Sage Journals: Discover world-class research

Abstract

The concept of acoustic comfort is hardly defined and used to refer to conditions of low noise levels or annoyance based on standardized descriptors. Airborne and impact sound measurements are used to rate acoustic comfort in dwellings, but they often do not express human perception of noise or comfort. If the descriptors are statistically associated with self-reported responses, they can be used as prediction models and considered sufficient for acoustic comfort assessment. This review article presents studies that approach acoustic comfort in dwellings via the association of acoustic data and subjective responses in laboratory tests. Specifically, we investigate the cases of impact sound, since it is usually reported as the most disturbing noise source in dwellings. We also evaluated the reviewed studies with the Bradford Hill’s criteria. The reviewed studies indicate that self-reported annoyance to impact sound is an important issue and it can be predicted well in overall. Various standardized descriptors are studied and associate sufficiently with subjective responses. Inclusion of low frequencies down to 50 Hz in measurements improves the association of impact sound descriptors to subjective responses. Some impact noise stimuli associate only with some descriptors but not all. From the standardized impact sources, the tapping machine is the most efficient to predict overall annoyance and the impact ball for human walking or typical impact sounds in dwellings.

Keywords

Acoustic comfort impact sound laboratory subjective responses association evaluation

Introduction

This article concerns a review of acoustic comfort evaluation for dwellings in laboratory tests. The reviewed publications present studies which were conducted in laboratory conditions and evaluate the association of acoustic data with subjective responses and thus approach acoustic comfort perception. Since impact sound has been reported in the literature as the most important noise source in dwellings,¹ this review is focused only on impact sound studies and results. The examined laboratory tests usually include acoustic data of measured sound insulation or recorded noise sounds of various types, which are deployed in controlled listening experiments where the subjects, that is, the participants, offer their self-reported responses.^2–15 In some of the presented cases, the acoustic data of the reviewed studies originate from field measurements or sound recordings in real buildings and not laboratory measurements. However, those data are still processed and used for listening experiments within a laboratory setup under controlled conditions in the reviewed studies.

Acoustic comfort is vaguely defined in the literature, despite being an important concept in engineering. It is typically used to denote a state of low or no noise and therefore lack of annoyance for the residents. A complete definition is provided in Rasmussen and Rindel¹ as “a concept that can be characterized by absence of unwanted sound, desired sounds with the right level and quality, opportunities for acoustic activities without annoying other people.”

Standardized measurements and relevant descriptors are used to assess building acoustic conditions. They do not always represent well how people perceive the living sound environment as occupants in their flats. Previous studies have shown that residents suffer from impact noise types: such noise types have dominant low-frequency characteristics which are usually neglected in a standardized measurement with a typical frequency range of 100–3150 Hz. Also, the impact sound sources used during measurements might offer different types of excitation than the real-life impact sounds. Then, there are various types of building constructions and components, which provide different structural and acoustical conditions to the tenants.^3–15

Therefore, it is important to test the association of the acoustic data from measured results to self-report responses; that association is tested with statistical analyses comparing objective and subjective data in many studies in this review. Sometimes, alternative versions of standardized descriptors are suggested in order to achieve better agreement of acoustic data with subjective responses. If a strong association can be established, then it is possible to formulate models for prediction of annoyance and comfort for the residents.

The understanding of acoustic comfort and development of prediction models would be essential for the design of proper acoustic conditions in buildings. For all the above reasons, comparing measured data to human perception is essential for the characterization of acoustic comfort in overall. In this review article, a set of selected studies are presented dealing with impact sound data compared and associated with subjective responses collected in laboratory tests.

Methods

A wide search for peer-reviewed publications and conference proceedings, which include examination between acoustic data and self-reported responses relevant to impact sound, has been done in the following databases: ScienceDirect, AIP Scitation, Ingenta Connect, ResearchGate, PubMed, Scopus, and Google Scholar. The search strategy included numerous searches in the databases using relevant keywords, such as objective, subjective, acoustic, psychoacoustic, self-report, rating, score, comfort, quality, impact, sound, insulation, noise, annoyance, assessment, association, and correlation. Several publications were subsequently found as references of the first selected papers.

Finally, this review article includes 10 Asian studies,^2–11 1 Canadian study,^12,13 and 4 European studies.^14–17 Requirements for inclusion of papers in this review were the comparison of results between impact sound measured data and subjective responses collected from tests in laboratory experiments. Overall, 37 papers were found during the search in databases or relevant references and were evaluated by title name, abstract reading full reading; 17 of them were included in this review. The selection was based on their relevance to this review: some publications did not offer statistical comparisons or did not consider impact sound laboratory tests and thus were excluded.^18–37 Other exclusion criteria were the year of publication and language: only articles published after 2000 in English were included. The bibliographic research took place between April 2015 and September 2017.

Summary of methods, metrics, and quantities in the reviewed studies

Many different indicators (or descriptors) have been used to represent different quantities in acoustic measurements. They are all standardized in international ISO standards or other compliable national standards. Many variations of them exist as well, since experimental research has been done to acquire better indicators than the standardized ones. A description of all indicators involved in this review is presented in Table 1. For the full methods used to acquire and calculate the indicators, please see the relevant standards.

Table 1.

List of acoustic indicators used in the review studies.

Indicator	Description	Standards	References
ACF	Autocorrelation function: correlation of a time signal with delayed versions of itself	–	2,3,7
IACC	Interaural cross-correlation function: covariance of delayed versions of the left and right ear time signal	–	2,3,8
L	Loudness: sound quality (SQ) metric defined by Zwicker & Fastl	ISO 532:1975	3,6,7,11,38,39
N₅, N₁₀	Percentile loudness: SQ metric defined by Zwicker & Fastl	–	5,6,11,39
N_max	Maximum loudness: SQ metric defined by Zwicker & Fastl	–	6,39
FS	Fluctuation strength: SQ metric defined by Zwicker & Fastl	–	3,7,39
T	Tonality: SQ metric defined by Zwicker & Fastl	–	3,6,39
UA	Unbiased annoyance: SQ metric defined by Zwicker & Fastl	–	3,39
S	Sharpness: SQ metric defined by Zwicker & Fastl	–	6,39
R	Roughness: SQ metric defined by Zwicker & Fastl	–	6,39
SPL	Sound pressure levels		5,8,9,10
L_Aeq	A-weighted sound pressure level in dB, equivalent to the total sound energy over a specific period of time	JIS A 1418, KS F 2810-2	4,5,40,41
L_Amax	Maximum A-weighted sound pressure level	JIS A 1418, KS F 2810-2	4,5,8,9,10,11,40,42
L _{i,Fmax, AW}	Maximum A-weighted impact source level	JIS A 1418, KS F 2810-2	6,8,9,10,11,12,13,40,42
DR	Decay rate: similar to reverberation time but for impact sounds		8,9,10
JND	Just noticeable difference		7,8,9
L_n,w	Impact sound insulation index characterizing a building element (laboratory measurements)	ISO 717-2, ISO 140-7, EN ISO 12354-2, ISO 16283-2	12,13,41,43,44,45
L′_n,w	Apparent impact sound insulation index (same as $L_{n, w}$ for field measurements)	ISO 717-2, ISO 140-7, EN ISO 12354-2, ISO 16283-2	14,15,16,17, 41,43,44,45
C	C is an A-weighted pink noise spectrum adaptation term	ISO 717-1, 717-2, EN ISO 12354-1, 12354-2	12,13,14,15,16,17,41,43,44

Several statistical methods such as analysis of variance (ANOVA), regression analysis, and principal component analysis (PCA) associate acoustic data to subjective responses. Details on the statistical methods can be found in relevant literature. The quality of statistical association is usually described with typical parameters such as the correlation coefficient, denoted as $r, ρ$ , or $R$ , and the coefficient of determination, denoted as $R^{2}$ . The p values and the confidence intervals (CIs) are measures of statistical significance.

Evaluation of included studies

The quality of evidence for studies in this review was evaluated Bradford Hill’s criteria²⁶ which is an evidence classification method often used in epidemiology and health review studies. The fulfilled criteria are rated in a scale of High (+++), Moderate (++), and Low (+). The evaluations are tabulated in Table 2, while the criteria are as follows.

Strength of association: it refers to the causality proven by the association between the studied variables (cause, effect size, and confounding factors).

Consistency: it indicates the degree of certainty when similar results are observed by different studies in different tests.

Specificity: specific factors and effects on a specific population lead to a more likely causal relationship.

Temporality: it is based on temporal relations between effects and used as an indicator for causality, meaning one effect happening after an exposure.

Biological gradient: it refers to the relation between exposure and effect; usually bigger exposure leads to greater effect, but not always, while the opposite outcome can happen as well.

Plausibility: it means that a biological explanation of why a cause leads to a certain effect supports a reasonable causality.

Coherence: it is a condition meaning that a stated causal relationship should not contradict with other accepted results or knowledge.

Experiment: it refers to the study design parameters that guarantee a reasonable causation, like randomization.

Analogy: the possibility of having or predicting analogous effects from similar factors without total evidence.

