Professional Documents
Culture Documents
Articulo Dano
Articulo Dano
Applied Acoustics
journal homepage: www.elsevier.com/locate/apacoust
a r t i c l e i n f o a b s t r a c t
Article history: The sound insulation provided by a certain construction element is usually presented through Single-
Received 11 February 2020 Number Quantities (SNQs). These SNQs are global values used to summarize the sound insulation that
Received in revised form 8 April 2020 a construction element has as function of the frequency. Although there are standardized SNQs, whose
Accepted 13 May 2020
calculation is described in the ISO 717–1 and ASTM E413-16 standards, there is not enough agreement
Available online 10 June 2020
on whether these SNQs are representative of the way humans perceive.
The purpose of this research is to assess the degree of correlation between the annoyance perceived by
Keywords:
humans and the most relevant existing SNQs for a set of façade insulation elements and urban-type
Subjective perception
Annoyance
sound samples. Likewise, the purpose is also to evaluate whether certain frequency bands have more
Sound insulation influence than others on the perception of annoyance. Furthermore, to propose a reference curve for
Single number quantities the calculation of a SNQ that is representative of the annoyance perceived by humans when faced with
sounds of an urban nature.
To this end, a 2-alternative choice (2-AC) listening test was carried out by 119 participants. In this test,
participants had to listen to stimuli in pairs and select the stimulus that was most annoying to them. The
stimuli were obtained by filtering different sound samples with the sound insulation of different con-
struction elements. In particular, five sound samples were used for this listening test: four urban-type
sound samples as well as pink noise. Also, the sound insulation of six façade insulation elements was
used.
The results of this listening test were then analyzed using Thurstonian models and the distance mea-
sure d’ (d-prime) was obtained for each comparison. Finally, an optimized reference curve for urban-type
sounds Lopturban was calculated based on the results of the listening test by means of a Sequential
Quadratic Programming (SQP) optimization algorithm.
Ó 2020 Elsevier Ltd. All rights reserved.
https://doi.org/10.1016/j.apacoust.2020.107432
0003-682X/Ó 2020 Elsevier Ltd. All rights reserved.
2 D. de la Prida et al. / Applied Acoustics 168 (2020) 107432
behavior of an element depending on the spectral features of the extended to 50 Hz and 5000 Hz respectively for the lower and
incident noise. Little research has been done, however, to assess the higher bounds. In recent years, the scientific dialogue has
whether the focus should be on modifying the reference curve or mainly focused on whether it is necessary to add frequency bands
the spectra used for the calculation of the spectral adaptation below 100 Hz or not. As described in [22], while for impact noise
terms and if so, what changes would be necessary to improve the the general trend tends to propose the inclusion of the bands in
relationship between the subjective perception and the Single- the range between 50 and100 Hz, for airborne sound insulation
Number Quantities. there is no generalized trend. While several studies show that
The purpose of this research is to evaluate the annoyance per- the inclusion of the low frequency bands does not improve or even
ceived by a significant sample of participants when presented with worsens the description, others reinforce their adequacy specially
urban sound samples filtered by common sound insulation ele- for music sounds.
ments used in the construction of façades. Further, to determine On the contrary, little research has been carried out to deter-
the relevance of each spectral band in the perception of annoyance mine whether the calculation method and the reference curve to
and to obtain a more representative SNQ than those existing. which the construction elements are compared should be modi-
For this purpose, a two-alternative choice (2-AC) listening test fied. In this sense, the early works of van den Eijk [23] and North-
was conducted [8], in which a sample of 119 participants reported wood [24] are especially relevant, trying to propose more
their perceived annoyance when listening to five different types of representative curves of loudness and annoyance. Also, the more
sound virtually insulated by the Sound Reduction Index (SRI) of six recent work of Scholl is worth mentioning [25], aimed at proposing
different façade insulation elements. Then, the results of this lis- further reference curves for inclusion in the at that time work-in-
tening test were analyzed by means of Thurstonian models and progress ISO 16717–1. Finally, the most recent Virjonen’s work
the correlation between the subjective judgments and several [26], which proposed a reference curve for airborne sound insula-
existing SNQs was obtained. Further, the results of the Thurstonian tion taking into account living sounds by means of an automatic
analysis were used as input in a SQP (Sequential Quadratic Pro- optimization method, is also noteworthy.
gramming) optimization algorithm, and five different optimized These proposals have, however, not been reinforced or refuted
reference curves were obtained, one for each sound type. Finally, by other researches with exception of [25], whose proposals have
an average optimized reference curve Lopturban was proposed, being been studied in [20,27,28].
