| [Download PDF]
|Year : 2010 | Volume
| Issue : 48 | Page : 166--181
Evaluation of annoyance from low frequency noise under laboratory conditions
Malgorzata Pawlaczyk-Luszczynska1, Adam Dudarewicz1, Wieslaw Szymczak2, Mariola Sliwinska-Kowalska1,
1 Department of Physical Hazards, Nofer Institute of Occupational Medicine, Lodz, Poland
2 Environmental Epidemiology, Nofer Institute of Occupational Medicine, Lodz; Institute of Psychology, University of Lodz, Lodz, Poland
Department of Physical Hazards, Nofer Institute of Occupational Medicine, 8 Sw. Teresy Str., 91-348, Lodz
The aim of the study was to investigate the annoyance of low frequency noise (LFN) at levels normally prevailing at workplaces in control rooms and office-like areas. Two different laboratory experiments were carried out. The first experiment included 55 young volunteers and the second one comprised 70 older volunteers, categorized in terms of sensitivity to noise. The subjects listened to noise samples with different spectra, including LFNs at sound pressure level (SPL) of 45-67 dBA, and evaluated annoyance using a 100-score graphical rating scale. The subjective ratings of annoyance were compared to different noise metrics. In both the experiments, there were no differences in annoyance assessments between females and males. A significant influence of individual sensitivity to noise on annoyance rating was observed for some LFNs. Annoyance of LFN was not rated higher than annoyance from broadband noises without or with less prominent low frequencies at similar A-weighted SPLs. In both the experiments, median annoyance rating of LFN highly correlated with A-weighted SPL (L Aeq,T ), low frequency A-weighted SPL (L LFAeq,T ) and C-weighted SPL (L Ceq,T ). However, it is only the two latter noise metrics (i.e. L LFAeq,T and L Ceq,T ) which seem to be reliable predictors of annoyance exclusively from LFN. The young and older participants assessed similar annoyance from LFN at similar L LFAeq,T or L Ceq,T levels. Generally, over half of the subjects were predicted to be highly annoyed by LFN at the low frequency A-weighted SPL or C-weighted SPL above 62 and 83 dB, respectively.
|How to cite this article:|
Pawlaczyk-Luszczynska M, Dudarewicz A, Szymczak W, Sliwinska-Kowalska M. Evaluation of annoyance from low frequency noise under laboratory conditions.Noise Health 2010;12:166-181
|How to cite this URL:|
Pawlaczyk-Luszczynska M, Dudarewicz A, Szymczak W, Sliwinska-Kowalska M. Evaluation of annoyance from low frequency noise under laboratory conditions. Noise Health [serial online] 2010 [cited 2022 Jun 30 ];12:166-181
Available from: https://www.noiseandhealth.org/text.asp?2010/12/48/166/64974
Low frequency noise (LFN), usually considered as a broadband noise with the dominant content of low frequencies from 10 (20) Hz to 200 Hz, has been recognized a special environmental pollutant, affecting mostly sensitive people in their homes.  But exposures to LFN are not limited to the living environment; they are also present at workplaces, particularly in industrial control rooms and office-like areas. ,,, The importance of LFN in the general environment was pointed out in the World Health Organization (WHO) document on community noise.  The specific regulations on its control in the living environment (in dwelling room) are in use in some European countries. ,,, However, so far, less attention has been paid to LFN in the occupational settings and its adverse effects have been less well recognized. ,
The primary and the most frequent adverse effect of LFN is annoyance. ,,, Annoyance may be defined as a feeling of displeasure associated with any agent or condition known or believed by an individual or a group to be adversely affecting them.  It is closely related to feelings described by the words: disturbance, irritation, discomfort, dissatisfaction, bother, nuisance, uneasiness and distress. ,
Annoyance related to noise (sound) is a combination of both physical and psychological factors. The degree of annoyance or disturbance due to noise is difficult to predict and assess accurately in individuals.  The same noise may result in totally different responses in different people, depending on cultural factors, activity at the time of exposure, attitude to the noise source, subject's sensitivity to noise, controllability of the stressor, and other individual differences.  Over the years, many studies have been conducted to evaluate annoyance of LFN. They included both real life and laboratory experiments. , However, there are fewer published studies on annoyance reactions due to LFN in the occupational settings. It has been shown that LFN differs from ordinary medium- or higher-frequency noise. LFN may cause different subjective symptoms and higher rating of annoyance than noises at comparable A-weighted sound pressure levels (SPLs) that are not dominated by low frequencies. Annoyance from LFN is frequently reported at relatively low SPLs (slightly above the hearing threshold) and increases rapidly with its level. , It is worth to underline that previous experiences reported on people disturbed by LFN in their homes have shown that subjects who become annoyed by this type of noise develop a specific sensitivity to its sources, while they rarely consider themselves sensitive to noise in general.  Contrary to residential areas, where noise levels are often low, LFN in the occupational environment is usually well above the hearing threshold and, therefore, represents a different problem than household LFN. Several reports suggest that LFN at levels normally occurring in control rooms and office-like areas (40-50 dBA) can be perceived as annoying and adversely affecting the human mental performance, particularly when more demanding tasks have to be executed. Moreover, subjects recognized as highly sensitive to LFN may be at higher risk. ,,, Thus, LFN may possibly affect work performance, particularly in jobs requiring selective attention and/or processing of high load of information.
Several authors have shown that A-weighting characteristic, commonly used to evaluate occupational exposure to noise, is a less suitable descriptor for assessing the effects of LFN. ,, The same seems to be true for the G-weighting characteristic intended for infrasound assessment.  Many attempts have been made to find alternative measures that can better predict the effects of LFN. For example, Vercammen  proposed to apply for this purpose a low frequency A-weighted SPL determined from results of the frequency analysis in 1/3-octave bands from 10 to 160 Hz. On the other hand, the aforesaid WHO publication says: "Since A-weighting underestimates the sound pressure level of noise with low frequency components, a better assessment of health effects would be to use C-weighting". 