Publication type: an additional criterion in order to rank the reviewed studies. Scientific journal papers are thoroughly peer reviewed, while conference papers are usually less well reviewed. There are study reports from research organizations that may be scientifically well conducted but not reviewed at all. There are others, for example, unofficial reports, which are excluded. Thus, publications were evaluated as scientific journal (+++), conference paper (++), and report (+).

Table 2.

Evaluation of the presented studies according to selected criteria.

Reference number	Publication type^a	Strength of association	Consistency	Specificity	Temporality	Biological gradient	Plausibility	Coherence	Experiment design	Analogy
Yeon²	++	+	++	+	+	++	++	+	++	++
Yeon and Jeong³	+++	+	+	+	+	+	+	+	++	+
Jeon et al.⁴	+++	++	++	++	++	+++	++	++	++	++
Jeon et al.⁵	+++	++	++	++	++	++	++	++	++	++
Lee et al.⁶	+++	+++	+++	+++	+++	++	++	+++	+++	+++
Jeon and Sato⁷	+++	++	++	++	++	++	++	++	++	++
Jeon et al.⁸	+++	++	+++	++	+++	++	++	++	++	++
Kim et al.⁹	+++	++	++	++	+	+	++	+	+	++
Jeon and Oh¹⁰	++	+	+	+	+	++	+	+	+	+
Ryu et al.¹¹	+++	++	++	+	++	+	+	++	+	+
Gover et al.¹²	++	+	++	++	++	+	++	++	++	++
Gover et al.¹³	++	+	++	++	++	+	++	++	++	++
Späh et al.¹⁴	+++	++	++	++	++	++	++	++	++	++
Kylliäinen et al.¹⁵	++	++	++	++	++	+++	++	++	+++	++
Kylliäinen et al.¹⁶	+++	++	++	++	++	+++	++	++	+++	++
Öqvist et al.¹⁷	+++	+++	++	++	++	++	++	++	+++	++

Types: scientific journal (+++), conference paper (++), and report (+).

The included studies were evaluated by the authors of this article, while the presented data were chosen according to their relation and importance for this review’s context. In Table 3, an overview of all the selected studies can be found, which are tabulated with summary of results, study design, methods, and conclusions. In Table 2, the evidence evaluation rating of the studies is presented according to the above criteria. Readers who would like to have a deeper insight into any specific study results or conclusions may read the original publications using the references. Essential information might also be missing from this review if they are not reported in the papers. The studies are presented in chronological order and analyzed in the next chapter.

Table 3.

Overview of studies’ summaries.

