| Article Access Statistics|
| Viewed||2901 |
| Printed||60 |
| Emailed||0 |
| PDF Downloaded||26 |
| Comments ||[Add] |
|Year : 2016
: 18 | Issue : 81 | Page
|Comparison of direct measurement methods for headset noise exposure in the workplace
Flora G Nassrallah1, Christian Giguere2, Hilmi R Dajani3, Nicolas N Ellaham4
1 Population Health Program, Faculty of Health Sciences, University of Ottawa, Ontario, Canada
2 School of Rehabilitation Sciences, Faculty of Health Sciences; School of Electrical Engineering and Computer Science, Faculty of Engineering, University of Ottawa, Ontario, Canada
3 School of Electrical Engineering and Computer Science, Faculty of Engineering, University of Ottawa, Ontario, Canada
4 School of Rehabilitation Sciences, Faculty of Health Sciences, University of Ottawa, Ontario, Canada
Click here for correspondence address
|Date of Web Publication||10-Mar-2016|
The measurement of noise exposure from communication headsets poses a methodological challenge. Although several standards describe methods for general noise measurements in occupational settings, these are not directly applicable to noise assessments under communication headsets. For measurements under occluded ears, specialized methods have been specified by the International Standards Organization (ISO 11904) such as the microphone in a real ear and manikin techniques. Simpler methods have also been proposed in some national standards such as the use of general purpose artificial ears and simulators in conjunction with single number corrections to convert measurements to the equivalent diffuse field. However, little is known about the measurement agreement between these various methods and the acoustic manikin technique. Twelve experts positioned circum-aural, supra-aural and insert communication headsets on four different measurement setups (Type 1, Type 2, Type 3.3 artificial ears, and acoustic manikin). Fit-refit measurements of four audio communication signals were taken under quiet laboratory conditions. Data were transformed into equivalent diffuse-field sound levels using third-octave procedures. Results indicate that the Type 1 artificial ear is not suited for the measurement of sound exposure under communication headsets, while Type 2 and Type 3.3 artificial ears are in good agreement with the acoustic manikin technique. Single number corrections were found to introduce a large measurement uncertainty, making the use of the third-octave transformation preferable.
Keywords: Acoustic manikin, acoustical standards, artificial ears, communication headsets, noise exposure
|How to cite this article:|
Nassrallah FG, Giguere C, Dajani HR, Ellaham NN. Comparison of direct measurement methods for headset noise exposure in the workplace. Noise Health 2016;18:62-77
|How to cite this URL:|
Nassrallah FG, Giguere C, Dajani HR, Ellaham NN. Comparison of direct measurement methods for headset noise exposure in the workplace. Noise Health [serial online] 2016 [cited 2018 Dec 19];18:62-77. Available from: http://www.noiseandhealth.org/text.asp?2016/18/81/62/178479
| Introduction|| |
Noise in the workplace is responsible for 16-24% of adult-onset hearing loss worldwide  and is the second most prevalent self-reported work-related injury.  In the United States, 22 million workers are exposed to potentially damaging noise levels,  and occupational noise-induced hearing loss affects 10 million workers.  In the European Union, 7.2% of workers report work-related hearing problems.  Among various factors, the increased use of wired and wireless communication headsets is raising concerns regarding exposure to potentially hazardous noise levels.  Communication headsets are commonly found in workplace settings such as call centers, retail stores, fast food outlets, airport ground and control tower operations, industrial and construction sites, and military sites. , They can be used, for example, to enhance the received communication signal in adverse noise  or to enable hands-free communication in less noisy environments.  While their use in occupational settings varies, in all cases the worker is exposed to the surrounding workplace noise as well as to the audio communication signals from the headset. Users typically adjust the volume setting of their communication headset to overcome the masking effects of the background noise entering the device in order to ensure proper reception of the audio signal such as speech. Depending on the situation, communication signals may occur continuously or intermittently; when present, they have been found to be a significant source of noise exposure. 
Given the increased use of communication headsets in the last decade, proper measurement tools and methods are required to assess noise exposure from these devices in the workplace. However, several methodological challenges arise when carrying out noise measurements with communication headsets. First, when a device is occluding the ear, measurements become dependent on the acousto-mechanical properties of the head, pinna, and ear canal  and in-ear recording techniques are typically required.  Second, both the audio signals generated by the headset and the external background noise passing through the headset contribute to the total noise exposure and must be accounted for. Third, the worker must be able to operate normally during the field recording period to achieve a valid assessment under realistic working conditions. Finally, since the assessment of noise exposure specified in occupational standards is based on free or diffuse sound field measurements, in-ear measurements must be converted to equivalent sound field exposure levels to enable a comparison with the regulatory exposure limit, for example, 85 dBA. ,
Despite these challenges, several methods have been proposed and used in the past forty years to assess noise exposure from the use of communication headsets in various workplaces. , Field studies have been conducted to evaluate noise exposure of radio operators, ,, military personnel,  call and communication center operators, ,,,,, as well as workers in others occupations. ,,, Results from these studies indicate that noise exposure from communication headsets often depends on the external background noise and could exceed regulatory limits in some workplaces or situations, especially in noisy environments. 
