F r e q u e n cy w eig h tin gs b a s e d on s u bjec tively do min a n t s p ec t r al

r a n g e sTorija M a r tin ez, AJ, Flind ell, IH, S elf, RH a n d Torija M a r tin ez,

Antonio J

Tit l e F r e q u e ncy w eig h tings b a s e d on s u bjec tively do min a n t s p ec t r al r a n g e s

Aut h or s Torija M a r tin ez, AJ, Flind ell, IH, S elf, RH a n d Torija M a r tin ez, Antonio J

Typ e Confe r e nc e o r Works ho p It e m

U RL This ve r sion is available a t : h t t p://usir.s alfor d. ac.uk/id/e p rin t/53 2 4 1/

P u bl i s h e d D a t e 2 0 1 5

U SIR is a digi t al collec tion of t h e r e s e a r c h ou t p u t of t h e U nive r si ty of S alford. Whe r e copyrigh t p e r mi t s, full t ex t m a t e ri al h eld in t h e r e posi to ry is m a d e fre ely availabl e online a n d c a n b e r e a d , dow nloa d e d a n d copied for no n-co m m e rcial p riva t e s t u dy o r r e s e a r c h p u r pos e s . Ple a s e c h e ck t h e m a n u sc rip t for a ny fu r t h e r copyrig h t r e s t ric tions.

For m o r e info r m a tion, including ou r policy a n d s u b mission p roc e d u r e , ple a s econ t ac t t h e Re posi to ry Tea m a t : u si r@s alford. ac.uk .

mailto:usir@salford.ac.uk

Frequency weightings based on subjectively

dominant spectral ranges

Antonio J. Torija, Ian H. Flindell, Rod Self

ISVR, University of Southampton, SO17 1BJ Southampton, UK.

Summary

The optimisation of frequency weightings for the assessment of environmental and transportation

noise has been an enduring research theme, with no real consensus achieved, other than increasing

acknowledgement that the widespread adoption of the A-frequency weighting is an essentially

arbitrary choice. Based on the results of 4 illustrative case studies, we propose a simple

hypothesis that the 'best' frequency weighting in any specific context depends on which part of the

frequency spectrum is subjectively dominant. This hypothesis helps to explain why different

frequency weighting schemes appear to work 'best' in different situations and allows the selection

of frequency weightings for use in assessment procedures to be carried out on a more rational and

possibly less partisan basis. Further work is of course necessary to develop and extend our

hypothesis to a wider range of different circumstances.

PACS no. 43.50.Rq, 43.50.Qp, 43.50.Lj

1. Introduction1

The significant influence of frequency spectrum-

related factors on perceived annoyance has been

highlighted by several studies carried out on

transportation noise [1,2].

Even with the extensive attention in the literature,

and widespread adoption in standards and

regulations, still there is a lack of consensus about

the frequency filter to be applied for evaluating

transportation noise. Based on the physical

dominance of the different frequency ranges: low

(20-250Hz), mid (315-2000 Hz) and high (2500-

20000 Hz) frequencies; diverse and, often

opposing results regarding the most influential

spectral region on perceived annoyance have been

found by different research studies [2-6]. Because

of this disparity of results, authors have pointed

out one or another of the existing frequency

weightings as the most appropriate for assessing

transportation noise annoyance. All these findings

seem to indicate that still there is an open debate,

and that the adoption of the A-frequency weighting

by most national standards is an essentially

arbitrary choice.

On the other hand, since the main purpose of

transportation noise assessment is to estimate the

magnitude of community annoyance, the issue

arising should not be the identification of the

physically dominant frequency spectrum range, but

the identification of the one perceived as

subjectively dominant.

The hypothesis underlying this research is that the

'optimal' frequency weighting in any specific

context depends on which part of the frequency

spectrum is subjectively dominant, and it is based

on the findings presented by Torija and Flindell

[7], where it was stated that whichever is the

physically dominant part of the frequency

spectrum is not necessarily a good guide to what

part of the spectrum will be perceived as

subjectively dominant.

In order to illustrate this hypothesis the results of 4

laboratory studies are presented and discussed.

Thus, the performance of the application of A- and

D-weighting filters is analyzed for assessing

annoyance evoked by (i) urban road traffic noise

under outdoor and indoor conditions and (ii) by

urban traffic noise with low and high low-

frequency content after the erection of different

noise barriers.