Reference number	Samples and experiment details	Variables as defined in studies	Models and results	Parameters for evaluation	Summary of results
Yeon²	20 subjects in a listening test (age = 21–31 years)	Independent: In situ sound recordings of impact sources:– bang machine (tire)– tapping machineDependent:V1: subjective loudness and noisiness	Test samples matching to pink noise in steps of 3 dB	No details provided	– Bang machine was perceived 6–7 dB noisier and louder than the tapping machine, which is too loud for an impact source.– ACF magnitude Φ (0) important for loudness and noisiness perception
Yeon and Jeong³	30 subjects in a listening test (27 males, 3 females, age = 24–41 years)	Independent:In situ sound recordings of impact sources:– Bang machine (tire)– Tapping machine– Impact rubber ballACF/IACCZwicker parametersDependent:SV: Subjective loudness	Correlations:– Tapping machine noise $S V$ to Φ(0); $S V$ to L—loudness; $S V$ to UA—unbiased annoyance;– Bang machine noiseSV to Φ(0); $S V$ to L —loudness; $S V$ to UA—unbiased annoyance;– Impact rubber ball noise $S V$ to Φ(0): $S V$ to L—loudness: $S V$ to UA—unbiased annoyance:Multiple regression models:SV_tapping = −17.761 + 0.065 Φ(0) + 11.51 τ₁ − 1.45ϕ₁ SV_tapping = −5.731 + 0.25 L + 2.23 FS + 1.16 T−0.0076 UA SV_Bang = −3.691 + 0.147 Φ (0) − 0.251 τ_e − 3.83 W_IACC SV_Bang = −0.534 + 0.22 L SV_Ball = −4.754 + 0.121 Φ (0)–0.2021 τ_e − 1.01 IACC + 0.992 τ_IACC SV_Ball = −1.431 + 0.177 L + 0.24 FS − 0.0012 UA	Coefficients (all p < 0.01):γ = 0.96γ = 0.94γ = 0.92γ = 0.94γ = 0.74γ = 0.72γ= 0.94γ= 0.94γ = 0.76Total coefficients (all p < 0.05): γ = 0.94γ = 0.98γ = 0.96γ = 0.74γ = 0.98γ = 0.95	– Subjective response of loudness was highly correlated with the maximum ACF amplitude Φ (0) for all impact sources– Subjective response of loudness was highly correlated with Zwicker’s Loudness and Unbiased annoyance.– Zwicker parameters are more reasonable for the tapping machine noise while for the bang machine and impact ball cases Φ (0) seems to affect more the loudness perception.– Spatial factors, pitch and sound energy are also significant parameters as indicated by the multiple regression models
Jeon et al.⁴	60 subjects in a listening test(30 Korean, 30 German)56 sound samples tested	Independent:Two standardized impact sources:S1: tapping machine $L_{e q}$ S2: tire machine L_max $L_{m a x}$ Measured and recorded below of eight types of floor structures in various configurations with insulation in the sending roomDependent: V1: loudness responsesV2: annoyance responses	Correlations:– V1 (loudness) to: $L_{e q}$ (S1) $L_{m a x}$ (S2)– V2 (annoyance) to: $L_{e q}$ (S1) $L_{m a x}$ (S2)– Impact noise differences: $Δ d B$	Coefficients:R² = 0.64 above 250 HzR² = 0.49–0.81 above 63 HzR² = 0.36 above 250 HzR² = 0.25–0.49 above 63 HzR² = 0.55 for KoreansR² = 0.55 for Germans	– Good ratings for the cases of sending rooms with insulation on floor and walls and for the extra suspended ceiling insulation.– Good associations in general above 250 Hz for the tapping machine and slightly good above 63 Hz for the tire machine.– Koreans are more sensitive to impact sounds due to bigger exposure
Jeon et al.⁵	30 subjects (students) in a listening test48 sound samples testedAdditional 98 subjects in on-site auditory experiment and 10 subjects in a listening experiment combined	Independent:Two standardized impact sources:S1: impact ballS2: tapping machineMeasured and recorded below of eight types of floor structures in various configurations with insulation in the sending roomDependent: V1: loudness responses	Regression models:– V1 (loudness) to:Mean SPL $L_{A m a x}$ $I n v e r s e - A$ $L_{n, w} + C_{I, 63 - 2000}$ $L o u d n e s s$ $P e r c e n t i l e l o u d n e s s N_{10}$	Coefficients:S1: R² = 0.57, S2: R²= 0.84S1: R² = 0.70, S2: –S1: R² = 0.69, S2: R² = 0.79S1: –, S2: R² = 0.73S1: R² = 0.74, S2: R² = 0.84S1: R² = 0.77, S2: R² = 0.88	– Tapping machine offers better association with subjective ratings than impact ball;– The impact ball spectra were found to be the most similar to real impact sounds in multistory residential buildings.– Three classes were suggested according to the $L_{i, F m a x, A W}$ levels: Class 1 (<44 dB), Class 2 (<49 dB) and Class 3 (<54 dB)
Lee et al.⁶	40 subjects in a listening test (28 males and 12 females, age = 24–35 years)54 sound samples tested	Independent:Impact ball excitation measured and recorded below of 35 floor structures. Descriptors used: $L_{i, F m a x, A W}$ , $L_{- n u m b e r}$ , $L_{A e q}$ $L_{A m a x}$ , $L L_{Z}$ , $N_{m a x}$ , $N_{5}$ , $L_{1 / 1 (63 - 500 Hz)}$ , $A S E L$ V1: annoyance responses in pairwise comparisons	Regression models:– V1 (annoyance) to: $L L_{Z}$ $L_{A m a x}$ $L_{i, F m a x, A W}$ Loudness $(L)$ Fluctuation strength $(F)$ Multiple linear regression: $S V_{a n n o y a n c e} = 0.77 L + 0.15 F$	Coefficients:r = 0.97, R² = 0.94, p < 0.05r = 0.92, R² = 0.85, p < 0.01r = 0.88, R² = 0.77, p < 0.01r = 0.81, R² = 0.66, p < 0.01r = 0.90, R² = 0.81, p < 0.05r = 0.90, R² = 0.81, p < 0.05	– Very good correlation with all metrics tested; $L_{A m a x}$ is suggested as the most practical descriptor. Sound quality metrics are difficult to derive.– Three dimensions revealed in a factor analysis; “1: reverberance and spaciousness,” “2: dullness,” and “3: loudness”
Jeon and Sato⁷	40 subjects in a listening test(20 students and 20 housewives)28 pairs of sound samples tested	Independent:Impact ball and bang machine excitation measured and recorded below of six floor structures. Descriptors: $L_{i, F m a x, A W}$ , ACF, and SQ metricsDependent:V1: annoyance	Correlations:– V1 (annoyance) to $Φ (0)$ $V A R_Φ (0)$ $V A R_φ_{1}$ Loudness $(L)$ Fluctuation strength $(F)$ Multiple regression models: $\begin{array}{l} S V_{a n n o y a n c e} \approx 0.61 Φ (0) + 0.15 V A R_Φ (0) - \\ 0.46 V A R_φ_{1} \end{array}$ $S V_{a n n o y a n c e} = 0.63 L + 0.34 F$	Coefficients:r = 0.66r = 0.13r = −0.29r = 0.66r = 0.38	– Sound energy amplitude $Φ (0)$ and loudness were the best correlated to subjective annoyance.– ACF parameters can be useful for prediction of subjective annoyance
Jeon et al.⁸	20 subjects in a listening test (age = 24–35 years)87 impact ball recordings processed to 9 sound stimuli for the test	Independent:Impact ball excitation measured and recorded in real floor structures.Dependent:V1: Just noticeable differences (JND); V2: Annoyance responses in pairwise comparisonsDescriptors used: $L_{A m a x}$ , $S P L$ , IACC	Multiple regression models:– V2 (annoyance) to $S V_{a n n o y a n c e} \approx - 0.34 (I A C C) + 0.95 (S P L)$	Individual coefficients: p < 0.01Total coefficients:r = 0.78, R² = 0.61, p < 0.01	– The JND for the $S P L$ was found 1.5 dB in terms of $L_{A m a x}$ and for the IACC levels between 0.12 and 0.13– The annoyance ratings increased as IACC decreased and $S P L$ increased; both measures contributed to the regression model significantly– Also, $S P L$ and temporal variance of IACC were found to contribute independently to annoyance
Kim et al.⁹	20 subjects (age = 20–35 years)92 impact ball recording transformed to 24 test sound samples	Independent:Impact ball test samples.Dependent:V1: Just noticeable differences (JND) of decay rate (DR)V2: Annoyance responses in pairwise comparisonsDescriptors used: $L_{A m a x}$ (SPL), DR30	Pairwise comparison test between samplesMultiple regression models:– V2 (annoyance) to $S V_{a n n o y a n c e} = - 0.02 D R + 0.18 S P L - 8.21$ Linear regression – V2 (annoyance) to $L_{A, F m a x}$ $L_{i, F m a x, c}$ (with correction)	Total coefficient:r = 0.84 (R² = 0.71, p < 0.01)r = 0.81, R² = 0.65r = 0.99, R² = 0.98	– JND of DR30 was found at a difference of 11 dB/s between test sound and reference.– SV increases when both SPL and DR increase.Longer DR30 causes higher annoyance with constant SPL.– Louder stimuli cause higher annoyance with constant DR.– Suggested correction to SPL considering the DR increases association of SPL to annoyance
Jeon and Oh¹⁰	30 subjects (age = 20–35 years)28 impact ball recorded sound stimuli	Independent:Impact ball test samples.Dependent: V1: Annoyance responses in pairwise comparisonsV2: Acceptabilitydescriptors used: $L_{A m a x}$ (SPL), DR30, DR60	Rating in 7-point scale for sound samples	Dose-response curves provided in paper No further statistical details	– No significant differences between DR30 and DR60 below 61 dB A—Some very significant above 67 dB A (p < 0.01).– Sounds evaluated with DR30 were rated as more annoying than DR60, thus above 60 dB A there is a significant effect.– The acceptable limit for impact sound in terms of LA_max was found at circa 50 dB A– A classification system for annoyance to impact sounds is developed
Ryu et al.¹¹	17 subjects in a listening test(age circa 20 years)	Independent:Impact ball and bang machine excitation measured and recorded in real floor structures.Dependent: V1: annoyance responses in pairwise comparisons	Correlations:– Part 1: V1 (annoyance) to: $L_{i, F a v g, F m a x}$ $N_{5}$ Part 2: V1 (annoyance) to: $L_{i, F a v g, F m a x}$ $N_{5}$	Coefficients:r = 0.96, p < 0.01r = 0.96, p < 0.01r = 0.84, p < 0.01r = 0.74, p < 0.01	– Arithmetic averages of octave-band $S P L$ like $L_{i, F a v g, F m a x}$ and Zwicker’s loudness percentile $N_{5}$ predict very well the subjective annoyance
Gover et al.^12,13	12 subjects in a listening test90 sound samples tested	Independent:Two impact noise types:S1: adult walking barefootS2: impact ballMeasured and recorded below of 19 types of lightweight floor-ceiling structuresDependent:V1: relative annoyance responses in pairwise comparisons	Correlations:– V1 (annoyance) to $L_{n, w}$ $L_{n, w} + C_{I, 50 - 2500}$ $L_{n, w} + C_{I, 100 - 2500}$ $L_{i, F m a x, r}$ $L_{i, F m a x, A W}$ $L_{i, F m a x (63 - 1 kHz)}$	Coefficients:S1: R² = 0.85, S2: R² = 0.89S1: R² = 0.87, S2: R² = 0.90S1: R² = 0.83, S2: R² = 0.96S1: R² = 0.70, S2: R² = 0.86S1: R² = 0.80, S2: R² = 0.93S1: R² = 0.80, S2: R² = 0.93Most differences between sound ratings statistically significant (p < 0.05)	– The standard tapping machine outcome was the best associated with the subjective ratings; it can be used adequately for subjective annoyance prediction– The rubber impact ball offered very good results as well
Späh et al.¹⁴	40 subjects in two similar listening tests	Independent:Impact sources: tapping machine, impact ball excitation measured, and human walking recorded in real and laboratory setups of floor structures.Dependent:V1: annoyance responses in a scale 0–10	Correlations:– V1 annoyance to walking: ${L^{'}}_{n T, w}$ ${L^{'}}_{n T, w} + C_{I, 50 - 2500}$ ${L^{'}}_{n, w} + C_{I, 50 - 2500}$ ${L^{'}}_{n T, H a g b e r g 03}$ ${L^{'}}_{n T, H a g b e r g 04}$ ${L^{'}}_{n T, B o d l u n d}$ – V1 annoyance to moving chair noise: ${L^{'}}_{n T, w} + C_{I, 50 - 2500}$ ${L^{'}}_{n T, B o d l u n d}$ ${L^{'}}_{n, T A 20 - 2500}$ ${L^{'}}_{n, T A 50 - 2500}$ ${L^{'}}_{n T, w} + C_{I, 50 - 2500}$	Coefficients:r = 0.62, R² = 0.38r = 0.76, R² = 0.58r = 0.78, R² = 0.61r = 0.79, R² = 0.63r = 0.79, R² = 0.62r = 0.77, R² = 0.58r = 0.85, R² = 0.72r = 0.85, R² = 0.73r = 0.91, R² = 0.82r = 0.92, R² = 0.84r = 0.91, R² = 0.82	– The tapping machine represents poorly walking annoyance– The Japanese impact ball is the most appropriate source to represent walking noise annoyance due to frequency spectrum similarities; the modified tapping machine offered slightly better associations but it is considered impractical.– Measuring down to 50 Hz helps to acquire good associations with subjective annoyance
Kylliäinen et al.^15,16	55 subjects in a listening test(25 males, 30 females, age = 25–57 years, mean = 27 years)54 sound samples tested	Independent:five sound samples of impact noise types:S1: walking with hard shoes;S2: walking with socks;S3: walking with soft shoes;S4: bouncing ball;S5: moving chair;filtered through nine types of floor impact SRI spectra.Dependent:V1: loudness responsesV2: annoyance responses	Regression models:– V1 (loudness) to ${L^{'}}_{n, w} + C_{I, 50 - 2500}$ ${L^{'}}_{n, w} + C_{I}$ V2 (annoyance) to ${L^{'}}_{n, w} + C_{I, 50 - 2500}$	Coefficients:R² = 0.56 (S1)R² = 0.37 (S3)R² = 0.53 (S5)R² = 0.57 (S1)R² = 0.39 (S3)R² = 0.50 (S5)R² = 0.49 (S1)R² = 0.31 (S3)R² = 0.47 (S5)All R² > 0.12 statistically significant (p < 0.01)	– ${L^{'}}_{n, w} + C_{I}$ , ${L^{'}}_{n, w} + C_{I, 50 - 2500}$ , ${L^{'}}_{n, w, F a s},$ ${L^{'}}_{n, w, F a s, 50}$ , ${L^{'}}_{n, w, G e r}$ , and ${L^{'}}_{n, w, B o d}$ were found to be the best indicators for both subjective loudness and annoyance– Low frequencies 50–100 Hz inclusion in the SNQs offers better correlation to the subjective responses
Öqvist et al.¹⁷	24 subjects in a listening test (12 males, 12 females, age mean = 27 years, SD = 5 years)4 sound samples tested	Independent:two sound samples of impact noise types:walking with socks,walking with hard shoes,recorded under two floors:a lightweight (LW)a concrete heavyweight (HW)Dependent:V1: annoyance responses	Pairwise comparison test between samples	No details provided	– Annoyance perception was significantly higher for the lightweight floor case– 20 Hz was indicated as the limit for perceived annoyance, as an important limit to evaluate walking with socks and impact sounds in LW

SD: standard deviation; SPL: sound pressure level; SRI: Sound Reduction Index; SNQ: single number quantities.