highly representative of the subjective perception of annoyance for A SNQ that is highly representative of subjective perception
the chosen urban sounds . hasn’t been defined yet, which means that there is currently no
consensus on which SNQ should be used to describe the sound
insulation of an element. Consequently, as developed in [29,30],
2. Historical perspective different SNQs are used in Europe to assess the sound insulation
depending on the country. It is therefore of great interest to per-
The first attempts to achieve a Single-Number Quantity that form further experiments with additional types of sound to prove,
would representatively describe the sound insulation were based improve or complement the already proposed reference curves
on one or several arithmetic averages (AA) of the sound insulation enhancements with the aim of obtaining a unique SNQ that is
in the range of frequencies between 100 and 3000 Hz [9]. However, highly representative of the perception of airborne sound
it soon became apparent that this procedure could give the same insulation.
SNQ values to several insulation elements with very different spec-
tral features. Surprisingly, many years later, some studies [3,10,11]
3. Material and methods
have pointed out that the AA index, while it cannot be considered
better than the standardized SNQs, does not perform as poorly as
In this section, brief information is first given about the design
initially thought.
and performance of the listening test. Further details and deeper
The next approach, which is the one used so far, was to establish
explanations about these facts can be found in [8], in which the
a comparative method that would allow to obtain a rating of the
methodology used for this listening test is explained at length.
insulation of a certain element by the comparison with a reference,
However, information is given here about the sound samples and
whose acoustic behavior was considered good. A reference curve
insulation elements, as well as about the sample of participants.
was proposed, which was a smoothed version of the sound reduc-
Detailed information is then given about the analysis by means
tion index (SRI) of a 9-inch brick wall, which was typically used for
of Thurstonian models.
the separation between dwelling units in the mid-20th century
Finally, the used optimization procedure, which is an adapta-
[1,2] in Europe but also in the United States.
tion of that proposed by Virjonen in [26], is explained in detail.
Surveys were then carried out to assess the appropriateness of
this reference curve in different countries such as England, Swe-
den, Holland and France [12–15]. While some of these studies rein- 3.1. Listening test
forced the idea of the use of this type of traditional brick wall as a
reference curve, as presented in [1], others highlighted that the use 3.1.1. Study design and performance
of this reference curve could be insufficient to describe the percep- A listening test was conducted in an insulated booth, located at
tion of sound protection perceived by the tenants. the Technical School of Architecture of the Technical University of
However, Single-Number Quantities based on this comparative Madrid (Spain), by using Sennheiser HD650 headphones. The fre-
method and reference curve have been widely used to describe the quency response of the headphones was compensated and the
sound insulation of construction elements over the years. playback chain calibrated to assure a correct playback of the used
Many studies have then been carried out to determine the ade- sound samples. Further details about the environmental conditions
quacy of existing SNQs to the subjective perception of the popula- and the equipment can be found in [8].
tion [3–5,11,16–21]. A deeper review of some of these studies can Following a 2-AC procedure, 119 participants were presented
be found in [22]. with different comparisons, each comprising two different stimuli
These studies have mainly focused on determining which is the and were asked to select the sample that, in their opinion, was the
most appropriate SNQ and also on whether the calculation range, most annoying. They were also able to answer ”I don’t know/ I
which is historically set between 100 and 3150 Hz, should be don’t perceive difference” if they were not able to make a clear
D. de la Prida et al. / Applied Acoustics 168 (2020) 107432 3
statement. These comparisons were performed in random order to to fulfill a survey with the gender, age, nationality and level of
assure statistical independence. studies. Also, the test was innocuous, given the low sound levels
The stimuli compared by the participants were five different emitted. However, the participants were specified that they could
sound samples filtered by the SRI of six different façade insulation finish the experiment at any time in case they felt uncomfortable
elements, making a total number of 30 stimuli, each stimuli being a or annoyed. Further information about these facts will be given
combination of one of those five sound samples and one of the six in the Listening test design section.
insulation elements. Considering a pairwise comparison strategy,
in which all possible combinations of stimuli are conducted, each
participant performed 15 comparisons (see Table 3) for each sound 3.2. Sound samples
sample and the process was repeated for the five different sound
samples. Therefore, each participant performed a total number of Five different sound samples were used in the experiment.