For making policy to noise control, it is important to set the relationship curves between annoyance and exposure levels. Many studies, in particular, surveys concerning transportation noise, have shown a large variation in individual annoyance reactions resulting in poor correlation between the noise exposure levels and the individual ratings.  Much better correlation coefficients were obtained for median of average ratings in subgroups.  An alternative method to determine median or average annoyance score is to evaluate the proportion of persons whose annoyance rating exceeds a certain annoyance level associated with given noise exposures. ,,,, Such exposure-response relationships have been determined for various types of environmental noise; ,,,, however, they are still missing for LFN in the occupational settings, although preliminary recommendations for its control to prevent annoyance and adverse effects on work performance have been recently proposed in Sweden and Poland. , Our recent observations from field studies comprising 276 workers confirmed that exposure to LFN in the industrial control rooms and in office-like areas at equivalent A-weighted SPLs (48-61 dB), which was well within current occupational exposure limits, could cause increased annoyance, especially in subjects recognized as more sensitive to this type of noise.  Our results were also in line with earlier findings that LFN was rated as more annoying than noise without (or with less prominent) low frequency content at similar dBA levels. Moreover, they suggested that the low frequency A-weighted SPL seemed to be a proper predictor for evaluation of the gravity of the annoyance experienced from LFN at workplace since, among the noise metrics, it yielded the highest correlation with subjective evaluations of LFN. 
The aim of the present study was to investigate annoyance from LFN at levels normally prevailing in the industrial control rooms and office-like areas under laboratory conditions. In particular, it has been attempted to:
Assess the effect of age, gender and individual sensitivity to noise on annoyance perception;Compare annoyance related to LFN and noise without or with less prominent low frequency content at comparable A-weighted SPLs;Examine some of the objective evaluation methods (noise metrics) that can give the best correlation with the subjective assessment of the LFN annoyance;Determine exposure-response relationships that show how annoyance ratings (expressed as a percentage of subjects whose annoyance rating exceeds a certain score) are associated with given LFN levels.
Materials and Methods
Two laboratory experiments on noise annoyance rating were carried out. The participants of the first study were young individuals, while an older population participated in the second study. In both the cases, persons were categorized in terms of individual sensitivity to noise in general and to LFN in particular. They listened to different noise samples, including LFNs, at A-weighted SPLs corresponding to levels normally occurring in industrial control rooms and office-like areas ,, and gave their subjective evaluation of noises.
The subjects received financial compensation for their participation in the experiments. The local Ethics Committee located at the Nofer Institute of Occupational Medicine at Lodz, Poland, approved the study design, methods and noise levels.
In experiment I, the study group comprised 55 pre-selected volunteers (27 females and 28 males), aged 18-28 years (mean 21.8 ± 2.1 years), who were high school or university graduates. Candidates were selected from among 117 young persons recruited by advertising, based on their scores on questionnaires used to separately evaluate individual sensitivity to noise in general and to LFN in particular. Additionally, each candidate underwent hearing test and only those with normal hearing (hearing threshold  was applied. The sensitivity to LFN was evaluated with a questionnaire comprising the following statements:
"I am not sensitive to noise with bass (low tones)",
"I think that even low intensity, monotonous humming (e.g. from a transformer) is unpleasant",
"I like listening to music when bass are turned on".
All aforesaid items had five response alternatives ranging from "do not agree at all" to "agree completely", graded from 1 to 5, respectively. The first and third items were scored in reverse order before the responses were summed.Both the questionnaires were validated before use. Their reliability was confirmed by Cronbach's alpha values equal to 0.881 and 0.500 in the case of the Weinstein noise sensitivity questionnaire and the LFN questionnaire, respectively. Subjects were categorized as high-sensitive or low-sensitive to noise in general or to LFN in particular on the basis of their questionnaires scores. The higher the result, the higher is the sensitivity to noise. Thus, persons who obtained at least median score (calculated for initial groups) were classified as high-sensitive to noise in general or to LFN in particular. The others were categorized as low-sensitive (NG- or LFN-). In the first step, only subjects who were recognized as high-sensitive to noise in general (NG+) and high-sensitive to LFN (LFN+) or low-sensitive to noise (NG-) and low-sensitive to LFN (LFN-) were eligible for the study. In the second step, in order to increase the number of cases, persons with a relatively high or low score in one of two questionnaires were also included to the study. In the first experiment, 30 subjects were recognized as high-sensitive to LFN and 33 as high-sensitive to noise in general, while in the second study 39 and 33 persons were classified as high sensitive to LFN and to noise, in general, respectively. However, in both the cases, the two sensitivity distributions were not identical, i.e. a fraction of high-sensitive to LFN subjects was also recognized as low-sensitive to noise in general and vice versa [Figure 1]. Nevertheless, there was a relatively high correlation between individual sensitivity to noise in general and sensitivity to LFN in terms of scores in questionnaires. Pearson correlation coefficients ranged from 0.596 to 0.804 (P Exposure to noise
In the first experiment, four stationary noises of artificial origin with different frequency contents were presented at nominal A-weighted SPLs of 45, 50, 55, 60 and 65 dB (5 series of 5 nominal levels Χ 4 spectra) [Figure 2]. All of them were steady-state broadband noises. Spectra A, B and C were LFNs, while spectrum D was noise without dominant low frequency contents with a more or less flat spectrum. (It had been assumed that the difference between C- and A-weighted SPLs (LC - LA ) exceeding 15 dB was a distinguishing feature of LFN.  ) Tonal components were present in spectra A (at 1/3-octave band of 200 Hz), C (at 1/3-octave bands of 40 and 80 Hz) and D (at 1/3-octave bands of 25, 50 and 200 Hz). (According to ISO 9612:1997,  if the level in a particular 1/3-octave band exceeded the level of the adjacent bands by 5 dB or more, the noise was said to be tonal.) All noise samples, each lasting 30 seconds, were played once in the same order to each subject. The sequence of level presentations as well as the order of spectra in each series was selected at random.