Results: associations of impact sound acoustic data with self-reported responses in laboratory tests

In Yeon,² a laboratory listening test with 20 participants (aged 21–31 years) to investigate the differences in perception of impact noise sounds was recorded in apartments. The standardized sources were a bang machine (tire) and a tapping machine. The subjects listened to the samples and had to adjust them to pink noise levels according to their perception of loudness and noisiness. First, the results of loudness and noisiness matching were highly and significantly correlated (r = 0.916, p < 0.01). The subjects raised the pink noise 2–3 dB higher to match the levels of the bang machine, while they lowered the pink noise 3–4 dB to match the tapping machine sound: subjects perceived bang machine 6–7 dB noisier and louder than the tapping machine as the author comments. Also, parameter values of the autocorrelation function (ACF) and the interaural cross-correlation function (IACC) were analyzed for both sources. The maximum amplitude $Φ$ (0) of the ACF is reported as highly correlated to perceived noisiness of the tapping machine noise. The author argues that perceived loudness and noisiness can be explained by the ACF and directivity of peaks by the IACC. However, this is not supported by any statistical testing, as only correlations among acoustic parameters are presented.

In Yeon and Jeong,³ a continuation of the previous study is presented as the evaluation of loudness. A typical concrete floor structure in a Korean residential building and nine different configurations with treatments of that structure were measured according to JIS A 1418.¹⁸ Recordings were made for the impact excitation sources: tapping, bang machine, rubber impact ball, and human jumping. A listening test with 30 subjects (27 males, 3 females, aged 24–41 years) was conducted where the test samples were evaluated in a pair of comparison test (108 comparisons) using a 5-point scale (−1, −0.5, 0, 0.5, and 1). Subjective responses of loudness were highly correlated with the maximum ACF amplitude $Φ$ (0) of tapping noise (r = 0.96, p < 0.01), bang machine noise (r = 0.94, p < 0.01), and impact ball noise (r = 0.94, p < 0.01). The same applied to the subjective loudness responses associated with Zwicker’s parameters: Loudness $(L)$ and unbiased annoyance $(U A)$ , which are psychoacoustic metrics defined in literatures.^18,19 Specifically, the loudness responses correlated highly with $L$ for tapping noise (r = 0.94, p < 0.01), sufficiently for bang machine noise (r = 0.74, p < 0.01), and highly for impact ball noise (r = 0.94, p < 0.01). They also correlated highly with $U A$ for tapping noise (r = 0.92, p < 0.01) and sufficiently for bang noise (r = 0.72, p < 0.01) and rubber ball noise (r = 0.76, p < 0.01). Also, $L$ values were highly correlated to $U A$ . The authors highlight that Zwicker parameters are more reasonable for the tapping machine noise; for the bang machine and impact ball cases, maximum amplitude $Φ$ (0) was associated with the loudness perception more than other parameters.

Furthermore, a multiple regression analysis was done which resulted in the following optimal models for the loudness perception, denoted as $S V$ , all with statistical significance (p < 0.05). For the tapping machine case, the model was SV_tapping = −17.761 + 0.065 Φ(0) + 11.51 τ₁ − 1.45ϕ₁, where $τ_{1}$ and $φ_{1}$ are parameters (for time and amplitude, respectively, at 1 ms) of the ACF. The total correlation coefficient of the model was r = 0.94. Another model was acquired using all the examined Zwicker parameters: SV_tapping = −5.731 + 0.25 L + 2.23 FS + 1.16 T – 0.0076 UA, where $F S$ and $T$ denote Fluctuation Strength¹⁹ and Tonality,¹⁹ respectively, with total $r$ = 0.98. The authors highlight that pitch and energy changes are important parameters.

For the case of bang machine noise, the derived models were SV_Bang = −3.691 + 0.147 Φ(0) − 0.251 τ_e −3.83 $W_{I A C C}$ (r = 0.96) and SV_Bang = −0.534 + 0.22 $L$ (r = 0.74). The term $τ_{e}$ denotes the effective duration of the envelope of normalized ACF and $W_{I A C C}$ is the width of IACC at time $τ_{I A C C}$ (inter-aural delay time), see details in Yeon and Jeong.³ Finally, for the impact ball noise, the models were SV_Ball = −4.754 + 0.121 Φ(0) − 0.2021 τ_e − 1.01 IACC + 0.992τ_IABC (r = 0.98) and SV_Ball = −1.431 + 0.177 L + 0.24 FS – 0.0012 UA (r = 0.95). Thus, the authors highlight the spatial factors and sound energy as important parameters for the sources: impact ball and bang machine.

Similar studies regarding floor impact sound and self-reported loudness and annoyance were continued in study.⁴ Eight floors in different apartments (same floorplan) of an unoccupied multi-story building in Seoul were measured following the standard JIS A 1418. Different configurations in the sending rooms including insulation for the floor, walls, and ceiling were tested with two different impactors, the tapping machine and the tire machine measured in $L_{A e q}$ and $L_{A m a x}$ , respectively. Sound recordings using a dummy head were taken as well, which were used in an auditory test with 60 participants (30 Korean and 30 German). The sound samples of the floor setups were tested in pairs, always using a floor structure with no additional insulation as a reference sound to be compared with the other seven floor types. Overall, 56 sound stimuli were tested in the listening experiment, and 28 initial pair of sounds were tested twice and in random orders. The participants had to rate loudness (in a scale from −2 to 2, 0 means equal loudness between stimuli) and annoyance (scale 1–9) for each pair of stimuli.

Lower levels of subjective loudness and annoyance were reported for the cases of sending rooms with insulated floor and walls or the same setup with an extra suspended ceiling insulation. These conclusions were made for both cases of impact noise sources. In this study, the parameter of different culture is featured as well. A comparison of impact noise level differences and subjective data offered determination coefficient values $R^{2}$ equal to 0.55 for the Korean and 0.86 for the German subjects; the results for Koreans are not so consistent due to higher impact noise sensitivity according to the authors. Also, correlation coefficients, dependent on frequency (1/3 octave bands) are presented for the data comparison; good correlations were found in general above 250 Hz for the tapping machine around 0.8 for loudness (R² = 0.64) and 0.6 for annoyance (R² = 0.36) and above 63 Hz for the tire machine with values between 0.7 and 0.9 (R²: 0.49–0.81) for loudness and 0.5 and 0.7 for annoyance (R²: 0.25–0.49). It is highlighted that the tire machine spectrum has dominant frequencies below 250 Hz and that could be reduced on thicker concrete slabs.

A continuation of the same study in Korea is presented in Jeon et al.⁵ Further measurements in the test building and floor structure configurations were conducted for another comparison of two impact sources: impact ball and tapping machine. A total of 30 students took part in a similar listening experiment rating 48 sound samples in the same loudness scale (–2 to 2). Several descriptors were tested for the association with subjective ratings. For the impact ball case, the results were sufficient with coefficients acquired by $L_{A m a x}$ , Zwicker’s Loudness $L$ , and Percentile Loudness $N_{10}$ offering $R^{2}$ of 0.70, 0.74, and 0.77, respectively. However, $L_{A m a x}$ is still suggested as a practical descriptor since the authors highlight that Zwicker’s parameters are not easy to determine due to instrumentation and calculations. For the tapping machine, the results were very good with $R^{2}$ of 0.84, 0.84, and 0.88 for mean sound pressure level (SPL) averaged for all measured structures $L$ and $N_{10}$ , respectively.

Two additional listening experiments were conducted in this study⁵ with few details provided:

An on-site auditory experiment with 98 subjects in a living room of the test building to rate annoyance (scale 1–9) to impact ball sounds dropped from various heights. Three categories were suggested for classification using this scale: “Audibility” (1–3), “Disturbance” (4–6), and “Amenity” (7–9). The level of $L_{i, F m a x, A W} = 54 dB$ corresponded to a level of annoyance of 4 in the rating scale. Three classes were suggested according to the $L_{i, F m a x, A W}$ levels: Class 1 (<44 dB), Class 2 (<49 dB), and Class 3 (<54 dB).

A listening test with 10 students was conducted to investigate the just noticeable differences (JND) for the perception of impact ball noise in SPL. The JND level was recognized at about 2 dB for both the tapping machine and the impact ball cases, as indicated by 86% and 89% of the participants in each case, respectively.