2. Road traffic noise under outdoor and indoor conditions

2.1. Laboratory study under outdoor

conditions

In [7] a recording of continuous urban road traffic

noise with physically prominent frequency

components at low-frequency and middle/high-

Copyright© (2015) by EAA-NAG-ABAV, ISSN 2226-5147All rights reserved

1645

frequency 3rd-octave bands (Fig. 1), was used as

the basis for all sounds reproduced in the listening

tests. Three frequency filters were applied for

boosting or cutting the low- frequency (LF), mid-frequency (MF), or high-frequency (HF) ranges. Twelve filtered sounds were produced by applying each filter with -9dB, -3 dB, +3dB and +9 dB relative gain setting.

Figure 1. Frequency spectra of master recording road

traffic noise stimuli used in outdoor and indoor

laboratory studies.

A five point semantic scale, according to standard ISO/TS 15666 [8]: “Not at all”, “Slightly”, “Moderately”, “Very”, and “Extremely”, was used for assessing perceived annoyance.

2.2. Laboratory study under indoor conditions

In [9] a continuous urban road traffic noise was

filtered for simulating typical frequency dependent

attenuation of double glazing sealed units made up

from 3mm glass, 3 mm air gap, and 3 mm glass,

according to the valued reported in [10].

Moreover, artificial reverberation at 0.5 second

reverberation time was added to increase the

subjective realism of the intended indoor

simulation. The outdoor and indoor frequency

spectra are shown at Fig. 1. Low pass and high

pass shelf filters were applied for boosting or

cutting the LF, or mid/high-frequency (MHF)

ranges by +9 dB, +3 dB, -3 dB, and -9 dB to the simulated indoor filtered sound, thereby producing 8 different sounds for the listening tests. In this laboratory study, the perceived annoyance was assessed using the relative magnitude estimation (RME) method. According to this method, the participants rated subjective annoyance of each stimulus numerically against a defined reference sound which was given an arbitrary rating of 100 (so that each reported annoyance value was referred to 100).

2.3. A- and D-weighting for assessing road

traffic noise annoyance

In Fig. 2, it is observed that the application of the

D-filter (right) for frequency weighting the sound

level achieves higher correlations with reported

annoyance than the A-filter (left) for the specific

sounds used in [7]. In [7], for the filtered road

traffic sounds tested, and under outdoor

conditions, the MF and especially HF contents

were found to be subjectively dominant. However,

as indicated by the results shown in Fig. 2, the

relatively high LF content in urban road traffic

noise should not be neglected. Under outdoor

conditions, the D-weighting better accounts for the

difference in contribution to road traffic evoked

subjective annoyance of LF, MF and HF ranges.

Meanwhile, the A-weighing underestimates the

contribution of LF and HF ranges, as can be seen

in Fig. 2 – left (triangle and circle symbols).

Fig. 3 shows the linear relationship between A-

and D-weighted sound levels and reported

annoyance for the stimuli used in the listening test

presented in [9]. In [9], even under conditions

where LF content was physically dominant (indoor

conditions), it was found that changes in LF

content made smaller contributions to reported

annoyance than might be inferred from such

physical dominance. Indeed, under the tested

indoor conditions, equivalent changes in LF and

MHF content led to similar changes in reported

annoyance. Under these conditions, and as it is

seen in Fig. 3, the A-weighting filter accounts for

the difference in contribution to subjective

annoyance between LF and MHF ranges, while D-

weighting filter overestimates the contribution of

LF.

-10

0

10

20

30

40

50

60

70

31

,5

50

8

0

12

5

20

0

31

5

50

0

80

0

12

50

2

00

0

31

50

5

00

0

80

00

1

25

00

So

un

d L

ev

el

(dB

)

1/3rd-octave Bands (Hz)

Road Traffic Outdoor

Road Traffic Indoor

EuroNoise 201531 May - 3 June, Maastricht

A.J. Torija et al.: Frequency...

1646

Figure 2. Linear relationship between the A-weighted

(left) and D-weighted (right) sound levels and

perceived annoyance. Triangle, square and circle

symbols correspond to LF, MF and HF filter gain,

respectively.