In the second experiment, seven stationary noises with diversified contents of low frequencies were presented three times, in the same order, at different A-weighted SPLs (from 45 to 67 dB); the differences resulted from different levels of amplification of the reproduced signals (0, 5 and 10 dB) [Figure 3]. All of them were steady-state broadband noises. Tonal components occurred in spectra nos. II (at 1/3-octave band of 50 Hz), V (at 1/3-octave bands of 25, 50 and 200 Hz) and VII (at 1/3-octave band of 200 Hz). Almost all the spectra simulated noise occurring in the work environment, including offices (spectrum no. I), ventilation systems (spectrum no. II), industrial control rooms (spectra nos. III-V) and laboratories with equipments for chemical analysis (spectrum no. VII). Spectra nos. I-VI were classified as LFNs, while spectrum no. VII had less prominent content of low frequencies. Each noise sample lasted 75 seconds and was reproduced once. Experiment I was performed in a sound proof room for audiometric tests (6.8 m 2 area), while experiment II was carried out in a cabin designed for psychological tests (6.2 m 2 area) and furnished as an office environment. Rooms were acoustically treated to fulfill the requirements of ISO 6189:1983.  In both the cases, the subjects were placed in the middle of the room in a sitting position. The noise was emitted by the TONSIL type FOCUS SE set of loudspeakers with the TONSIL type AKTIV 200 active subwoofer (TONSIL SA, Wrzesnia, Poland). Recorded noise samples were reproduced directly from a personal computer (in experiment I) or using the SONY type D-NE005 portable CD player (Sony Corporation, China ) (in experiment II) and amplified by the DENON type AVR-1603 AV surround receiver (Denon Ltd., China). During the listening sessions, noise exposure conditions were monitored via a microphone located close to the head of each participant, approximately 0.10 m from the entrance to the external canal of the right ear. The measuring system consisted of the Brόel and Kjζr (BandK) type 4190 microphone, BandK type 2231 sound level meter (Brόel & Kjζr Sound & Vibration Measurement A/S, Nζrum, Denmark), SVANTEK type SVAN 912 and 912E sound analyzers (SVANTEK Sp. z o. o., Warsaw, Poland) and the Hewlett-Packard type 3569A frequency analyzer (Hewlett-Packard Company, Everett, USA) was used for monitoring.
Objective evaluations of noise samples
Basic noise parameters such as equivalent-continuous A-, C- and G-weighted SPLs (LAeq,T, LCeq,T and LGeq,T) were measured. In addition, SPLs in 1/3-octave bands from 10 Hz to 10 kHz were determined. Each single measuring period lasted 20 and 40 seconds during the first and second experiments, respectively. A linear averaging was applied for this purpose.
From the results of frequency analysis, additional measures, i.e. low frequency A-weighted SPL (LLFAeq,T), total SPL (LTot ), low frequency total SPL (LLF Tot ), and excess of hearing threshold level (HTL) (LLF HTL ) were calculated. The LLFAeq,T level was determined from 10 to 200 Hz bands and the nominal A-weighting corrections were added to the spectra.  The LTot and LLF Tot levels were calculated from summation of 1/3-octave bands exceeding the HTL according to ISO 226:2003  across the frequency range 20-10,000 Hz and 20-200 Hz, respectively. The LLF HTL level corresponded to the highest value of excess of HTL by SPLs in 1/3-octave bands in the frequency range from 20 to 200 Hz.
Subjective evaluations of noise samples
In experiment I, subjects listening to different noises were asked to imagine that in such noise conditions they would have to perform jobs requiring increased mental processing and selective attention. Immediately after completion of each noise sample, during which they did not perform any real tasks, they assessed the noise annoyance, its loudness and the degree of its disturbing effect in case of mentally demanding tasks as well as simple routine jobs.
Prior to the listening session, the subjects were familiarized with noise examples. After the session, they were also asked to complete a questionnaire intended to capture symptoms experienced while listening to the noise samples.
In the second experiment, the noise samples were assessed in a similar way. However, in this case the listening session was preceded by performing logical tasks for 30 minutes during exposure to LFN at A-weighted SPL of ca. 40 dB [Figure 3]h. Subjects were executing these tasks in order to help "imagine" themselves performing mentally demanding tasks as well as to get used to a slightly noisy environment.
In both the experiments, instructions to the subjects were given orally and the same information was provided to them in the written form. Participants provided their evaluation of noise samples on a paper form, using 100-score graphical rating scales with verbally labeled two poles:
"not at all annoying" and "very annoying" for noise annoyance rating,"not at all loud" and "very loud" for loudness assessment and
"not at all disturbing" and "very disturbing" for the evaluation of disturbing effects of noise in case of routine jobs and more demanding tasks involving mental processing and selective attention.
During the first experiment, each session lasted for a total of about 35 minutes, while in the second experiment it took ca. 75 minutes.
In both the experiments, relations between various subjective evaluations of noise samples (i.e. annoyance, loudness and disturbing effects of noise) were analyzed using Pearson's correlation coefficient. For each individual noise sample, differences between annoyance and loudness as well as differences in assessment of disturbing effects in case of mentally demanding tasks and simple ones, were analyzed using t-test for dependent data. Different statistical tools were used to determine the influence of gender and individual sensitivity to noise on annoyance ratings of each individual noise sample. One-factorial (between-subjects design) analysis of variance (ANOVA) for independent data was performed to evaluate the impact of gender. To assess the influence of sensitivity to noise in general or sensitivity to LFN in particular and to identify a possible interaction between these two different sensitivities, two-factorial (between-subjects design) ANOVA for independent data was applied.The influence of nominal A-weighted SPL (experiment I) or the amplification level of noise samples (experiment II) on subjective assessments was analyzed (individually for each type of spectrum) using within-subject ANOVA for dependent data, supplemented by analysis of contrasts or t-test for dependent variables. The same tools were applied to compare annoyance rating for noise samples of different frequency spectra and at similar A-weighted SPLs.
Due to variations in individual annoyance ratings, in particular, unsymmetrical distributions of subjective evaluations in case of the lowest and highest values of nominal presentation/amplification level of noise samples, a median annoyance score was determined for each noise sample. The relations between objective evaluations of noises and their median annoyance ratings were analyzed using Pearson's correlation coefficient. The resultant correlation coefficients were compared in pairs using t-test. In order to compare the annoyance assessment related to LFN at similar levels in the first and second experiments, linear regression models were applied. Logistic regression analysis was used to determine the relationships between noise parameters and percentage of subjects whose annoyance rating exceeded a certain score. Similarly, to transportation noise, the cutoff at 28 on a 0-100 scale was ascribed to a term of (at least) "a little annoyed" (LA), while the cutoffs at 50 and 72 were related to terms "annoyed" (A) and "highly annoyed" (HA), respectively.  The statistical analysis was done with an assumed level of significance α = 0.05, excluding the comparisons in pairs using t-test for dependent variables. For the latter purpose, α-value (α = 0.05) divided by number of possible comparisons was set as the limit for statistical significance. The statistical analyses were carried out using SPSS (version 14.0 for Windows, SPSS Inc., Chicago, IL, USA) and STATISTICA (version 6. StatSoft, Inc.) software packages.