In a further study in South Korea,⁶ impact ball sounds were again recorded in 35 different typical apartments (100–120 m²), which were box-frame-type reinforced concrete constructions with slab thickness 150–180 mm. They were clustered in three groups based on their frequency characteristics and they were then used for two auditory experiments with 40 participants (28 males, 12 females, age span 24–35 years). The first experiment concerned successful indicators of perceived annoyance; 87 impact ball sound samples ( $S P L$ between 38 and 64 dB, divided in three groups) were evaluated in pair comparisons. The sound quality (SQ) metrics reported and used for the assessment were $L_{i, F m a x, A W}$ , $L_{- n u m b e r}$ , $L_{A e q}$ , $L_{A m a x}$ , $L L_{Z}$ (Zwicker’s loudness level), $N_{m a x}$ , (maximum loudness), $N_{5}$ , and $L_{m, 1 / 1 (63 - 500 Hz)}$ . They all showed good correlations with annoyance, especially $L_{A m a x}$ , $L_{A e q}$ , $L L_{Z}$ , $N_{m a x}$ , and $L_{m, 1 / 1 (63 - 500 Hz)}$ , which were concluded to be good descriptors of subjective annoyance, with reported correlation coefficients higher than 0.88 for all impact ball groups. Zwicker’s loudness $L L_{Z}$ showed the highest correlation r = 0.97 (R² = 0.94, p < 0.05), $L_{A m a x}$ was sufficient with r = 0.92 (R² = 0.85, p < 0.01), while the lowest coefficient was r = 0.88 (R² = 0.77, p < 0.01) for $L_{i, F m a x, A W}$ . The authors emphasize on the importance of loudness level $L L_{Z}$ for predicting the annoyance response and $L_{A m a x}$ is suggested as the most practical descriptor, due to easy measuring with a sound level meter.

In a second test, 36 stimuli sounds were evaluated by the same participants in pair comparisons to explore the effects of the psychoacoustic metrics as variables: loudness $(L)$ , sharpness $(S)$ , roughness $(R)$ , and fluctuation strength $(F)$ on the annoyance. In a regression analysis, loudness and fluctuation strength were found to be highly correlated with subjective responses, with coefficients r = 0.81 (p < 0.01) and r = 0.90 (p < 0.05), respectively, in the individual linear models. A multiple regression model for the subjective variable annoyance was chosen, using the best combination of metrics as $S V_{a n n o y a n c e} = 0.77 L + 0.15 F$ , with a total coefficient r = 0.90 (R² = 0.81, p < 0.05). Thus, the authors highlight that except the main effect of loudness, temporal variations in low frequencies play a role as well in the annoyance perception.

In addition, a semantic differential test took place for a set of 12 adjective pairs for evaluating floor impact sound after a selection process. The same 40 people participated and used a bipolar scale (with an adjective and its opposite) to characterize the given sound stimuli. Their responses were processed using the method of factor analysis, revealing three dimension groups, entitled by the authors as “1: reverberance and spaciousness,” “2: dullness,” and “3: loudness.” The first dimension was well correlated with roughness (r = 0.69, R² = 0.48, p < 0.05), the second with fluctuation strength (r = 0.71, R² = 0.50, p < 0.05), as well as the third (r = 0.73, R² = 0.53, p < 0.01), which was also associated with loudness with r = 0.75 (R² = 0.56, p < 0.05). The authors conclude that several frequency characteristics can be described by those three reported categories.

In Jeon and Sato,⁷ the annoyance of floor impact sounds was evaluated using the ACF and SQ metrics. Two impact sources were used, the bang machine and the impact ball for measurements in six apartments with different insulation configurations. Binaural recordings were taken also with a dummy head to create 28 pairs of sound stimuli for a pairwise comparison. The stimuli were classified into three groups according to their spectral behavior. Then, 40 subjects (20 students and 20 housewives) took part in a laboratory listening test; 35 of them distinguished various levels of annoyance (p < 0.05) and the agreement among all responses was significant (p < 0.05).

Single and multiple regression analyses were performed. Three ACF parameters were selected for a regression model: $S V_{a n n o y a n c e} \approx 0.61 Φ (0) + 0.15 V A R_Φ (0) - 0.46 V A R_φ_{1}$ . $Φ (0)$ stands for the maximum amplitude of sound energy, $φ_{1}$ is the maximum ACF amplitude, and $V A R$ denotes the variance of the parameters. The correlation coefficients between annoyance responses and the chosen parameters were 0.66 for $Φ (0)$ , 0.13 for $V A R_Φ (0)$ , and −0.29 for $V A R_φ_{1}$ . Regarding SQ parameters, loudness $(L)$ and fluctuation strength $(F S)$ correlated best with subjective annoyance and provided $r$ values 0.66 and 0.38, respectively. They were selected for the model $S V_{a n n o y a n c e} = 0.63 L + 0.34 F S$ , which is different from the model presented before in Hongisto et al.²⁶ However, the total coefficients for the above models are not reported. Overall, $Φ (0)$ and loudness were the most correlated from the studied parameters. It is highlighted that the variance of $Φ (0)$ and $φ_{1}$ can play a role in annoyance prediction since they are related to the pitch (tonality) of the noise signal. Floor structures with higher resonance frequencies had lower sound levels from the heavy impact sources. Floors with viscoelastic damping materials had reduced impact sound levels and thus corresponded to lower annoyance ratings. However, structures with resilient isolators (floating floor types) did not offer reduced annoyance in all cases, as it might be expected. Jeon and Sato⁷ state that this happens because “isolators amplify low-frequency noises (below 100 Hz) generally produced by heavyweight impacts.”

In Jeon et al.,⁸ the interaural cross-correlation (IACC) function was used in the evaluation of floor impact annoyance. Impact ball measurements inside Korean apartments and 87 binaural recordings took place: they were used in a laboratory listening test with 20 participants (aged 24–35 years). In the first part, random pairs of stimuli were presented to the subjects who were asked to choose the stronger sound. The JND of the $L_{A m a x}$ levels (manipulated SPL) and IACC levels were explored. The JND value was acquired when 75% of the subjects could distinguish between a test sample and the reference with different measures of $L_{A m a x}$ and IACC values. Overall, the JND for the SPL was found 1.5 dB differences of $L_{A m a x}$ and for the IACC levels between 0.12 and 0.13.

Then, nine of the stimuli were chosen for the second part where the subjects rated relative annoyance in pair comparisons again. The effects of $S P L$ and IACC were found statistically significant in the ANOVA (p < 0.01) but not their interactions. Then, a regression model was determined as $S V_{a n n o y a n c e} \approx - 0.34 (I A C C) + 0.95 (S P L)$ , with statistically significant individual coefficients (p < 0.01) and total correlation coefficient (r = 0.78, p < 0.01). The annoyance ratings increased as IACC decreased and SPL increased. SPL and IACC contributed to the regression model by 79.3% and 20.4%, respectively. The temporal variations of IACC (T.var_IACC) were explored as well in association with subjective annoyance; the subjects offered consistently and significantly (p < 0.05) agreed that $S P L$ and T.var_IACC contribute independently to annoyance at 94.2% and 2.7%, respectively (p < 0.01). Also, it was concluded that for the floor structures with damping materials, the IACC values are greater than floors with resilient isolator: there is better energy absorption and less sidewall transmission with damping layers in floors.

A continuation study of Jeon et al.⁸ is presented in Kim et al.⁹ that deals with the temporal decay of impact sounds and how that affects subjective perception. For that investigation, the JND of decay rate (DR) was used for impact ball sound samples. The test samples were created after processing of 92 field recordings in apartments of concrete box-framed buildings; they were classified in three spectrum groups according to Jeon and Sato.⁷ In addition, the authors mention that the effects of floor and room conditions on the recordings were investigated with ANOVA and found statistically significant, specifically factors such as floor thickness, area, room volume, and type. However, no details are provided for those variables. The metric DR is similar to reverberation time (RT) and corresponds to the decay of a signal (normalized to 0 dB): for example, from −5 to −35 dB for DR30. The subjects rated the sounds using pairwise comparisons in a laboratory listening test. If more than 75% of the subjects distinguish the reference sound and the test sample, JND is valid according to this study.

In the first test, 15 test sound stimuli were judged by the participants if they sound similar; the JND was determined when the DR difference of the stimuli was 11 dB/s between test sounds and reference. That means the subjects started to decide that the tested stimuli were different sounds when their actual difference in DR was more than 11 dB/s (slope of 11 dB drop per second). In the second test, the participants offered annoyance ratings of nine test sounds. It was found that the annoyance values increase when both $L_{A E}$ (SPL) and DR increase. Also, longer decays (DR) cause higher annoyance when $L_{A E}$ is constant; when DR is constant, then louder stimuli cause higher annoyance. ANOVA results indicated that the interactions of the factors SPL, DR, and spectrum group were not statistically significant. A multiple regression model was developed as $S V_{a n n o y a n c e} = - 0.02 D R + 0.18 S P L - 8.21$ with reported total coefficient R = 0.84 (R² = 0.71, p < 0.01). The contribution of the modeling factors was 23% for DR and 76% for SPL. Therefore, a correction for the A-weighted maximum level $L_{A, F m a x}$ rating index is proposed considering the effect for DR as $L_{i, F m a x, c} = L_{i, F m a x} - a (D R_{i} / D R_{m e a n})$ . When the latter correction was applied, the linear regression results were drastically improved: the subjective annoyance was associated with $L_{A, F m a x}$ with R² = 0.98 instead of R² = 0.65 (without correction).