75 comparisons. Given that the aim of the experiment was to address the correla-
In order to reduce the sources of bias that could affect the qual- tion between existing SNQs and the sound insulation for urban sit-
ity of the results, as well as to reduce their possible impact, the lis- uations and to propose a better reference curve for these cases, the
tening test was carried out following a predefined procedure. This selected sound samples were representative excerpts of different
procedure helped to reduce the influence of various sources of bias, urban noise situations recorded for this purpose in the city of
as described in [8]. Madrid. In particular five different sound samples were used.
Two of the samples were binaural recordings of traffic noise at
two different road traffic conditions, under different proportions
3.1.2. Participants of vehicle types and road speeds. Another sample was a binaural
The participants were recruited mainly at the Superior Techni- recording of an aircraft flyover. The fourth sample was a binaural
cal School of Architecture of the Technical University of Madrid by recording of a common situation in a pedestrian commercial street,
email, phone call or visit. The potential participants were briefly during the passage of a small emergency vehicle sounding its siren.
explained the procedure and the estimated duration of the exper- Details about the recording, analysis and the criteria for the selec-
iment. They were also explained that, although no compensation tion of these recordings can be found in [31,8].
would be given for their participation, a hearing level test would Additionally, pink noise was used as a fifth sample during the
also be performed and that they could get it printed and explained test. This sample was chosen for reference based on its flat spec-
for their information. More than 200 people, academic and non- trum in 1/3 octave bands and its stable and constant temporal
academic, were given with information about the test. features.
Finally, the listening test was performed by 119 people, from Fig. 1 shows the 1/3 octave spectrum of the five sound samples
which the 54.6% were women (N = 65) and the 45.4% were men selected. As it can be seen, the spectral behavior of each sound type
(N = 54). The age of the participants was gathered in ranges of five is very different, although the first four are representative samples
years between the ranges ‘‘15–19 years” and ‘‘>=65 years”. The of usual sounds in urban areas. This fact can help to assess whether
exact age of the extreme cases is known. Therefore, it can be pre- the proposal of a unique reference curve might be done or not.
cisely stated that the participants were aged 19 to 71 years old. Also, the analysis of the results for the Siren sound sample can lead
Table 1 shows the different age ranges, the proportion of the to interesting conclusion given the fact that its main energy contri-
total amount of participants belonging to each range and the pro- bution is centered in the 1/3 octave frequency band of 1250 Hz.
portion of women and men inside each range. This frequency band is considered (among other adjacent bands)
As it can be seen in Table 1, the age group ”20–24 years” was the to be particularly relevant in the reference curves currently in
biggest one over all the others, although its participants repre-
sented only the 28.6% (N = 34) of the total. However, to ensure that
this higher proportion of participants wouldn’t have a significant
effect on the bias, a Student’s t-test was conducted for both the
age group ”20–24 years” and a group containing all other partici-
pants. The outcome of this test is presented in the Results section.
The test was performed by people from 13 different countries.
However, most of the participants were from Spain, bringing
together more than the 83% (N = 99) of the sample.
The test was anonymous, in order to safeguard the ethical integ-
rity of the experiment. Therefore, no sensitive information was
gathered from the participants. The participants were only asked
Table 1
Ranges of age and proportion of participants, men and women, belonging to each
range.
use. Therefore, high correlations are to be expected between the First, the participant was welcome on a specific date and time
subjective judgments for this sound sample and the existing SNQs. that had been previously appointed. The participant was then
invited to enter the insulated booth and asked to answer an anony-
3.3. Façade insulation elements mous demographic survey about age, gender, nationality and level
of studies. This survey was carried out for two reasons. First, know-
Six SRIs (W1 to W6) were selected for comparison in the listen- ing factors such as the gender and age of the participants can be
ing test. As described in detail in [8], the window, that is, the ele- helpful on addressing whether these factors can evoke subtle dif-
ment that is typically fixed in the hollow part of the façade, is ferences in the perception. Second, performing this easy and non-
usually the weakest element of it. Therefore, the six SRIs were cho- acoustic task prior to the listening test might help the participants
sen to belong to three different window types with a rolling shut- to get used to the test environment, before beginning the listening
ter fixed, both with the rolling shutter extended and retracted. test.
Therefore, each façade insulation element comprised a window Once the survey was completed, a document was provided,
from a specific type regarding its glazing and frame, the box of where the procedure of the listening test as well as safety and eth-
the shutter, and the shutter itself either extended or retracted. ical information were given. As the participant was finished with
The study of the shutter position on perception, however, is not the reading, the experimenter asked if everything was clearly
addressed in this article. understandable and assisted the participant aseptically if questions
The selection of this six elements was based on three principles: arose.