Subjective evaluations of noise samples
Subjective assessments of noise samples played in experiment I and experiment II are presented in [Figure 4] and [Figure 5]. The same tendency was observed in both the experiments. Annoyance related to noise was rated higher than its loudness for majority of the noise samples, excluding spectrum A at nominal level of 65 dBA, spectra B and C at nominal 60 dBA (−2.400 ≤ t(54) ≤ −1.227, 0.020 ≤ P ≤ 0.225, α = 0.05/6) in the experiment I and spectra nos. I, II and III at an amplification level of 5 dB and spectrum no. VI at an amplification level of 10 dB (−2.685 ≤ t(69) ≤ −1.634, 0.009 ≤ P ≤ 0.107; α = 0.05/6) in experiment II [Figure 4]a and [Figure 5]a. Similarly, in both the experiments, subjective ratings of disturbing effects related to noise were significantly higher for mentally demanding tasks than for simple ones (7.628 ≤ t(54) ≤ 12.366 and 7.212 ≤ t(69) ≤ 9.290, Pα = 0.05/6) [Figure 4]b and [Figure 5]b. Regardless of the subjects' sensitivity to noise, close relations between all subjective assessments were found (Pearson's correlation coefficients varied from 0.764 to 0.934, Pr=0.934, Pr = 0.893, Pr = 0.884, PF(1,53) ≤ 1.771, 0.925 ≥ P ≥ 0.189; experiment II: 0.004 ≤ F(1,68) ≤ 2.226, 0.947 ≥ P ≥ 0.140). A significant impact of sensitivity to noise in general (NG) or to LFN in particular as well as an interaction between these two sensitivities was observed in the annoyance assessment related to some noise samples in experiment I [Figure 4]a. On the other hand, in experiment II, the only significant differences were those resulting from different sensitivity to NG [Figure 5]a. In experiment I, the subjects recognized as LFN+ assessed the annoyance related to spectrum A at a nominal level of 60 dBA (F(1,51) = 5.308, P=0.025) and 65 dBA (F(1,51) = 5.289, P=0.026) as well as that related to spectra B and D at 60 dBA (F(1,51) = 5.494, P=0.023; F(1,51) = 5.274, P=0.026) higher than the LFN individuals [Figure 6]. The same relation was observed in case of the NG+ and NG- persons for spectrum B at 45 and 50 dBA (F(1,51) = 4.389, P=0.041; F(1,51) = 5.174, P=0.027). An interaction between sensitivity to NG and sensitivity to LFN was noted for spectrum A at the level of 60 dBA (F(1,51) = 6.707, P=0.012). In this case, in the NG subgroup, the subjects recognized as LFN+ rated the annoyance higher than low-sensitive ones, while in the NG+ subgroup no significant difference in annoyance assessment between persons of different sensitivities to LFN was found [Figure 7]. Regardless of the nominal level, there were no significant differences resulting from different sensitivities to noise in case of spectrum C (0.001 ≤ F(1,51) ≤ 3.726, 0.982 ≥ P ≥ 0.059). In experiment II, subjects recognized as NG+ rated annoyance related to spectra nos. II, III, IV and VI at amplification levels of 0, 10 and 5 dB (only in case of spectrum IV) higher than the low-sensitive ones (4.004 ≤ F(1,66) ≤ 6.470, 0.049 ≥ P ≥ 0.013) [Figure 8]. For other noise samples (spectra nos. I, V and VII at various amplification levels), there were no significant differences between subjects showing different sensitivities to noise (0.567 ≤ F(1,66) ≤ 3.754, 0.454 ≥ P ≥ 0.057). As expected, the nominal level of noise samples (in experiment I) or amplification level of reproduced signals (in experiment II) had a significant influence on the subjective evaluations of each type of noise spectra (experiment I: 363.291 ≤ F(1,54) ≤ 621.809, PF(1,69) ≤ 747.510, PF(1,54) ≤ 1164.818, PF(1,54) = 3.361, P=0.072; F(1,54) = 2.019, P=0.161) and C at 65 dBA (F(1,54) = 2.498, P=0.120) [Figure 9]a. For other cases (combinations of spectra and/or nominal levels), the LFN samples were assessed as less annoying than samples of broadband noise without dominant content of low frequencies (10.229 ≤ F(1,54) ≤ 36.502, P≤0.002). Similarly, while comparing annoyance related to different LFN samples at similar levels, no significant differences were noted for spectra B and C (for all nominal levels), i.e. in case of noises characterized by similar values of the LC - LA difference (−2.485 ≤ t(54) ≤ 1.159, 0.016 ≤ t ≤ 0.251, α = 0.05/6). No significant differences were also observed for spectra A and B at 55 dB (t(54) = 2.414, P=0.019, α = 0.05/6) or spectra A and C at 65 dB (t(54) = 0.830, P=0.410, α = 0.05/6).
In the second experiment, noises which were reproduced at the same amplification level differed in the A-weighted SPLs [Figure 3]. Therefore, seven noise samples with different kinds of spectra but similar level of 57 ± 1.2 dBA were chosen for comparison (i.e. spectra nos. I, II, III at an amplification level of 10 dB, spectra nos. IV and V at 5 dB and spectra nos. VI and VII at 0 dB). There were no significant differences in annoyance rating between a broadband noise with less dominant content of low frequencies (spectrum no. VII) and some LFNs, i.e. spectra nos. II, III and VI (0.028 ≤ F(1,68) ≤ 2.604, 0.867 ≥ P ≥ 0.111) [Figure 9]b. The other LFN samples (spectra nos. I, IV and V) were rated as less annoying than spectrum no. VII (17.679 ≤ F(1,68) ≤ 28.690, Pt(69) ≤ 7.777, 0.00023 > P>0.00000, α = 0.05/21) Thus, in both the experiments, annoyance related to LFN was not assessed higher than annoyance from noises without or with less prominent low frequency content at similar A-weighted SPLs.