In another continuation study,¹⁰ the classification of annoyance and acceptability of SPL and temporal decay levels (DR) was explored. Similar sound stimuli as in Jeon et al.⁸ were used and 30 subjects in a listening test rated their annoyance in a 7-point-scale and acceptability (yes/no). Both DR30 and DR60 were tested for decays of 30 and 60 dB, respectively. No significant differences were reported between DR30 and DR60 below impact level differences of $L_{A m a x} = 61 dB A$ , but significant differences were found above $L_{A m a x} = 67 dB A$ (p < 0.01). Sounds evaluated with DR30 were rated as more annoying than DR60, indicating that above a DR slope of 60dB A/s, there is a significant effect of temporal decay on annoyance. Dose-response curves for the percentage of annoyed subjects relevant to SPL $(L_{A m a x})$ and DR30 and DR60 are presented, but no statistical details given. A classification system for annoyance from impact sound was developed with four classes based on the percentage of annoyed subjects (%A who rated “3—Moderately” and higher). Class A includes the upper quantile 0%A–25%A of annoyed subjects (criteria in $L_{A m a x}$ for cases of DRs: DR30 < 44.5 dB A or DR60 < 45.4 dB A), and then, other classes were defined as Class B (25%A–50%A, DR30 < 49.2 dB A or DR60 < 50 dB A), Class C (50%A–75%A, DR30 < 53.8 dB A or DR60 < 54.5 dB A), and Class D (75%A–100%A, DR30 > 53.8 dB A or DR60 > 54.5 dB A). A similar classification system is proposed for the percentage of highly annoyed subjects (%HA who rated “4—Considerably” and higher). The acceptable limit for impact sound in terms of $L_{A m a x}$ was found at circa 50 dB A, which corresponds Class to A and B (%A) from the developed system, thus the authors consider it as reliable.

In Ryu et al.,¹¹ a study for the relation between subjective annoyance and single number quantities (SNQs) for impact sounds in wooden buildings in Japan is reported. Excitation by bang machine and impact ball was used for measurements and mono-aural recordings on 26 floors of 12 real buildings; 2 typical spectra were defined for the study, SP-1 and SP-2 to be used as reference, and another 11 stimuli for each typical spectrum were created with manipulation of the frequency responses. In all, 17 subjects (aged circa 20 years) took part in a listening experiment where they rated the 24 sound stimuli in a pair comparison test (55 pairs) using a relative annoyance scale from −3 to 3 (0 for equal annoyance between stimuli). The various impact sound levels (with different types of weighting) were defined in the Japanese standard: $L_{i, F m a x, r}$ , $L_{i, F m a x, A W}$ , $L_{i, F m a x}$ , and $L_{i, F a v g, F m a x}$ were assessed for the sound stimuli. Loudness $(N_{5})$ was used too. They were all very well correlated to annoyance with $r$ values ranging from 0.89 to 0.99: the best correlations were equal to 0.99 for $N_{5}$ and 0.96 for $L_{i, F a v g, F m a x}$ in both cases of SP-1 and SP-2. All responses were found to be significantly different (p < 0.01).

A second experiment took part in the same study with 31 subjects (aged circa 20 years) where impact sounds dependent on the $S P L$ were compared to a reference sound (SP-2). Two separate levels of 55 and 65 dB in $L_{i, F m a x, r}$ (denoted L55 and L65) were used for the compared stimuli in pair comparisons using the same methodology as before. The associations between SNQs and relative annoyance differed a lot; correlation coefficients varied from 0.39 to 0.93, while bigger associations with annoyance were found for the L55 stimuli. Most results were statistically significant (p < 0.01) especially for the L65; $r$ values ranging from 0.81 for $L_{i, F a v g, F m a x}$ to 0.91 for $N_{5}$ , while the same values for L55 stimuli were from 0.84 for $L_{i, F a v g, F m a x}$ to 0.74 for $N_{5}$ . It is concluded that arithmetic averages of octave-band $S P L$ like $L_{i, F a v g, F m a x}$ and Zwicker’s loudness percentile $N_{5}$ describe well the subjective annoyance and can be used as sufficient SNQ, but $N_{5}$ is characterized difficult to calculate, as also in Lee et al.⁶

A wide research study took place in the National Research Council of Canada in Ottawa,^12,13 specifically for the ranking of lightweight (LW) wood framed floor-ceiling structures based on the subjective response of participants. First, a wide set of 19 various bare floor assemblies was measured in laboratory conditions (two vertically adjacent reverberation rooms with a specimen opening). All standardized excitation sources were used; the standardized tapping machine, the modified tapping machine (i.e. the standardized one on a resilient layer), the heavy/soft rubber ball dropped from the heights of 10, 50, and 100 cm, and additionally the tire machine was used as well. Sound recordings were taken for the rubber ball cases and additionally with a human source: an adult walking barefoot on the test floors. A total of 90 samples were used in a pairwise comparison test; 12 participants took part in the laboratory test and rated the sounds in a relative annoyance scale from 1 to 9 (1—‘Sound 2 much less annoying,’ 5—‘equally annoying sounds,’ and 9—‘Sound 2 much more annoying’). Sound 1 was always the same reference and Sound 2 was the tested sample.

A correlation analysis was performed to investigate the relationship between the subjective annoyance and the acoustic data collected in the measurements. The highest association was reported between the annoyance levels and the metrics derived with the standard tapping machine; $L_{n, w}$ , $L_{n, w} + C_{I, 50 - 2500}$ , and $L_{n, w} + C_{I, 100 - 2500}$ . The determination coefficients $R^{2}$ were equal to 0.85, 0.87, and 0.83, respectively, for the walking sounds case and 0.89, 0.90, and 0.96, respectively, for rubber ball impact noise. The relevant results for the measurements with the hard/soft impact ball (according to JIS A 1418 and KS F 2863) and the metrics $L_{i, F m a x, r}$ , $L_{i, F m a x, A W}$ , and $L_{i, F m a x (63 - 1 kHz)}$ were also satisfactory with $R^{2}$ values 0.70, 0.80, and 0.80, respectively, for walking and 0.86, 0.93, and 0.93, respectively, for rubber ball annoyance. The tire machine outcome was the worst, while the modified tapping machine outcome was sufficiently associated with $R^{2}$ values ranging from 0.71 to 0.84. Summing up, according to this study, the use of the standard tapping machine is adequate for predicting the subjective annoyance, without using any other sources. The use of rubber ball is also a good choice since it has shown correlations with subjective annoyance. However, that conclusions were derived using a small group of 12 participants only for the test.

In Späh et al.,¹⁴ the European research program AcuWood is presented, which concerns impact noise annoyance in wooden buildings. Measurements of timber floor structures and binaural recordings took place in real buildings and in laboratories following the same methods. Different coverings on the floors were tested during laboratory measurements too. Several impact sources were explored: the standardized tapping machine and the modified one (according to ISO 10140-5), the Japanese impact ball, and “real” impact sources (male walkers with socks and shoes and a female walker with hard heeled shoes and a chair which was drawn). Two separate listening tests took place using the stimuli created from all floors, while a field measurement was common in both tests as a reference. The tests involved 18 and 22 subjects, which provided ratings of annoyance (scale 0–10) according to ISO 15666.²⁷

The results indicate that the typically used ${L^{'}}_{n T, w}$ (range = 100–3150 Hz) was poorly associated with the annoyance due to walking (r = 0.62, R² = 0.38) but with using the lower frequency range and the adaptation term, the result becomes better for ${L^{'}}_{n T, w} + C_{I, 50 - 2500}$ (r = 0.76, R² = 0.58). Different rating curves proposed for evaluation of the ISO 717-2²⁰ method for assessment of impact noise levels with the tapping machine were tested; the best associations between walking noise annoyance and impact noise descriptors were found for ${L^{'}}_{n, w} + C_{I, 50 - 2500}$ (r = 0.78, R² = 0.61), ${L^{'}}_{n T, H a g b e r g 03}$ (r = 0.79, R² = 0.63), ${L^{'}}_{n T, H a g b e r g 04}$ (r = 0.79, R² = 0.62) and ${L^{'}}_{n T, B o d l u n d}$ (r = 0.77, R² = 0.58). The last three descriptors are variations of ${L^{'}}_{n, w} + C_{I}$ , with correction spectra $C_{I}$ differentiated from the standardized ones: they were acquired from past field research and tested again in the laboratory.¹⁴ For the moving chair annoyance, the best associations with the descriptors for tapping machine measurements were found for ${L^{'}}_{n T, w} + C_{I, 50 - 2500}$ (r = 0.85, R² = 0.72) and ${L^{'}}_{n T, B o d l u n d}$ (r = 0.85, R² = 0.73). The modified machine descriptors offered better results for ${L^{'}}_{n, T A 20 - 2500}$ (r = 0.91, R² = 0.82) and the best for ${L^{'}}_{n, T A 50 - 2500}$ (r = 0.92, R² = 0.84). The impact ball descriptor relates very well to moving chair annoyance: ${L^{'}}_{n T, w} + C_{I, 50 - 2500}$ (r = 0.91, R² = 0.82) as well. It is concluded that the Japanese impact ball is the most appropriate source to represent walking noise annoyance due to frequency spectrum similarities; the modified tapping machine offered slightly better associations but it is considered impractical. The need of measuring down to 50 Hz to acquire good associations with subjective annoyance is highlighted.