Figure 3. Linear relationship between the A-weighted

(left) and D-weighted (right) sound levels and

perceived annoyance. Triangle and circle symbols

correspond to LF and MHF filter gain, respectively.

3. Urban road traffic noise with noise barriers

3.1. Stimuli and procedure

For this laboratory study, two master recordings of

continuous urban road traffic noise were selected,

one with relatively high LF content (similar to the

MF content), hereafter called RT1, and another

one with relatively low LF content (as compared

to the MF content), hereafter called RT2. These

two master recordings were used as the basis for

synthesising all the experimental sounds for the

listening tests.

To each of these two road traffic sounds, a series

of frequency filters were applied for simulating

the insertion loss (IL) generated by the presence of

a selection of 10 noise barriers. The IL provided

for each of the noise barriers was obtained from

[11]. It should be noted that in [11], only 5 noise

barriers were presented, so that the other 5 noise

barriers simulated here were derived by keeping

the same spectral pattern but reducing the

simulated IL by 6 dB. In Fig. 4, the frequency

spectra of each of the original master recording

and of each of the filtered sound are shown.

The 30 participants in this listening experiment

assessed the annoyance evoked by all the filtered

sounds using the RME method. Thus, for each

master road traffic sound, RT1 and RT2, the

participants rated subjective annoyance of each of

the 10 stimuli (road traffic sounds filtered to

simulate the presence of the 10 noise barriers)

numerically against the reference sounds which

were give an arbitrary rating of 100.

3.2. A- and D-weighted sound levels vs.

perceived annoyance

As can be seen in Table I, for the road traffic noise

with low LF content (RT2), the application of A

and D filters for the frequency weighting of the

sound level achieves similar results in assessing

the perceived annoyance of the 10 filtered noise

barrier sounds.

PA = 0,1104·LAeq - 4,6554

R² = 0,5019

1,5

2

2,5

3

3,5

4

4,5

55 60 65 70 75 80

Pe

rce

ive

d A

nn

oy

an

ce (

Fiv

e-P

oin

ts

Sca

le U

nit

s (1

to

5))

A-weighted Sound Level (dBA)

PA = 5,66·LAeq - 187,44 R² = 0,97

PA = 4,30·LDeq - 166,23 R² = 0,84

60

80

100

120

140

160

40 50 60 70 80

Re

lati

ve

Ma

gn

itu

de

Est

ima

tio

n o

f P

erc

eiv

ed

A

nn

oy

an

ce

Sound Level

PA = 0,11·LAeq - 4,66

R² = 0,50

1,5

2

2,5

3

3,5

4

4,5

55 60 65 70 75 80

Pe

rce

ive

d A

nn

oy

an

ce (

Fiv

e-P

oin

ts

Sca

le U

nit

s (1

to

5))

A-weighted Sound Level (dBA)

PA = 0,17·LDeq - 9,43 R² = 0,78

60 65 70 75 80 85

D-weighted Sound Level (dBD)

EuroNoise 201531 May - 3 June, Maastricht

A.J. Torija et al.: Frequency...

1647

Figure 4. Frequency spectra of the original and

frequency filtered sounds to simulate the presence of

the different noise barriers, for master urban road

traffic noises 1 (RT1 – high LF content) and 2 (RT2 –

low LF content).

However, when the road traffic noise has high LF

content (RT1), higher correlations with reported

annoyance are obtained by using the D-weighting

filter than by using the A-weighting filter. These

results are consistent with the results presented in

Section 2.3, i.e. for typical road traffic noise under

outdoor conditions, the frequency regions

subjectively dominant are MF, and especially HF,

but if there is a strong component in the LF

region, this should not be neglected.

With high LF content, simulation of the presence

of the different noise barriers made the LF region

subjectively more important, and as observed in

Table I, the A-weighting filter then

underestimated the effect of the LF contribution,

whilst the D-weighting filter was able to give a

better representation of the LF contribution.

Table I. Results of the Linear Regression analysis

(N=10) for estimating perceived annoyance from A-

weighted and D-weighted sound levels. p≤ 0.05.

