduc ed, classi ied
by ecei e loca ion and na u e o he inpu signal. On he le : measu ed s. simula ed-IR au aliza ion (a). On he igh :
be o e- s. a e - e u bish au aliza ion (b).
Recei e
A-1 A-2 B
Numbe o co ec esponses
0
5
10
15
20
25
30
Cello
Piano
Timbals
Recei e
A-1 A-2 B
Numbe o co ec esponses
0
5
10
15
20
25
30
Cello
Piano
Timbals
In e na ional Semina on Vi ual Acous ics VALENCIA, No embe 24-25, 2011
159
(FFT) e sus ime in e ms o A-weighed sound
p essu e le el show an i egula ene ge ic
dis ibu ion in he equency ange o analysis o
he anechoic inpu signals in use. Whe eas he
piece o piano music goes wi h undamen al
equencies alling wi hin he 500-he z-cen ed
oc a e band, cello and imbales e eal g ea e
amoun s o ene gy wi hin he 250 and 125-he z
oc a e bands co espondingly.
On he o he hand, i is well known ha some
acous ic pa ame e s a e closely ela ed o ce ain
subjec i e ea u es. Likewise, sound s eng h (G)
is a good indica o o he subjec i e le el o sound.
So applies o pe cei ed e e be ance and ea ly
decay ime (EDT), pe cei ed cla i y and cla i y
i sel (C80), and spa ial imp ession and in e au al
c oss-co ela ion coe icien s (IACC). As a
consequence, salien misma ches in he
compa ison be ween measu ed and simula ed
igu es o hese desc ip o s mus ha e been
ansla ed in o no iceable di e ences be ween
hei subsequen au aliza ions.
In conside a ion o he exempla y da a plo ed in
Figu e 6, co esponding o he alues o he
abo e-men ioned acous ic pa ame e s a he
ecei e labelled as A-1, he mos ou s anding
de iances be ween measu ed and simula ed
ciphe s concen a e in he 125 and 250-Hz-
cen ed oc a e equency bands, mainly a ec ing
s eng h (G) and cla i y (C80) espec i ely –
coun ing up o 10 JND and 5 JND in ei he case-.
Tha is he eason why di e ences happen o be
ai ly mo e pe cep ible when imbales and cello
exce p s, in ha o de , a e adminis e ed o he
assesso s wi hin he expe imen al sessions.
Asked abou he na u e o he di e ence, an
o e whelming majo i y o he panellis s was
inclined owa ds colo a ion. As illus a ed by he pola plo in Figu e 7, he choices o he
indi iduals unde going he lis ening es s a ou ed colo a ion and e e be ance o e loudness
and spaciousness. Going back o he desc ip o s whose ciphe s ha e been spec ally depic ed
in Figu e 6, he assesso s’ esponses o his ques ion co ela e ai ly well o he indings in he
g aphs, ega ding signi ican de ia ions be ween measu ed and simula ed nume ical da a in
s eng h (G), ea ly decay ime (EDT), and cla i y (C80).
No no ewo hy dependence upon he loca ion o he ecei e has been ound in he expe imen .
Figu e 15. F equency spec a (FFT) e sus ime o
A-weighed sound p essu e le el (SPL) in dBA o he
o iginal agmen s o anechoic eco dings au alized
o he expe imen s. F om op o bo om: cello, piano
and imbales sample spec og ams.
In e na ional Semina on Vi ual Acous ics VALENCIA, No embe 24-25, 2011
160
Figu e 16. Measu ed -in black- and simula ed -in blue and ed, ega ding be o e and a e e u bish- da a o some oom
acous ics pa ame e s a ecei e A-1. F om op o bo om, and le o igh : sound s eng h (G), ea ly decay ime (EDT),
cla i y (C80), and in e au al c oss-co ela ion coe icien (IACC). Fo he sake o compa ison, simula ed alues om
ene ge ic analysis (E) –in con as o hose om IRs (h)- a e depic ed wi h dashed ace.
7.2. Simula ed-BRIRs Au aliza ion: Be o e e sus A e Re u bishmen .
As o he second o he expe imen s conduc ed, whose esul s a e b oken down in Figu e 4(b),
a signi ican di e ence be ween au aliza ions om impulse esponses espec i ely e e ing o
he condi ions be o e and a e he e u bishmen o he audi o ium canno be concluded. Jus in
he scena io o he cello anechoic eco ding being con olu ed wi h he binau al oom impulse
esponse ob ained a he backs age sea included in his s udy, a s a emen o signi ican
di e ence would be s a is ically jus i ied.
Su p isingly enough, p ominen de iances in e ms o spa ially-a e aged e e be a ion ime
(T30) be ween he s a es in compa ison a e mani es in he da a plo ed in Figu e 8. Mo eo e ,
such misma ches ge o a peak alue beyond 5 JND alling a he 250-he z-cen ed oc a e
band, which should ha e yielded an easie ecogni ion, abo e all, wi h cello music. None heless,
i mus also be said ha none o he es o acous ic pa ame e s he ein analyzed pu on display
dispa i ies su passing 2 JND –wi h he excep ion o ea ly decay ime (EDT), connec ed o he
subjec i e disce nmen o e e be ance as well- as shown in Figu e 6. On he whole, howe e ,
he numbe o co ec esponses does no seem o ha e been linked o ei he he ecei e ’s
loca ion o he na u e o he s imulus adminis e ed in his case.