Another study in Finland took place^15,16 exploring the associations of descriptors derived from impact sound on concrete floors and subjective annoyance; the relation of eight impact noise descriptors to subjective ratings was studied. A listening test was conducted with 55 subjects (25 males and 30 females, age 25–57 years, mean 27 years) who offered their ratings on a set of five recorded impact sounds through nine floor configurations in a psychoacoustic listening experiment at the Finish Institute of Occupational Health. A floor construction was measured in a laboratory, being bare concrete or with eight different floor covering types, according to ISO 140-7. The eight SNQs explored were ${L^{'}}_{n, w}$ , ${L^{'}}_{n, w} + C_{I}$ , ${L^{'}}_{n, w} + C_{I, 50 - 2500}$ (according to ISO 717-2²²), ${L^{'}}_{n, w, F a s}$ , ${L^{'}}_{n, w, F a s, 50},$ ${L^{'}}_{n, w, G e r}$ , ${L^{'}}_{n, w, B o d}$ , and ${L^{'}}_{n, w, H a g}$ . The last five descriptors are variations of ${L^{'}}_{n, w} + C_{I}$ , with correction spectra $C_{I}$ differentiated from the standardized ones: they were acquired from past field research and tested again in the laboratory.^15,16 The recorded sound types were walking with hard shoes, socks and soft shoes, a bouncing ball, and a moving chair. The participants were asked to rate the sound samples in terms of perceived loudness and annoyance in a scale of 0–10 (0—‘Not audible,’ 1—‘Not at all …’ and 10—‘Extremely …’) and also in terms of acceptability in a scale of 0–3.

For three sound types S1, S3, and S5 (walking with hard shoes, soft shoes, and moving chair), the correlations were considered sufficient and statistically significant (p < 0.01) for most SNQs, with determination coefficient $R^{2}$ values ranging from 0.25 to 0.60. Overall, ${L^{'}}_{n, w} + C_{I, 50 - 2500}$ is proposed as the most suitable indicator for S1, S3, and S5 having good associations with both loudness (reported $R^{2}$ values 0.56, 0.37, and 0.53, respectively) and annoyance ( $R^{2}$ values 0.49, 0.31, and 0.47, respectively). This is in agreement with the results presented in Rychtáriková et al.²⁰ The other standardized descriptor ${L^{'}}_{n, w} + C_{I}$ is considered good for perceived loudness prediction as well with reported $R^{2}$ values 0.57, 0.39, and 0.50, respectively, for S, S3, and S5.

For the other sound types (S2: walking with socks and S4: bouncing ball), the associations were weak with $R^{2}$ ranging from 0.03 to 0.16. The metrics ${L^{'}}_{n, w} + C_{I}$ , ${L^{'}}_{n, w} + C_{I, 50 - 2500}$ , ${L^{'}}_{n, w, F a s}$ , ${L^{'}}_{n, w, F a s, 50},$ ${L^{'}}_{n, w, G e r}$ , and ${L^{'}}_{n, w, B o d}$ were found to be the best indicators for both subjective loudness and annoyance. For the acceptability, it is only reported that the determination coefficients are similar to the ones acquired for the loudness and annoyance perception cases. It is concluded that inclusion of low frequencies 50–100 Hz in the SNQs offers better correlation between a SNQ and the subjective responses. They summarize also that more SNQs should be developed to represent all types of typical impact noise sounds in buildings and their spectra.

In Öqvist et al.,¹⁷ a study is presented where the authors investigate the effect of the frequency range 20–50 Hz in the perception of walking sound annoyance. A listening experiment with 24 Swedish subjects (12 males and 12 females, age mean 27 years, standard deviation (SD) = 5 years) took place, where walking sound samples were evaluated. The latter concerned recordings of a male walker with socks or shoes through two construction cases: a wooden LW and a concrete heavyweight (HW). They were tested in a pairwise comparison test which showed that the percentage of subjects perceiving a difference in annoyance was significantly higher for the LW floor case; 20 Hz was indicated as the limit for perceived annoyance and as an important limit to evaluate walking with socks. It is highlighted that existing impact sound SNQs are not sufficient in terms of correlation to subjective responses. It was confirmed that frequencies down to 20 Hz are necessary to evaluate impact sounds in LW, while 40 Hz was the lowest limit for walking with socks in HW and 100 Hz for shoes in HW. In addition, the highest correlation between annoyance responses and standardized descriptors is reported for ${L^{'}}_{n T, w} + C_{I, 20 - 2500}$ and ${L^{'}}_{n T, w} + C_{I, 25 - 2500}$ , so they are considered the optimized SNQs in this study. This is in agreement with the previous findings as in Gover et al.^12,13 However, statistical details for correlation and significance are not reported.

Discussion

In the presented studies, various descriptors have been used to associate to self-reported responses, mostly for annoyance or loudness. However, the lack of a proper SNQ that could work efficiently for all types of impact noise is apparent or directly concluded in many studies.^14–16

The inclusion of low frequencies (down to 50 Hz) seems to be an important concern. Many of the reviewed studies indicate that extended frequency spectra which include low frequencies down to 50 Hz correlate better with subjective responses of annoyance.^14–17 Variations exist as well regarding several types of impact sources tested in different studies, but the overall associations of subjective responses to impact sound are sufficiently good and become better with extended spectra. That is a general issue discussed in the field of building acoustics.^1,14

The indicators for the standardized tapping machine seem to predict well the overall subjective noise annoyance in many studies,^4,5,12–16 but do not associate well enough with walking noise.^12,13 The Japanese impact ball seems to represent better impact sounds induced by human walking as demonstrated in many Korean and Japanese studies;^5–11 it is summarized that impact ball as an impact source corresponds better to the usual impact noise spectra found in residential multistory buildings, especially human walking and kids jumping. It is also noticeable that Korean researchers differentiate between HW (impact ball and bang machine) and LW impact sounds (tapping machine) in their publications.

In some studies, both loudness and annoyance ratings were included for the self-reported assessment of the participants,^4,16 and loudness scale only was used in one study.⁵ Some similar results have been between loudness and annoyance ratings,¹⁶ but overall no final conclusion has been done on the differences and similarities for the case of impact sound perception related to loudness or annoyance.

In some studies, SQ metrics are examined for the subjective annoyance assessment.^3,5–7 In Lee et al.,⁶ the authors highlight the significance of Zwicker loudness level, $L L_{Z}$ , for predicting annoyance response. Some of the studies focus on the effects of ACF and IACC. Few studies focus on the effect of SQ metrics only² or their combination to ACF/IACC.^3,7 Some studies explore specifically the effect of temporal decay with DR.^9,10 In overall, they all conclude that temporal characteristics are important for the prediction of self-reported annoyance in literatures.^2,3,7–10 In many cases, the parameter of maximum amplitude $Φ$ (0) was highlighted as significant.^2,3,6 Furthermore, additional properties of sound signals such as modulation and fluctuation were mentioned as important.^3,7,8

Many multiple regression models have been presented for the prediction of self-reported annoyance.^3,6,7 The most successful regression models are presented in Yeon and Jeong,³ and they both have total correlation coefficient r = 0.98 (p < 0.05) and concern annoyance prediction based on acoustic measurements from the following:

Tapping machine data: SV_tapping = −5.731 + 0.25 L + 2.23 FS + 1.16 T − 0.0076 UA.

Impact ball: SV_Ball = −4.754 + 0.121 Φ(0) − 0.2021 τ_e − 1.01 IACC + 0.992 τ_IACC.

The variability of impact noise sensitivity due to different culture is featured in only one study,⁴ where subjects from Germany and Korea took part in the presented experiment. A big difference was revealed; therefore, intercultural responses to impact noise might be an interesting issue for further studies.

Classification took place in two studies only. In Jeon et al.,⁵ 98 subjects evaluated impact ball noise and the following three categories were proposed using an annoyance scale from 1 to 9: “Audibility” (1–3), “Disturbance” (4–6), and “Amenity” (7–9). In addition, in Jeon and Oh,¹⁰ four classes were developed based on self-reported annoyance percentages (Class A–B, %A), and minimum SPL levels of the DR for every class were defined.

Most of the studies have a good level of presentation and evaluation of research evidence as can be seen in Table 2. Many statistical evaluations took place; some were incomplete with missing important parameters or some details were not reported at all.^2,7 In some listening tests, very small amounts of subject have participated.^11–13 This fact weakens the strength of association, the consistency, the biological gradient, and the analogy of the acquired results, as demonstrated also in Table 2.

Conclusion

This review shows that annoyance perception due to impact sound is an important issue and can be associated well in overall to acoustic measurements. Many standardized SNQs and alternative descriptors have been evaluated and associate well with subjective responses collected in laboratory listening tests. The standardized descriptors based on the tapping machine measurements are considered sufficient, but the highest correlations have been found between SQ metrics and subjective ratings. Inclusion of low frequencies down to 50 Hz in measurements helps impact sound descriptors to relate better to subjective responses. Furthermore, all descriptors do not relate well to all kinds of impact sound related. The impact sources suggested as efficient are the standardized tapping machine for overall annoyance, the Japanese impact ball for human walking annoyance, or typical impact sounds in dwellings. Additional properties of noise signal such as modulation, decay, and other temporal characteristics evaluated by the ACF, the IACC, the DR, or SQ metrics are indicated to play an important role in annoyance rating and perception.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support of the research, authorship, and/or publication of this article: This article was written within the research project “Acoustic Comfort in Building Apartments” funded by Saint-Gobain Weber.