RT1

R2 Adjusted

R2

Standard Error of the Estimate

A-weighting 0.83 0.81 3.66

D-weighting 0.90 0.89 2.80

RT2

R2 Adjusted

R2

Standard Error of the Estimate

A-weighting 0.97 0.97 2.32

D-weighting 0.97 0.97 2.45

4. Conclusions

In this paper, 4 case studies are presented to

illustrate the hypothesis that the 'optimal'

frequency weighting in any specific context

depends on which part of the frequency spectrum

is subjectively dominant. According to the results

presented here, the selection of the most

appropriate frequency weighting filter should not

be made solely on the basis of which frequency

region is physically dominant, but should also take

into subjective dominance into account.

10

20

30

40

50

60

70

31,5 63 125 250 500 1000 2000 4000 8000 16000

So

un

d L

ev

el

(dB

)

Octave Bands (Hz)

RT1

Original NB_Timber NB_Metal NB_Transparent NB_Concrete NB_Vegetation NB_Timber_2 NB_Metal_2 NB_Transparent_2 NB_Concrete_2 NB_Vegetation_2

31,5 63 125 250 500 1000 2000 4000 8000 16000

Octave Bands (Hz)

RT2

EuroNoise 201531 May - 3 June, Maastricht

A.J. Torija et al.: Frequency...

1648

Differences in the relative contributions to

subjective annoyance made by different spectral

regions, i.e. low-, mid- and high-frequencies, are

important for the selection of optimum frequency

weightings.

This finding may help to explain why different

frequency weighting curves appear to work 'best'

in different situations, but could also inform the

selection of frequency weightings for use in

assessment procedures to be carried out on a more

rational and possibly less biased basis.

Of course, further work would be necessary to

develop and extend this research hypothesis to a

wider range of different circumstances, i.e. across

a wider range transportation and other noise

sources and contexts.

Acknowledgement

This work is funded by the University of Malaga

and the European Commission under the

Agreement Grant no. 246550 of the seventh

Framework Programme for R & D of the EU,

granted within the People Programme, «Co-

funding of Regional, National and International

Programmes» (COFUND), and the Ministerio de

Economía y Competitividad (COFUND2013-

40259).

References

[1] G. Leventhall: A review of published research on low frequency noise and its effect. Report for Department for Environment, Food and Rural Affairs (DEFRA), May 2003. Available at http://archive.defra.gov.uk/environment/quality/noise/research/lowfrequency/documents/lowfreqnoise.pdf. (date last viewed 11/06/14).

[2] H. G. Leventhall: Low frequency noise and annoyance. Noise and Health 6 (2004) 59-72.

[3] M. E. Nilsson: A-weighted sound pressure level as an indicator of perceived loudness and annoyance of road-traffic sound. Journal of Sound and Vibration 302 (2007) 197-207.

[4] G. R. Watts: A comparison of noise measures for assessing vehicle noisiness. Journal of Sound and Vibration 180 (1995) 493-512.

[5] N. J. Versfeld, J. Vos: Annoyance caused by sounds of wheeled and tracked vehicle. Journal of the Acoustical Society of America 101 (1997) 2677-2685.

[6] J. Kim, C. Lim, J. Hong, S. Lee: Noise-induced annoyance from transportation noise: Short-term responses to a single noise source in a laboratory. Journal of the Acoustical Society of America 127 (2009) 804-814.

[7] A. J. Torija, I. H. Flindell: Differences in subjective loudness and annoyance depending on the road traffic noise spectrum. Journal of the Acoustical Society of America 135 (2014) 1-4.

[8] ISO, Acoustics-Assessment of noise annoyance by means of social and socio-acoustic surveys.

ISO/TS15666:2003(E). 2003, ISO: Geneva, Switzerland.

[9] A. J. Torija, I. H. Flindell: The subjective effect of low frequency content in road traffic noise. Journal of the Acoustical Journal of America 137 (2015) 189-198.

[10] J. D. Quirt: Sound transmission through windows II. Double and triple glazing. Journal of the Acoustical Society of America 74 (1983) 534-542.

[11] J. Y. Hong, H. S. Jang, J. Y. Jeon: Evaluation of noise barriers for soundscape perception through laboratory experiments. Proceedings of the Acoustics 2012, Nantes, France, 2012, pp. 2159-2162.

EuroNoise 201531 May - 3 June, Maastricht

A.J. Torija et al.: Frequency...

1649