Low frequency noise annoyance rating verses noise metrics
In both the studies, linear relationships between median LFN annoyance rating and all noise metrics, excluding G-weighted SPL in the experiment II, were found [Table 1]. Statistically significant Pearson's correlation coefficient for G-weighting noted in the first experiment was relatively low compared to the high correlation coefficients obtained for other noise parameters (significant differences (0.001 ≤ P ≤ 0.004) were noted for almost all cases excluding LCeq,T).
The measures based on hearing threshold levels (i.e. LTot , LLF Tot and LLF HTL ) gave the highest value of correlation coefficient for subjective annoyance rating in the first experiment with young participants, while in the second experiment (with involvement of older persons) C-weighting provided the highest correlation [Table 1]. The next best measures in experiment I were either A-weighted SPL or low frequency A-weighted SPL. In experiment II, relatively high correlation coefficients were also noted for the low frequency and "normal" A-weighing (LLFAeq,T and LAeq,T ), while the lowest value was obtained for the LLF HTL level. Subjective assessments of LFN were significantly lower correlated with LTot , LLF Tot and LLF HTL , i.e. noise measures which were calculated on the basis of HTL  in case of older subjects (experiment II) in comparison with younger persons (experiment I) (0.001PLAeq,T and LLFAeq,T levels were similar. The same relationship was also obtained for the LTot and LLF Tot levels. These results are not surprising. Due to the frequency content of the LFN samples used in both the experiments, the aforesaid pairs of noise parameters highly correlated (Pearson's correlation coefficients were 0.999 and 0.983 (PLAeq,T and LTot levels. The correlation of the other noise parameters (i.e. LLFAeq,T, LLF Tot and LCeq,T) with annoyance assessment was significantly lower (0.001PLAeq,T and LTot levels. As stated above, in both the experiments, subjective evaluations of LFNs highly correlated with the LAeq,T, LLFAeq,T, LCeq,T and LLF Tot levels. However, for the latter parameter, a significant difference between the experiments in correlation coefficients was found (P=0.007). On the other hand, the equivalent A-weighted SPL highly correlated with median annoyance of all noise samples irrespective of their frequency content. Therefore, only the equivalent-continuous low frequency A-weighted SPL and the equivalent-continuous C-weighted SPL seem to be suitable for the assessment of annoyance related exclusively to LFN.
The linear regression models relating median LFN annoyance rating (dependent variable) with the LLFAeq,T or LCeq,T levels (predictors) are presented in [Figure 10]. In case of the first experiment, the linear regression models had little higher slopes and lower intersections than the regression models for the second experiment, but their confidence intervals overlapped [Figure 10]a and b. This means that annoyance assessments of LFNs at similar LLFAeq,T or LCeqT levels were similar in young and older groups.
Based on individual annoyance ratings, the variables of being highly annoyed (HA), annoyed (A) and a little annoyed (LA) were calculated as dichotomic data, and an attempt was made to determine exposure-response curves expressed as probability (percentage) of HA, A or LA as function of noise parameter. The logistic regression analysis was applied for this purpose with the logistic model expressed as follows:
y - is the dichotomic variable (0 or 1) of being highly annoyed, annoyed or a little annoyed, i.e. y = 0 for annoyance rating lower than cutoff (28, 50 or 72) and y = 1 for other cases, x - is the explanatory variable, i.e. noise exposure metrics (LLFAeq,T or LCeq,T), β0 , β1 - are the regression coefficients in the logistic model.
The estimates of β0 and β1, fitted to the data from both the experiments, are presented in [Table 2] with their estimated standard errors and confidence probabilities in Wald test. [Figure 11] and [Figure 12] show the exposure-relationship curves that were determined by the aforesaid logistic fit procedure.According to these curves, the LFN at the LLFAeq,T levels above 54 dB might be assessed as annoying by at least 50% of subjects performing mentally demanding tasks, while by 23 and 82% of them as highly or a little annoying, respectively [Figure 11]. Over half of the persons were predicted to be highly annoyed by LFN at low frequency A-weighted or C-weighted SPLs exceeding 83 and 62 dB, respectively [Figure 11]a and [Figure 12]a. The same portion of subjects might perceive LFN as a little annoying at the LLFAeq,T or LCeq,T levels above 45 and 66 dB, respectively [Figure 11]c and [Figure 12]c.