Several national and international standards (e.g., ANSI/ASA S12.19,  ISO 9612  ) describe methods for noise measurements in occupational settings using sound level meters and noise dosimeters. These standards assume that noise sources are not in close proximity to the ears; as such, they are not directly applicable for measurement of noise under communication headsets. Consequently, ISO 11904 , describes two specialized methods for the measurement of noise for sources close to the ears. The first method described in ISO 11904-1  defines a microphone in a real ear (MIRE) method where acoustic measurements are performed using miniature or probe microphones inserted in the ears of workers and then converted to equivalent free-field or diffuse-field sound levels. This method provides a most direct estimate of noise exposure and likely has the best face validity.  The second method described in ISO 11904-2  defines sound measurements taken with an acoustic manikin comprising an embedded ear simulator and microphone. The manikin's construction allows a simulation of the acousto-mechanical properties of the torso, head, pinna, and ear canal for an average adult. With either method, the in-ear unweighted sound pressure levels are analyzed in third-octave bands and converted to the free or diffuse sound field, then A-weighted using third-octave band attenuation factors. , The resulting A-weighted free or diffuse-field-related sound exposure level can then be compared to the regulatory limit. Of note, the manikin transfer function specified in ISO 11904-2 used to convert to free-field or diffuse-field related sound levels is corrected to yield the same results as the MIRE technique in ISO 11904-1, a mean value for a human population.
Special considerations must be given to field logistics since workers using headsets must be able to carry out their tasks normally while the measurements are being taken. Use of the MIRE technique can restrict head and body movements in some situations and as such it may be resisted by the workers when sustained for a long period of time. With the manikin technique, the issue is that the worker can no longer continue to communicate when the headset is placed on the manikin. ISO 11904-2  does not provide procedures to overcome this challenge. Several approaches have been devised such as duplicating the electrical signal to the headset and using two matched headsets, one worn by the worker and the other fitted on the manikin placed in close proximity to the worker. , Another approach is to record the electrical signal feeding the worker's headset in the field and either playing it back on the manikin in a laboratory setting or filtering the signal through the electrical-to-eardrum transfer function of the manikin. , Still, the manikin method is considered to be quite cumbersome to use in the field, and the instrumentation is not widely available.
In an effort to simplify the manikin measurements and analysis procedures, the Australian/New Zealand Standard AS/NZS 1269-1  specifies the use of a general-purpose artificial ear (IEC 60318-1,  a Type 1 artificial ear under ITU-T P.57  ) for headphones and the use of an ear simulator (IEC 60318-4,  a Type 2 artificial ear under ITU-T P.57  ) for insert earphones. Single number corrections that can be applied directly to the A-weighted measurements are also proposed for each artificial ear as well as for the manikin technique. , In addition, use of the Type 3.3 artificial ear (ITU-T P.57  ), a device combining a pinna simulator, the IEC 60318-4  ear simulator and a cheek-plate, offers another alternative for carrying out communication headset sound measurements.  The recent revision of Canadian standard CSA Z107.56,  specifies all these alternative methods and emphasizes field logistics while conducting headset sound measurements in the workplace. Since these methods are less expensive and less cumbersome to use, they are attractive when a compact setup is needed. However, while these alternative methods and provisions offer more practical and accessible options compared to using an acoustic manikin and/or carrying out third-octave band sound field conversions, as specified in ISO 11904-2,  their effectiveness has yet to be determined. Furthermore, procedures for fitting supra-aural earphones and circum-aural ear cups of various shapes and models on artificial ears have not been formalized and studied in the context of communication headset measurements.
Given the array of techniques proposed for noise exposure assessments under communication headsets, a recent study was conducted to evaluate the accessibility of these different measurement methods to relevant stakeholders.  A survey was distributed to occupational health and safety professionals and hearing loss prevention professionals in Canada. Results indicated that there is a wide range of expertise regarding noise measurement from communication headsets and that access to basic or specialized equipment varies greatly across the different types of professionals. Consequently, there is a need for different direct and indirect measurement methods tailored to different groups of professionals while considering their respective roles, expertise, and access to equipment. However, few studies have used more than one method to conduct measurements, and little is known about the degree of agreement between the various test measurement setups.