Next, the participant was presented, on the computer screen,
1. The selected elements had to be representative of façade insu- with the answer form, in which the judgments had to be given,
lation elements commonly used in construction. and a short training was carried out comprising two dummy pair-
2. The selected elements had to be in a sufficiently small range to wise comparisons. If the familiarization process ended without
allow the perceptual assessment of subtle differences in terms incidents, the listening test could begin.
of the existing SNQs. The listening test was performed as five individual tests, one for
3. The spectra of the selected elements had to be significantly dif- each type of sound, with a short break between each individual
ferent even when their SNQs were very similar. test. During the break, in a relaxed manner and avoiding it to look
as part of the evaluation, the participant was interviewed about the
Taking these conditions into account can help to draw very rele- test that had just been done. Sometimes interesting information
vant conclusions about the perception of the different elements, arose from these interviews and was, therefore, noted for future
given that the effect of the spectral difference on perception can consultation.
be evaluated, as well as the extent to which subtle differences in Once the five tests were completed, all participants were given
the existing SNQs are perceived by the participants. a screening audiometry to assess their hearing capabilities. Each
From a database of laboratory measurements, carried out by the participant received the result of the audiometry as well as an
Laboratory for Acoustics and Vibration of the Technical University explanation about the quality of hearing.
of Madrid over several years, a SRI was obtained for each window
type and shutter position, by averaging all existing measurements 3.5. Analysis of the results
for windows of the same category and same shutter position. This
procedure allowed obtaining highly representative SRIs for each of Different analysis methods are described in detail in the previ-
the six configurations. Further information about these measure- ous article, regarding methodology [8], pointing out the fact that
ments and the effect of the rolling shutter on the SRI of these an analysis by means of Thurstonian models allows obtaining more
façade elements can be found in [32,33]. precise, specific and lower influenced by bias information from the
The frequency range was fixed in the range between 100 and listening test. In this paper, Thurstonian models are used as the
5000 Hz since previous researches have already pointed out the main method for the analysis of the results.
low influence of the frequency bands below 100 Hz on the subjec- Furthermore, for the enhancement of the reference curve, an
tive perception of airborne sound insulation [16,17,19]. optimization method was used, based on the one proposed in
The final SRIs were in the range between 32 and 39 dB (Rw ), [26]. The procedure was, however, adapted to allow the use of
therefore having a DRw ¼ 7. The main features of each façade insu- the Thurstonian measure of distance as the subjective response
lation element are shown in Table 2. variable and considering the objective difference in terms of SRI
The 1/3 octave SRI for each of the six façade insulation elements as the input data used for the improvement of the reference curve.
is shown in Fig. 2.
3.5.1. Thurstonian models
3.4. Test procedure Thurstonian models consider that the perception of intensity of
a stimulus can be described as a normal probability distribution,
The listening test was carried out on a one person per date given that in some cases the same stimulus will be perceived to
basis. Therefore, only one participant could perform the listening be more intense than in others, depending mostly on physiological
test at the same time. variables [34]. Therefore, the difference between two confusable
Table 2
Nomenclature, physical features (frame material, opening type, drum filling and shutter position) and Rw for the six different elements selected.
Table 3
Elements judged by the participants in each comparison and nomenclature.
X
n
S¼ ðyi ðA þ Bxi ÞÞ2 ð2Þ
i¼1
(3), 15xi values were obtained, each of them for a possible combina- 4.2. Thurstonian Analysis
tion i, in which K frequency bands were considered for the calcula-
tion. Since the frequency range was fixed between 100 and 5000 Hz, The results of the listening test are analyzed by means of
K = 18. Thurstonian models. This analysis allows addressing, for each of
Other than that, the procedure used to obtain the optimal refer- the comparisons performed by the participants, the difference per-
ence curve has been programmed in MatlabÒ, using the Sequential ceived between the two sound insulation elements compared. This
Quadratic Programming (SQP) algorithm and the same boundary difference is shown by means of the Thurstonian measure of dis-
conditions as those described in the appendix of [26]. Also, the ref- tance d’.
erence spectrum for the C adaptation term in the range between For the sake of simplicity, Table 3 shows the sound insulation
100 and 5000 Hz was used as initial L guess in the algorithm fol- elements judged in each comparison, as well as a nomenclature
lowing the rationale of [26]. The calculation of the variables A that will be used for each comparison in the rest of this section.