The general objective of the study was to analyze the annoying effects of LFN at the A-weighted SPLs corresponding to those occurring in offices and industrial control rooms. For this purpose, two laboratory investigations were carried out. They differed in exposure conditions, age of participants and the experimental procedure, since only in the second experiment listening to noise sample was preceded by performing logical tasks. In both the cases, the subjects evaluated steady-state broadband noises with different types of spectra, including LFNs (spectra A, B, C in the first experiment and spectra nos. I-VI in the second experiment), and noises without dominant content of low frequencies (spectrum D in the first experiment) or noises with less prominent content of low frequencies (spectrum no. VII in the second experiment). Tonal components were present in some of LFNs as well as in noises without or with less prominent content of low frequencies. It is worth noting that the calculation of the difference between C- and A-weighted SPLs is commonly used to identify the frequency composition of noise. A greater difference between these values indicates a greater content of low frequencies in the spectrum. It is often assumed that the difference exceeding 15 dB is an indicator of actual LFN.  However, some definitions of LFN rely on a difference of at least 20 dB.  Samples of LFN used in both the experiments were characterized by the LC - LA differences varying from 15 to 21 dB. Our earlier studies on occupational exposure to LFN showed that SPLs measured at workplaces in industrial control rooms and office-like areas varied from 48 to 61 dBA. ,, The incidence of low frequency components in the spectra was diversified, the differences between C- and A-weighted SPL ranged from 15 to 20 dB. Thus, LFNs which were used in both the studies (as far as dBA levels and frequency contents characterized by the LC - LA differences were concerned) corresponded quite well to noise normally occurring in the work environment. In the first experiment, each noise sample lasted 30 seconds, while in the second experiment the duration was 75 seconds. This difference seems to be of less importance. As it was shown by Moller  in his study on audible infrasound, annoyance ratings were independent of the exposure times used in the experiment (i.e. 30 seconds, 3 minutes and 15 minutes).Over the years, there have been a large number of laboratory determinations of annoyance from LFNs, but usually subjects were asked to imagine themselves relaxing at their homes in the evening or in similar conditions.  Whereas in the present study, subjects from both the experiments were asked to imagine that in such noise conditions they would have to perform jobs requiring increased mental processing and selective attention, and after completion of each noise sample, they assessed the noise annoyance, loudness and the degree of disturbing effect of noise in case of routine jobs and mentally demanding tasks. However, during listening to noise samples, they did not perform any real tasks. Regardless of the noise sample, for almost all cases (combination of spectra and levels), noise annoyance was rated higher than its loudness. Similarly, disturbing effects related to noise were significantly higher in the case of the mentally demanding than simple tasks. The latter results are in agreement with those of Kjellberg and Landstrom  and Landstrφm et al.  In their review, Kjellberg and Landstrφm  suggest that noise tends to be more annoying when complex tasks are being carried out. Accordingly, Landstrφm et al., found that the tolerance level was about 6 dB lower in a difficult reasoning task than in a simple reaction-time task while estimating annoyance thresholds. In both the experiments, there were close correlations between all subjective evaluations of noise samples. This is why annoyance rating was the only parameter taken into account in further evaluations. One of the detailed objectives of this study was to analyze the possible influence of gender, age and sensitivity to noise on subjective evaluations of noise sample. However, significant differences between females and males in noise annoyance rating were not observed in any of the two experiments. On the other hand, younger and older subjects similarly assessed annoyance related to LFN at similar low frequency A-weighted SPLs or C-weighted SPLs. These results are in line with earlier observation that demographic variables, especially sex and age, were found to have no important effect on annoyance from environmental noises.  Previous studies on the effects of community LFN (in dwelling rooms) have shown that subjects sensitive to this type of noise were not necessarily sensitive to noise in general as measured by general noise sensitivity scales.  Moreover, later investigations concerning the influence of LFN on performance confirmed that sensitivity to this special type of noise was somewhat different from sensitivity to noise in general.  This is why these two types of noise sensitivities were taken into consideration in this study and they were analyzed together (using two-factorial between-subjects design ANOVA).Contrary to age and gender, in this study, individual sensitivity to noise had a significant impact on noise annoyance rating. However, it was observed only for some noise samples. In the first experiment, significant main effect of sensitivity to NG or sensitivity to LFN as well as an interaction of these two factors was found in the case of 5 out of 15 LFN samples, while in the second experiment only significant influence of sensitivity to NG was noted for 9 of 18 LFN samples. Thus, the influence of sensitivity to LFN on annoyance rating was not as obvious as expected. This might be due to short time (30 or 75 seconds) of noise presentation, especially as the relationship between annoyance and individual sensitivity to noise (in general or to LFN in particular) was noted in some of our earlier laboratory studies dealing mainly with possible influence of LFN on performance, in which longer time of exposure to noise was used. , It was also observed in our previous field study cited above, concerning occupational exposure to LFN.  For example, Persson Wye et al.,  (in their article cited below in more detail) showed that exposure to LFN was rated as more annoying by subjects recognized as high-sensitive to this type of noise than by low-sensitive ones. Interestingly, their evaluation of sensitivity to LFN was based on statements or questions taking into consideration subjective sensations related to LFN (e.g. "Are you sensitive to low frequency noise?" or "I am sensitive to rumbling noise from ventilation system"). Thus, it was similar to the way of classification used in this study.
Another aim of the study was to compare annoyance related to LFN and noise without or with less prominent low frequency content at comparable A-weighted SPLs. However, contrary to our expectations, annoyance related to LFN was not rated higher than annoyance from noises without (experiment I) or with less prominent content (experiment II) of low frequencies at similar A-weighted SPLs. These findings did not confirm results from the majority of previous laboratory investigations intended to compare annoyance assessments of various sounds with and without dominant contents of low frequencies , as well as our earlier observation from field study.  On the other hand, higher assessments of LFN annoyance relative to noises without dominant content of low frequencies but similar A-weighted SPLs were not so obvious in laboratory studies dealing mainly with possible influence of LFN on performance. ,,, Persson et al., found out that subjects executing various high-load tasks for 2 hours reported a higher degree of annoyance and impaired working capacity when working under LFN conditions than those exposed to broadband noise with rather flat frequency spectrum, both at a level about of 40 dBA. However, the same authors in their earlier study did not observe significant differences in noise annoyance rating in subjects performing the high-load tests in similar noise exposure conditions (ca. 40 dBA), but for a shorter time (1 hour).  Similar results were also noted in other studies. , LFN at 45-50 dBA was more detrimental to performance than noise without dominant content of low frequencies (at the same level), while no significant differences in annoyance assessment between noises of different spectra were noted. ,
It has to be noted that many factors inherent in the noise (e.g. temporal variation, tonal steadiness, tonal frequency or non-tonality) may influence its annoying effect. Since all noise samples played in both experiments had steady-state temporal pattern, possible differences in annoyance related to noises at similar A-weighted SPLs apart from different frequency content might result from tonal character of noise. However, in the first experiment, regardless of the shape of spectra (presence or absence of low frequency tonal components), there were no significant differences in annoyance ratings between LFNs characterized by a similar LC - LA difference, i.e. spectra C and B [Figure 2]. Moreover, a tendency was observed to assess annoyance related to aforesaid spectra lower than that related to spectrum A or spectrum D. But such relations were not observed in the second experiment. Some studies showed the importance of spectrum balance between high and low frequencies for acceptable versus unacceptable LFNs. , Attempts were made to determine what slope of spectrum is unacceptable or strong rumble and what slope is considered to be acceptable, neutral or pleasant. A "good" slope seems to be in the range of 4 dB per octave, while a "bad" and potentially annoying sound seems to have a slope of about 6-7 dB per octave or more. , However, it was difficult to draw unambiguous conclusions as to why in our present study some LFN spectra were rated as more annoying than other ones. Several previous experimental and field studies have found that A-weighted SPL is a less suitable descriptor of LFN. ,, Therefore, the next part of the study was intended to examine some noise measures that can give the best relation with the assessment of the LFN annoyance. For this purpose, we compared subjective ratings of LFN with various noise metrics. Together with basic measures, i.e. the equivalent-continuous A-, C- and G-weighted SPLs (LAeq,T, LCeq,T and LGeq,T ), additional measures such as low frequency A-weighted SPL (LLFAeq,T), total SPL (LTot ), low frequency total SPL (LLF Tot ), and excess of hearing threshold level in the 20-200 Hz frequency range (LLF HTL ) were analyzed.