To he ques ion on how he odd sample was di e en , ewe answe s we e collec ed om his
expe imen han he p eceding one. Among he assesso s who did p o ide a esponse,
e e be ance and loudness we e he mos popula . Bu , gi en he ac ha a low pe cen age o
he panellis s we e uly able o ell he di e ence be ween he s imuli being p esen ed o
[Hz]
125 250 500 1000 2000 4000
G [dB]
5
10
15
20
25
Measu ed
Simula ed h
Simula ed E
Re u bished h
Re u bished E
[Hz]
125 250 500 1000 2000 4000
EDT [s]
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Measu ed
Simula ed h
Simula ed E
Re u bished h
Re u bished E
[Hz]
125 250 500 1000 2000 4000
C
80
[dB]
-10
-8
-6
-4
-2
0
2
4
6
Measu ed
Simula ed h
Simula ed E
Re u bished h
Re u bished E
[Hz]
125 250 500 1000 2000 4000
IACC
0.0
0.2
0.4
0.6
0.8
1.0
Measu ed
Simula ed h
Re u bished h
In e na ional Semina on Vi ual Acous ics VALENCIA, No embe 24-25, 2011
161
compa ison, hese answe s mus ha e been he esul o a guesswo k, and a e, he e o e,
inapp op ia e o u he analysis.
8. CONCLUSIONS AND FUTURE WORK
I has been p o en ha he e is s ill a
signi ican di e ence be ween au aliza ions
om expe imen ally eco ded binau al oom
impulse esponses and hei simula ed
equi alen s ob ained by means o a
nume ically- alida ed acous ic model o he
enclosu e. Consequen ly, sligh ly di e en
app oaches should be a emp ed in he
nea u u e in o de o each oom acous ics
digi al models speci ically sui ed o easible
au aliza ion. To do so, a s a egy based on
calib a ing he geome ic model by using
igu es o he bes -known desc ip o s
coming ou o pos -p ocessed impulse
esponses – ollowing s anda d p ocedu es
analogous o hose conside ed when
handling in-si u egis e s-, ins ead o
eso ing o da a om ene ge ic echog ams
migh be ad isable.
In e es ing conclusions associa ed o he
use o s imuli o a ied na u e can be
eached om he lis ening es s conduc ed.
The highly equency-dependen beha iou
usually demons a ed by mos o he
con en ional oom acous ics pa ame e s
makes i c ucial o ca y ou speci ic
spec al analysis o hose anechoic inpu
signals o be u ilized in au aliza ion p io o
op ing o one o con enience. I has been
demons a ed ha such a choice does play
a undamen al ole in disclosing
dissimila i ies in cen al acous ic ea u es
be ween au aliza ions e en ually subjec ed
o compa ison.
Fu he mo e, conside able di e gences
ound in desc ip o s co ela ing wi h ce ain
subjec i e aspec s o sound ha e been
e ealed o be capable o co ec ly
o eseeing la e no iceable di e ences
om e en ual panellis s’ assessmen s in
his so o senso y expe imen s. Tha is he
case o ea ly decay ime (EDT) and cla i y
(C80) -con e sely connec ed o e e be ance- and sound s eng h (G) -in close ela ion o
pe cei ed sound le el-. The e o e, analyzing nume ical da a o some acous ic pa ame e s
could help wa n o juiceless and unnecessa y campaigns o lis ening es s be o e being
unde aken.
Howe e , such a ansla ion om nume ical in o pe cep ual di e ences u ns ou no o be so
s aigh when i comes o e e be a ion ime (T30) being he only desc ip o showing no ewo hy
a ia ions. Gi en he ac ha he sea s emo ed om he conce hall we e s ill in seemingly
sound condi ion a he ime o eplacemen , he ou come om he second expe imen would
ha e made i acous ically poin less. Ne e heless, no only is Acous ics in he a chi ec ’s mind
Numbe o esponses
0
25
50
75
100
125
150
Colo a ion
Spaciousness
Re e be ance
Loudness
Measu ed s. Simula ed
Be o e s. A e Re u bish
Figu e 17. Combined pola diag am o panellis s’
p e e ences a p opos o he na u e o he di e ence be ween
s imuli compa ed in he expe imen s. In blue: measu ed s.
simula ed-IR au aliza ion. In ed: be o e- s. a e - e u bish
au aliza ion.
[Hz]
125 250 500 1000 2000 4000
T
30
[s]
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Measu ed
Simula ed h
Simula ed E
Re u bished h
Re u bished E
Figu e 18. Spa ially-a e aged e e be a ion ime (T30)
assessed a he h ee ecei e s in his wo k. Measu ed –in
black- and simula ed eco ds –dis inc i ely using blue and
ed colou s o be o e- and a e - e u bish s a es- a e shown.
Ene gy (E) and p essu e analyses (h) ha e been also
di e en ia ed by dashed plo o he la e .