and B in Eq. (2) also followed the formulations described in [26] Table 4 shows the values for the Thurstonian measure of differ-
and are not described here in detail for the sake of extension, as ence d’, calculated from the judgments of the listening test. Marked
these variables also depend on secondary variables depending on with asterisks, the range of the p-value for each comparison is
xi , yi and the relation between them. given: one asterisk (*) is used for 0.05 < p < 0.1; two asterisks
This optimization procedure is applied independently to the (**) are used for 0.0001 < p < 0.05; three asterisks (***) are used
results obtained for each sound sample. Therefore, five optimized for p-value < 0.0001. Also, no asterisks are used for p > 0.05. There-
reference curves are obtained, each of them being based on the fore, the more asterisks are given, the higher the difference per-
judgments given by the participants for each of the sound samples. ceived by the participants during the listening tests. In particular,
in those cases in which no asterisks are used it can be stated that
the two elements in a certain comparison were not significantly
4. Results
different in terms of subjective perception of annoyance for a
95% significance level.
This section presents the results of the different analyses that
As it can be seen on Table 4, representative perceptual differ-
have been carried out to obtain an optimized reference curve.
ence could be found by the participants for most comparisons.
First, through a Student’s t-test analysis, the effect that the lar-
Additionally, due to the high number of participants, the judg-
gest proportion of participants in the ”20–24 years” age range
ments obtained for most comparisons have a high (**) or very high
could have on the bias is assessed. Then, the results of the analysis
significance (***), which allows drawing very relevant conclusions.
of the participants’ judgments and their correlation with the exist-
Also, from Table 4 several conclusions can be drawn regarding
ing SNQs are presented. Additionally, the results of optimizing the
the variability of the perceived difference between the insulation
reference curve of the sound insulation based on the listening test
elements for different sound types. While for some comparisons
judgments are shown. Finally, the obtained optimized curve is
(C3, C6, C7, C8, C13) the difference is clearly perceptible with a very
compared with those defined in the standards and to that obtained
high significance (***), for others (C9, C12) the differences are
by [26] for living noises.
mostly not perceived. In addition, for some comparisons such as
C2, C4 and C15 the judgments vary widely depending on the spec-
4.1. Student’s t-test tral features of each sound type, although this behavior is not very
common.
Since the age group ”20–24 years” had a higher proportion of While for each particular sound sample different values of d’ are
participants than the others, it was assessed whether the results obtained, which can vary significantly from one to another, the
of d’ for that age group were significantly different from those of tendency of the participants is to select the same element as the
the rest of the participants. The aim was to determine whether this most annoying in a generalized manner. This behavior highlights
higher proportion could bias the calculation based on the pooled the fact that the sound sample may have some influence on the
data. For this purpose, d’ values were calculated for each compar- degree of perceived difference but is not that relevant as to create,
ison (see Table 3) and each sound sample for two groups of partic- in a generalized way, dramatic changes in judgment to the partic-
ipants: the first one comprising the participants of the age group ipants. However, the particular spectral features of the Siren sound
”20–24 years” and the second one comprising the rest of the par- sample have a relevant impact on the perception of difference,
ticipants. Then, a Student’s t-test analysis was performed on the making it easier for the participants to address the differences
d’ values for each of these combinations to address the difference for this particular sound sample.
between the two age groups. Table 4 can also be used to visually inspect the effect that the
This analysis reported that in only three of the 75 combinations differences, in terms of SRIs, have on the subjective perception of
there was a significant difference between the two age groups. annoyance and its relation with the existing SNQs. As an example,
These three combinations were: Figs. 3 and 4, which show the overlap of the SRIs for the compar-
isons C6 and C11 and C2 and C4 respectively, are presented below
Comparison between insulation element W5 and W6 for the and inspected in conjunction with Table 4.