References

Rasmussen

Rindel

JH.

Concepts for evaluation of sound insulation of dwellings—from chaos to consensus? In: Proceedings of the Forum Acusticum 2005, Budapest, 2005, http://orbit.dtu.dk/files/2460104/oersted-dtu1603.pdf

Yeon

JY.

Subjective evaluation of floor impact noise based on the model of ACF/IACF. J Sound Vib 2001; 241(1): 147–155.

Yeon

Jeong

JH.

Objective and subjective evaluation of floor impact noise. J Temp Des Architect Environ 2002; 2(1): 20–28.

Jeon

Jeong

Vorländer

et al . Evaluation of floor impact sound insulation in reinforced concrete buildings. Acta Acust United Ac 2004; 90: 313–318.

Jeon

Ryu

Jeong

et al . Review of impact ball in evaluating floor impact sound. Acta Acust United Ac 2006; 92: 777–786.

Lee

Kim

Jeon

JY.

Psychoacoustical characteristics of impact ball sounds on concrete floors. Acta Acust United Ac 2009; 95: 707–717.

Jeon

Sato

Annoyance caused by heavyweight floor impact sounds in relation to the autocorrelation function and sound quality metrics. J Sound Vib 2008; 311: 767–785.

Jeon

Lee

Kim

et al . Subjective evaluation of heavy-weight floor impact sounds in relation to spatial characteristics. J Acoust Soc Am 2009; 125(5): 2987–2994.

Kim

Ryu

Jeon

JY.

Effect of temporal decay on perception of heavy-weight floor impact sounds. J Acoust Soc Am 2013; 134(4): 2730–2738.

10.

Jeon

. Psychoacoustical evaluation of heavyweight floor impact sounds in apartment buildings. In: Proceedings of the inter-noise 2014, Melbourne, VIC, Australia, 16–19 November 2014.

11.

Ryu

Sato

Kurakata

et al . Relation between annoyance and single-number quantities for rating heavy-weight floor impact sound insulation in wooden houses. J Acoust Soc Am 2011; 129(5): 3047–3055.

12.

Gover

Bradley

Schoenwald

et al . Subjective ranking of footstep and low-frequency impact sounds on lightweight wood-framed floor assemblies. In: Proceedings of the Forum Acusticum 2011, Aalborg, 26 June 2011.

13.

Gover

Bradley

Zeitler

et al . Objective and subjective assessment of lightweight wood-framed floor assemblies in response to footstep and low-frequency impact sounds. In: Proceedings of the inter-noise 2011, Osaka, Japan, 4 September 2011.

14.

Späh

Hagberg

Bartlomé

et al . Subjective and objective evaluation of impact noise sources in wooden buildings. Build Acoust 2013; 20(3): 193–214.

15.

Kylliäinen

Hongisto

Oliva

et al . A laboratory listening experiment on subjective and objective rating of impact sound insulation of concrete floors. In: Proceedings of the inter-noise 2016, Hamburg, 21–24 August 2016.

16.

Kylliäinen

Hongisto

Oliva

et al . Subjective and objective rating of impact sound insulation of a concrete floor with various coverings. Acta Acust United Ac 2017; 103: 236–250.

17.

Öqvist

Ljunggren

Johnsson

. Listening test of walking noise from 20 Hz in dwellings. In: Proceedings of the inter-noise 2017, Hong Kong, China, 7 December 2017, pp. 3935–3942. Reston, VA: Institute of Noise Control Engineering.

18.

Monteiro

Machimbarrena

De la Prida

et al . Subjective and objective acoustic performance ranking of heavy and light weight walls. Appl Acoust 2016; 110: 268–279.

19.

Rychtáriková

Muellner

Chmelík

et al . Perceived loudness of neighbour sounds heard through heavy and light-weight walls with equal R_w+ C_(50–5000). Acta Acust United Ac 2016; 102(1): 58–66.

20.

Rychtáriková

Roozen

Müllner

et al . Listening test experiments for comparisons of sound transmitted through light weight and heavy weight walls. Akustika 2013; 19: 8–13.

21.

Rychtáriková

Müllner

Urbán

et al . Influence of temporal and spectral features of neighbour’s noise on perception of its loudness. In: Proceedings of the inner-noise 2013, Innsbruck, 15–18 September 2012.

22.

Pedersen

Antunes

Rasmussen

. Online listening tests on sound insulation of walls—a feasibility study. In: Proceedings of the inner-noise 2012, Prague, 2012, http://vbn.aau.dk/files/69825530/EuroNoise2012_Pedersen_Antunes_Rasmussen_OnlineListeningTests_Feasibility.pdf

23.

Park

Bradley

Gover

BN.

Evaluating airborne sound insulation in terms of speech intelligibility. J Acoust Soc Am 2008; 123(3): 1458–1471.

24.

Park

Bradley

Gover

BN.

Evaluation of airborne sound insulation in terms of speech intelligibility, Research Report (National Research Council Canada. Institute for Research in Construction); no. RR-228, NRC Institute for Research in Construction. National Research Council Canada: Canada, 2007.

25.

Park

Bradley

JS.

Evaluating standard airborne sound insulation measures in terms of annoyance, loudness and audibility ratings. J Acoust Soc Am 2009; 126(1): 208–219.

26.

Hongisto

Oliva

Keränen

Subjective and objective rating of airborne sound insulation—living sounds. Acta Acust United Ac 2014; 100: 848–864.

27.

Virjonen

Hongisto

Oliva

Optimized single-number quantity for rating the airborne sound insulation of constructions: living sounds. J Acoust Soc Am 2016; 140(6): 4428–4436.

28.

Ljunggren

Simmons

Hagberg

Correlation between sound insulation and occupants’ perception—proposal of alternative single number rating of impact sound. Appl Acoust 2014; 85: 57–68.

29.

Ljunggren

Simmons

Hagberg

. Findings from the AkuLite project: correlation between measured vibro-acoustic parameters and subjective perception in lightweight buildings. In: Proceedings of the inter-noise 2013, Innsbruck, 15–18 September 2013.

30.

Simmons

Ljunggren

Aku20—searching for optimal single number quantities in EN ISO 717-2 correlating field measurements 20-5000Hz to occupant’s ratings of impact sounds—new findings for concrete floors. In: Proceedings of the Euronoise 2015, Maastricht, 31 May–3 June 2015.

31.

Ljunggren

Simmons

Öqvist

Correlation between sound insulation and occupants’ perception—proposal of alternative single number rating of impact sound-part II. Appl Acoust 2017; 123: 143–151.

32.

Ljunggren

Simmons

Öqvist

. Evaluation of impact sound insulation from 20Hz. In: Proceedings of the 24th international congress on sound and vibration, ICSV24, London, 23–24 July 2017.

33.

Guigou-Carter

Balanant

Villenave

. Acoustic comfort evaluation in lightweight wood-based buildings. In: Proceedings of the Forum Acusticum, Krakow, 7–12 September 2014.

34.

Milford

Høsøien

Løvstad

et al . Socio-acoustic survey of sound quality in dwellings in Norway. In: Proceedings of the inter-noise 2016, Hamburg, 21 August 2016.

35.

Høsøien

Rindel

Løvstad

et al . Impact sound insulation and perceived sound quality. In: Proceedings of the inter-noise 2016, Hamburg, 21 August 2016.

36.

Hagberg

Evaluating field measurements of impact sound. J Build Acoust 2010; 17: 105–128.

37.

Hongisto

Mäkilä

Suokas

Satisfaction with sound insulation in residential dwellings—the effect of wall construction. J Build Environ 2015; 86: 309–320.

38.

ISO 532:1975. Acoustics—method for calculating loudness level.

39.

Zwicker

Fastl

Psycho-acoustics: facts and models. Berlin: Springer-Verlag, 1990.

40.

KS F 2810-2:2001. Method for field measurement of floor impact sound insulation —part 2: method using standardized heavy impact sources.

41.

ISO 717-2:1996. Acoustics—rating of sound insulation in buildings and of buildings elements—part 2: impact sound insulation.

42.

JIS A 1418-2:2000. Acoustic—measurement of floor impact sound insulation of buildings—part 2: method using standard heavy impact sources.

43.

ISO 140-7:1998. Acoustics—measurement of sound insulation in buildings and of building elements—part 7: field measurements of impact sound insulation of building elements.

44.

EN ISO 12354-2:2017. Building acoustics—estimation of acoustic performance of buildings from the performance of elements—part 2: impact sound insulation between rooms.

45.

ISO 16283-2:2014. Field measurement of sound insulation in buildings and of building elements—part 2: impact sound insulation.

Review of acoustic comfort evaluation in dwellings: part II—impact sound data associated with subjective responses in laboratory tests

Abstract

Keywords

Introduction

Methods

Summary of methods, metrics, and quantities in the reviewed studies

Evaluation of included studies

Results: associations of impact sound acoustic data with self-reported responses in laboratory tests

Discussion

Conclusion

Footnotes

Declaration of conflicting interests

Funding

References