It is worth noting that the majority of existing or proposed evaluating methods for indoor LFN are based on the frequency analysis in 1/3-octave bands in the frequency range from 8 to 200 Hz. In most of the cases, measured SPLs are compared with various criterion curves more or less close to hearing threshold levels.  However, in the Danish method based on these results, the low frequency A-weighted SPL (in the frequency range of 10-160 Hz) is calculated, and if necessary, a 5-dB penalty for impulsive noise is added.  Likewise, in the German method, if the noise is not tonal, the low frequency A-weighted SPL in the 10-80 Hz frequency range is determined, but only based on bands exceeding the hearing threshold. 
The aforesaid exposure criteria for LFN in the domestic environment had been compared under laboratory study by Poulsen.  He played to subjects different environmental LFNs, including impulsive noises from drop forge and discotheque music, at relatively low A-weighted SPLs (20-35 dB) and found that the Danish method was the one best correlated with subjective evaluations (mean values) of noise annoyance. When he excluded the impulsive discotheque sounds from analysis, it turned out that the Swedish method was as good as the Danish one. It is worth to underline that the Danish method includes the 5-dB penalty for impulsive LFN. However, when we had re-analyzed the Poulsen  data after prior elimination of this penalty for discotheque and drop forge noises, we found that the evaluation method based on low frequency A-weighted SPL still gave the best correlation with subjective annoyance rating. Please note that in the aforesaid study the relationship between A-weighted SPL and annoyance rating was not analyzed. It was only pointed out that the noise spectrum, the nominal level of presentation and the measured A-weighted SPL had a significant influence on the subjective evaluations of various samples of LFN.
Recently, Subedi et al.,  measured annoyance of low frequency pure and combined tones in a laboratory experiment. They compared these results (median values) with the evaluation obtained from three objective methods, namely a method based on Moore's loudness model, the total energy summation model (in this study referred to as the total SPL), and the low frequency A-weighting. Among these methods the latter one gave the best correlation. Similar results were obtained in the field study during which various evaluation methods were compared with subjective assessment of LFN at workplaces in the industrial control rooms. 
In this study, a linear relationship between the LFN annoyance rating and almost all analyzed noise metrics was noted. But subjective assessments of LFN were significantly less correlated with HTL-based noise measures (especially LLF Tot and LLF HTL ) in older subjects (experiment II) compared with younger persons (experiment I). For the other noise parameters (excluding LGeq,T) no significant differences between the experiments were noted. On the hand, in both younger and older participants, median annoyance rating of LFN highly correlated with A-weighted SPL (LAeq,T), low frequency A-weighted SPL (LLFAeq,T) and C-weighted SPL (LCeq,T). However, in our opinion, among the analyzed noise metrics, only the low frequency A-weighted SPL and C-weighted SPL seem to be suitable predictors of annoyance exclusively from LFN. These findings confirm the results of the above mentioned studies as well as the suggestion found in the WHO publication that using C-weighting would provide a better assessment of LFN. 
For making policy to control noise, it is important to set the relationships between annoyance and exposure levels. Therefore, similar to transportation noise and other environmental noises, ,,,, exposure-response relationship curves expressed as percentage of highly annoyed, annoyed or a little annoyed persons associated with given exposure to LFN have been determined. Since these curves are based on limited data and were obtained under laboratory conditions, their interpretations, especially with relevance to normal work situations should be done with great care. Similar limitations are also valid for other observations from this study. Nevertheless, results presented here suggest that LFN at the low frequency A-weighted SPLs above 62 dB might be assessed as highly annoying by at least 50% of subjects performing mentally demanding tasks. Likewise, over half of the persons might be highly annoyed after exposure to LFN at the C-weighted SPLs exceeding 83 dB. To sum up, our findings confirm that LFN at levels occurring in the industrial control rooms and office-like areas can be perceived as annoying. To evaluate the degree of the annoyance experienced from LFN in occupational settings, the low frequency A-weighted and C-weighted SPLs seem to be reliable predictors. However, since data on LFN in the work environment are limited, further studies are needed.