Siren sound sample (p = 0.04, t = 2.109). Fig. 3 shows the overlapped SRIs of the elements addressed by
Comparison between insulation element W4 and W5 for the the participants in two different comparisons (C6, C11) in which
Pink sound sample (p = 0.02, t = 2.384). the differences for the Rw are of 1 dB, as it can be drawn from
Comparison between insulation element W4 and W6 for the Table 2. In the left of Fig. 3 the spectral differences between W2
Pink sound sample (p = 0.04, t = 2.101). and W3 (C6) are shown. Furthermore, at the right of Fig. 3 the spec-
tral differences between W3 and W5 (C11) are shown. As it can be
An overall Student’s t-test was also performed combining the data seen at a glance, both figures have a similar behavior, in which one
from all sound samples and all comparisons for the two age groups. of the elements provides higher sound insulation at lower frequen-
The results showed that the two groups were equal regarding the cies, while the other provides better sound insulation at higher fre-
Thurstonian measure of distance d’ (p = 0.81, t = 0.24). Therefore, quencies. In these two particular cases, those differences are
the significant differences obtained for the three particular cases somehow compensated, making both elements in each comparison
did not contribute to an overall bias to the test. to be addressed as similar regarding their SNQs. However, if a look
D. de la Prida et al. / Applied Acoustics 168 (2020) 107432 7
Table 4
Values of the Thurstonian measure of difference, d’ (d-prime) for each comparison and sound sample. (*) 0.05 < p<0.1; (**) 0.0001 < p<0.05; (***) p < 0.0001; () p > 0.05.
Fig. 3. Example cases in which the Rw is similar but the elements are clearly distinguishable by the participants. Upper figures represent the SRI for each element in the
comparison (Left: W2 and W3; Right: W3 and W5). Lower figures represent the difference between the SRI values between the two elements (DSRI).
is taken at Table 4, it can be seen that perceptual differences were range of middle and high frequencies than the other. This higher
found to be significant for all the sound samples in the first case sound insulation from the middle to high frequencies is consistent
and for four out of five sound samples in the second. with the expected preference in terms of Rw but it does not follow
Also, Fig. 4 shows the overlapped SRIs for two comparisons (C2 the subjective judgments given by the participants. However, in
and C4) in which the elements have differences higher than 2 dB the range of middle to low frequencies, an inversion is shown
for the Rw , but are however not perceived as significantly different regarding the insulation features of the compared elements, having
for most sound types. In the left of Fig. 4 the spectral differences for higher insulation the opposite element. Although in most cases this
the comparison C2 are presented. In addition, in the right of Fig. 4 change in behavior is small if compared to the large differences
the spectral differences for the comparison C4 are presented. found in the range of middle to high frequencies, it seems to have
As a summary, in all the presented cases, one of the elements in a significant relevance when determining subjective annoyance.
each comparison provides much higher sound insulation in the
8 D. de la Prida et al. / Applied Acoustics 168 (2020) 107432
Fig. 4. Example cases in which the Rw is different but the elements are not clearly distinguishable by the participants. Upper figures represent the SRI for each element in the
comparison (Left: W1 and W3; Right: W1 and W5). Lower figures represent the difference between the SRI values between the two elements (DSRI).
Having all of this into account it seems that the existing refer- nificance follows the above-mentioned criteria presented in
ence curves, which are giving more relevance to some frequency Table 4.
bands than others, may not have a good agreement with the per- As it can be drawn from Table 5, the correlation between the
ception of annoyance. It is, therefore, necessary to address the existing SNQs and the subjective perception of annoyance is fair
influence that each frequency range has on the subjective percep- to good. In particular, the Rw is shown, on average, as the best
tion of annoyance, in order to improve the SNQs in a way that bet- existing SNQ among those used in this article. This result strongly
ter represents the subjective annoyance. supports previous researches [5,19] in which the Rw was also found
to be the best descriptor among those existing in most situations.
However, the correlation is no better than the 70% for the average
4.3. Linear regression in the best case scenario basis, which still gives room for improve-
ment. These correlations are in good agreement with those pre-
In this section, linear regression is presented between the per- sented in [26,4] even considering different sound samples,
ceived differences judged by the participants and the differences insulation elements and methodologies for the performance of
in terms of the existing SNQs, to address the particular correlation the test as well as for the analysis of the results.
between the subjective judgments of annoyance and the most used Especially interesting are the results for the sound sample Siren.
existing SNQs in the frequency range considered in this As it can be drawn from Table 5, the correlation between the sub-
experiment. jective judgments and the usual SNQs is higher for this sample for
For this purpose, the d’ values were used as dependent variable all the tested SNQs, being especially relevant for the Rw , the
and the differences between each possible combination of insula- Rw þ C1003150 , the Rw þ C1005000 and the STC. Considering the spec-
tion elements in terms of the different SNQs were used as indepen- tral features of the Siren sound sample and the relevance given by
dent variables. the existing reference curve to the frequency bands in which this
Table 5 shows the correlation, by means of the goodness-of-fit sample has its main energetic contribution, these high correlations
measure R2 , between each SNQ and each sound sample as well were to be expected.
as all these correlations on average for each SNQ (Column Avg). On the other hand, the correlation between the subjective judg-
These R2 values are presented corrected by the number of degrees ments for annoyance and the SNQs calculated on the basis of the
of freedom, to reduce as much as possible the dependence of the traffic reference curve (Tr) is rather low for all sound samples.
correlation on the number of observations. Also, asterisks are used Especially remarkable are the results of the sound samples in
in Table 5 to denote the degree of significance of each regression which vehicles are the main contribution (Traffic 1 and Traffic 2).
value. The number of asterisks used to represent each range of sig- As it is stated in the ISO 717–1:2013 [6] there are two different
D. de la Prida et al. / Applied Acoustics 168 (2020) 107432 9
Table 5
Regression coefficients (R2) for the linear regression between the subjective judgments and the selected existing SNQs for all sound samples.
Table 6
Regression values for each sound sample and the most correlated SNQs, Rw , Rw þ Coptliv ing and Rw þ Copturban
ception of annoyance. It is, however, strongly recommended per- [11] Park HK, Bradley JS, Gover BN. Evaluating airborne sound insulation in terms of
speech intelligibility. J Acoust Soc Am 2008;123(3):1458–71.
forming further experiments on this issue to increase the knowl-
[12] Parkin PH. Field measurements of sound insulation between dwellings. 33. HM
edge in this field as well as to improve the optimization curves Stationery Office; 1960.
through the open collaboration and sharing of the results between [13] Gray PG, Cartwright A. Parkin PH. Noise in three groups of flats with different
different institutions. floor insulations. HM Stationery Office; 1958.
[14] Bitter C, Weeren PV. Sound nuisance and sound insulation in blocks of
dwellings. Research Institute for Public Health Engineering TNO, Report No 24.
1955;.
CRediT authorship contribution statement
[15] Josse R, Plagnol R, Coblentz A, Alexandre M. Etude sociologique de la
satisfaction des occupants de locaux conformes aux règles qui sont
Daniel de la Prida: Conceptualization, Software, Methodology, supposées garantir un confort acoustique suffisant;.
[16] Hongisto V, Mäkilä M, Suokas M. Satisfaction with sound insulation in
Validation, Formal analysis, Investigation, Resources, Visualization,
residential dwellings–the effect of wall construction. Build Environ
Writing - original draft, Writing - review & editing. Antonio Ped- 2015;85:309–20.
rero: Conceptualization, Methodology, Validation, Investigation, [17] Kylliäinen M, Takala J, Oliva D, Hongisto V. Justification of standardized level
Resources, Writing - original draft, Writing - review & editing, differences in rating of airborne sound insulation between dwellings. Appl
Acoust 2016;102:12–8.
Supervision. María Ángeles Navacerrada: Conceptualization, [18] Park HK, Bradley JS. Evaluating standard airborne sound insulation measures
Methodology, Investigation, Resources, Writing - original draft, in terms of annoyance, loudness, and audibility ratings. J Acoust Soc Am
Writing - review & editing, Funding acquisition. Alexander Díaz- 2009;126(1):208–19.
[19] Rychtáriková M, Muellner H, Chmelík V, Roozen NB, Urbán D, Garcia DP, et al.
Chyla: Conceptualization, Investigation, Resources, Writing - orig- Perceived loudness of neighbour sounds heard through heavy and light-
inal draft, Writing - review & editing. weight walls with equal Rw + C50-5000. Acta Acustica United Acustica
2016;102(1):58–66.
[20] Rychtáriková M, Roozen B, Müllner H, Stani M, Chmelík V, Glorieux C. Listening
Declaration of Competing Interest test experiments for comparisons of sound transmitted through light weight
and heavy weight walls. Akustika 2013;19(1):8–13.
[21] Vian JP, Danner WF, Bauer JW. Assessment of significant acoustical parameters
The authors declare that they have no known competing finan-
for rating sound insulation of party walls. J Acoust Soc Am 1983;73
cial interests or personal relationships that could have appeared (4):1236–43.
to influence the work reported in this paper. [22] Vardaxis NG, Bard D. Review of acoustic comfort evaluation in dwellings: Part
III–airborne sound data associated with subjective responses in laboratory
tests. Build Acoust 2018;25(4):289–305.
Acknowledgments [23] van den Eijk J. My neighbour’s radio. In: Proc. 3rd Int. Congress Acoustics;
1959. p. 1041–1044.