This study was supported by sixth European Framework Project under the Marie Curie Host Fellowship for the Transfer of Knowledge "NoiseHear" (Contract MTKD-CT-2004-003137) and the Ministry of Science and Higher Education of Poland (grant IMP 18.5/2004−IMP 2005)
|1||Leventhall HG. Low frequency noise and annoyance. Noise Health 2004;6:59-72. |
|2||Berglund B, Hassmen P, Job RF. Sources and effects of low-frequency noise. J Acoust Soc Am 1996;99:2985-3002.|
|3||Pawlaczyk-Luszczynska M, Dudarewicz A, Waszkowska M. Assessment of low frequency noise annoyance in steering premises according to subjective annoyance rating of workers: A pilot study. Med Pr 2001;52:465-70.|
|4||Kaczmarska A, Mikulski W, Pawlaczyk-Luszczynska M. A study of annoyance of low-frequency noise in rooms for office and conceptual work. Bezp Pr 2006;1:16-9.|
|5||Pawlaczyk-Luszczynska M, Dudarewicz A, Waszkowska M, Sliwinska-Kowalska M. Annoyance related to low frequency noise in subjective assessment of workers. J Low Frequency Noise Vib Active Cont 2009;28:1-17. |
|6||Berglund B, Lindvall T, Schwella D, Goh KT, editors. Guidelines for Community Noise. Geneva: World Health Organization; 2000. |
|7||DIN 45680. Measurement and assessment of low frequency noise immissions in the neighborhood. Berlin: Deutches Institut Fur Normung;1997.|
|8||Jakobsen J. Danish guidelines on environmental low frequency noise, infrasound and vibration. In: Moller H, Lydolf M, editors. Proceedings of the 9th International Meeting on Low Frequency Noise and Vibration; 2000 May 17-19; Aalborg, Denmark: Department of Acoustics, Aalborg University; 2000. p. 57-64. |
|9||Mirowska M. Evaluation of low-frequency noise in dwellings: New Polish recommendations. J Low Frequency Noise Vib Active Cont 2001;20:67-74.|
|10||Moorhouse A, Waddington D, Adams M. Proposed criteria for the assessment of low frequency noise disturbance. DEFRA Contract no NANR45. Salford: University of Salford, 2005). Available from: HYPERLINK http://www.defra.gov.uk/environment/" http://www.defra.gov.uk/environment/noise/research/lowfrequency/pdf/nanr45-procedure.pdf [Last cited on 2009].|
|11||Persson Waye K. Effects of low frequency noise in the occupational environment - present knowledge base. In: Selamet A, Singh R, Maling GC, editors. Proceedings of the Inter-Noise 2002, The 2002 International Congress and Exposition on Noise Control Engineering; 2002 Aug 19-21; Dearborn, MI, USA. Ames, IA: Institute of Noise Control Engineering Inc; 2002 [CD-ROM, paper no. N273].|
|12||Persson Waye K. Adverse effects of moderate levels of low frequency noise in the occupational environment. ASHRAE Trans 2005;111:672-83.|
|13||Broner N. The effects of low frequency noise on people. J Sound Vib 1978;58:483-500.|
|14||Persson Waye K, editor. On the effects of environmental low frequency noise [Ph.D. dissertation]. Gothenburg: Gothenburg University; 1995.|
|15||Borsky PN. Sonic boom exposure effects II.4: Annoyance reactions. J Sound Vib 1972;20:527-30.|
|16||Guski R. Personal and social variables as co-determinants of noise annoyance. Noise Health 1993;3:45-56.|
|17||Guski R, Flescher-Suhr U, Schuemer R. The concept of noise annoyance: How international experts see it. J Sound Vib 1999;223:513-27. |
|18||Persson Waye K, Rylander R, Benton S. Effects on performance and work quality due to low frequency ventilation noise. J Sound Vib 1997;205:467-74.|
|19||Persson Waye K, Bengtsson J, Kjellberg A, Benton S. Low frequency noise pollution interferes with work performance. Noise Health 2001;4:33-49.|
|20||Bengtsson J, Persson Waye K, Kjellberg A. Evaluation of effects due to low frequency noise in a low demanding work situation. J Sound Vib 2004;278:83-99. |
|21||Pawlaczyk-Luszczynska M, Dudarewicz A, Waszkowska M, Szymczak W, Sliwinska-Kowalska M. The impact of low-frequency noise on human mental performance. Inter J Occup Med Environ Health 2005;18:185-98.|
|22||ISO 7196. Acoustics - Frequency weighting characteristic for infrasound measurements. Geneva: International Organization for Standardization;1995.|
|23||Vercammen ML. Low frequency noise limits. J Low Frequency Noise Vib 1992;11:7-13.|
|24||Schultz TJ. Synthesis of social surveys on noise annoyance. J Acoust Soc Am 1978;64:377-405.|
|25||Miedema HM, Vos H. Exposure-response relationships for transportation noise. J Acoust Soc Am 1998;104:3432-45.|
|26||Miedema HM, Oudshoorn CG. Annoyance from transportation noise: Relationships with exposure metrics DNL and DENL and their confidence intervals. Environ Health Perspect 2001;109:409-16.|
|27||Pedersen E, Persson Waye KP. Perception and annoyance due to wind turbine noise - a dose response relationship. J Acoust Soc Am 2004;116:3460-70.|
|28||Lim C, Kim J, Hong J, Lee S. The relationship between railway noise and community annoyance in Korea. J Acoust Soc Am 2006;120:2037-42.|
|29||Pawlaczyk-Luszczynska M, Dudarewicz A, Sliwinska-Kowalska M. Proposals of exposure criteria for the assessment of low frequency noise at workplaces in industrial control rooms and office-like areas. Arch Acoust 2007;32:303-12.|
|30||Weinstein ND. Individual differences in reactions to noise: A longitudinal study in a college dormitory. J Appl Psychol 1978;63:458-66.|
|31||ISO 9612. Acoustics - Guidelines for the measurement and assessment of exposure to noise in the working environments. Geneva: International Organization for Standardization; 1997. |
|32||ISO 6189. Acoustics - Pure tone air conduction threshold audiometry for hearing conservation purposes. Geneva: International Organization for Standardization:1983. |
|33||ISO 226. Acoustics - Normal equal-loudness contours. Geneva: International Organization for Standardization; 2003.|
|34||Moller H. Annoyance of audible infrasound. J Low Frequency Noise Vib 1987;6:1-7.|
|35||Kjellberg A, Lundstrom U. Noise in the office: Part II - The scientific basic (knowledgebase) for the guide. Int J Ind Ergon 1994;14:93-118.|
|36||Landstrom U, Kjellberg A, Bystrom M. Acceptable levels of sounds with the different spectra characteristics during the performance of a simple and a complex non-auditory task. J Sound Vib 1993;160:533-42.|
|37||Kryter KD, editor. The handbook of hearing and the effects of noise, physiology, psychology, and public health. San Diego: Academic Press;1994.|
|38||Belojevic G, Ohrstrom E, Rylander R. Effects of noise on mental performance with regard to subjective noise sensitivity. Int Arch Occup Environ Health 1992;64:293-301.|
|39||Poulsen T. Comparison of objective methods for assessment of annoyance of low frequency noise with the results of a laboratory listening test. J Low Frequency Noise Vib Active Cont 2003;22:117-31.|
|40||Subedi K, Yamaguchi H, Matsumoto Y, Ishihara M. Annoyance of low frequency tones and objective evaluation methods. J Low Frequency Noise Vib Active Cont 2005;24:81-96.|