[24] Northwood TD. Sound insulation and the apartment dweller. J Acoust Soc Am
This work was funded by the Spanish Ministry of Economy and 1964;36(4):725–8.
Competitiveness (MINECO) and the Spanish Ministry of Science, [25] Scholl W, Lang J, Wittstockh V. Rating of sound insulation at present and in
Innovation and Universities through the projects [BIA 2015- future. The revision of ISO 717. Acta acustica united with acustica. 2011;97
(4):686–698.
68914-R] and [RTI2018-094656-B-I00].
[26] Virjonen P, Hongisto V, Oliva D. Optimized single-number quantity for rating
the airborne sound insulation of constructions: Living sounds. J Acoust Soc Am
References 2016;140(6):4428–36.
[27] Monteiro C, Machimbarrena M, de la Prida D, Rychtarikova M. Subjective and
objective acoustic performance ranking of heavy and light weight walls. Appl
[1] Yaniv SL, Flynn DR. Noise criteria for buildings: a critical view. 499. US Dept. of Acoust 2016;110:268–79.
Commerce, National Bureau of Standards: for sale by the Supt. of...; 1978. [28] Mašović DB, Šumarac Pavlović DS, Mijić MM. On the suitability of ISO 16717–1
[2] Moreno A. Factors influencing dB (A) ratings for sound insulation: incident reference spectra for rating airborne sound insulation. J Acoust Soc Am
noise spectrum and shape of the transmission loss curve. J Sound Vibr 1984;97 2013;134(5):420–5.
(2):337–48. [29] Rasmussen B, Rindel JH. Sound insulation between dwellings–Descriptors
[3] Park HK, Bradley JS. Evaluating signal-to-noise ratios, loudness, and related applied in building regulations in Europe. Appl Acoust 2010;71(3):171–80.
measures as indicators of airborne sound insulation. J Acoust Soc Am 2009;126 [30] Rasmussen B, Gutiérrez MM. Existing Sound Insulation Performance
(3):1219–30. Requirements and Classification Schemes For Housing Across Europe. In:
[4] Hongisto V, Oliva D, Rekola L. Subjective and objective rating of the sound Towards a common framework in building acoustics throughout Europe;
insulation of residential building façades against road traffic noise. J Acoust Soc 2013. p. 25–48.
Am 2018;144(2):1100–12. [31] de la Prida D, Pedrero A. Ángeles Navacerrada M, Díaz C. Relationship between
[5] Hongisto V, Oliva D, Keränen J. Subjective and objective rating of airborne the geometric profile of the city and the subjective perception of urban
sound insulation–living sounds. Acta Acustica United Acustica 2014;100 soundscapes. Appl Acoust 2019;149:74–84.
(5):848–63. [32] Díaz C, Pedrero A. An experimental study on the effect of rolling shutters and
[6] ISO 717-1:2013 Acoustics – Rating of sound insulation in buildings and of shutter boxes on the airborne sound insulation of windows. Appl Acoust
building elements – Part 1: Airborne sound insulation. Geneva, Switzerland: 2009;70(2):369–77.
International Organization for Standardization; 2013. [33] Díaz C, Díaz A, Navacerrada MA. An experimental study on the effect of rolling
[7] E413-10: Classification for Rating Sound Insulation. West Conshohocken, PA, shutters on the field measurements of airborne sound insulation of façades.
USA: ASTM International; 2010. Appl Acoust 2013;74(1):134–40.
[8] de la Prida D, Pedrero A, Ángeles Navacerrada M, Díaz-Chyla A. Methodology [34] Lee HS, O’Mahony M. Sensory difference testing: Thurstonian models. Food Sci
for the subjective evaluation of airborne sound insulation through 2-AC and Biotechnol 2004;13(6):841–7.
Thurstonian models. Appl Acoust 2020;157:107011. [35] Christensen RHB, Brockhoff PB. sensR–An R-package for sensory
[9] Brandt O. European Experience with Sound-Insulation Requirements. J Acoust discrimination; 2018.
Soc Am 1964;36(4):719–24. [36] Team RC. R: A language and environment for statistical computing. 2013;.
[10] Tachibana H, Hamada Y, Sato F. Loudness evaluation of sounds transmitted [37] Christensen RHB, Lee HS, Brockhoff PB. Estimation of the Thurstonian model
through walls–basic experiment with artificial sounds. J Sound Vibr 1988;127 for the 2-AC protocol. Food Quality Preference 2012;24(1):119–28.
(3):499–506.