A pilot study was conducted to examine if results from different measurement setups, headset fitting methods, and conversion factors to relate measurements to the diffuse field were in agreement.  One expert positioned various types of headsets on four measurement setups (ISO 11904-2  manikin technique, and ITU-T P.57  Type 3.3, Type 2, and Type 1 artificial ears) while acoustic measurements were collected for six different audio signals from 12 fit/refits of various types of communication headsets. Results indicated that while the acoustic manikin and Type 3.3 and Type 2 artificial ears were in agreement, compatibility of the Type 1 artificial ear depended on the fitting method which greatly affected results. In addition, the difference between manikin measured levels and the equivalent diffuse-field levels was somewhat dependent on the headset and the noise type, which implied greater measurement uncertainty when using a single number conversion compared to using standardized ISO third-octave band corrections factors.
This present work expanded on the pilot study using multiple participants to compare the different measurement setups in order to provide a more comprehensive account of the different sources of measurement variability, including both within-subject and between-subject variability for various types of audio signals and communication headsets. The main goals were to investigate:
The measurement repeatability for each individual setup,
The measurement agreement between the different test setups, and
The increased measurement uncertainty associated with using single number corrections instead of third-octave band procedures to convert in-ear sound levels to the diffuse field.
| Methods|| |
Twelve individuals (6 males, 6 females), with technical and/or clinical expertise in the field of interest, participated in this study. Participants were asked to complete a questionnaire focusing on their professional expertise and employment history as well as on their experience fitting hearing devices and using sound measurement equipment. The participants had an average of 21.5 years of experience (range 1.5-48 years) in the fitting of earphones, hearing protectors and/or headsets on real ears, manikins, and/or artificial ear test measurement setups. Six participants were engineers or physicists and six others were audiologists. Nine of the 12 participants worked in research and/or consulting and three participants worked in a clinical setting. Ten of the 12 participants had experience working with a noise dosimeter and/or a sound level meter. All 12 participants had some experience working with an acoustic manikin, artificial ears, the MIRE technique, and/or a hearing aid analyzer and were expected to conduct noise exposure assessments related to communication headsets or hearing devices as a part of their scope of practice or employment.
Four different audio communication signals were used in this study as presented in [Table 1]. ICRA 1 and IEC 60268-1 are two artificial signals simulating the long-term spectral characteristics of speech and speech-music programming, respectively, and exemplify signal transmission over a noiseless communication channel. The two other audio signals consist of real speech mixed with real noises. The speech sentences were from the Hearing-In-Noise Test (HINT), a standardized test designed to evaluate speech recognition in noise. The two noise signals (coast guard plane and industrial riveter) are examples of background noises that can be picked up by the talker's microphone and transmitted over communication channels together with the talker's voice. The plane is a low-frequency continuous noise and the riveter is a high-frequency impact noise (26.5 impacts/s). Background noises and HINT sentences were mixed at a signal-to-noise ratio (SNR) of 0 dB to mimic transmission over a noisy communication channel. Tests were conducted to verify that the choice of SNR reflected challenging but realistic conditions of speech intelligibility. At an SNR of 0 dB, the extended speech intelligibility index  corresponding to the HINT/plane (HP) and HINT/riveter (HR) mixtures is 0.45 and 0.49, respectively, slightly above the limit for poor communication systems (ANSI S3.5  ). Informal listening of the two mixtures indicated word intelligibility around 90%.
The spectral characteristics of the four audio signals are shown in [Figure 1]. The four signals were concatenated into a sound file for presentation to the headsets, as shown in [Figure 2]. Each audio signal was 12 s long, separated by 3 s of silence. The analysis window consisted of the middle 10 s section of each signal. Tests with longer analysis windows of 20 and 30 s confirmed that a 10 s window was optimal to obtain accurate results (within 0.2 dB) while keeping the experimental session as short as possible (within 2 h). The four different audio signals were equalized to the same A-weighted root-mean-square amplitude on the sound file.
|Figure 2: Time waveform of the electrical signal of the complete sound excerpt used. The recording is preceded by a 1000 Hz initial pure tone to synchronize the data analysis. Dashed vertical lines delimit the sound measurement analysis window for each audio signal|
Click here to view
Four different headsets used in a wide range of occupational settings and applications were chosen for this experiment [Table 2]. The David Clark 40642G-01 (DVC) is a computer compatible circum-aural headset designed for high noise environments. The Plantronics HW261N (PLA) is a supra-aural telephone headset typically found in call centers. The 3M™ Peltor™ MT32H01 (PEL) is a lightweight single-sided nonattenuating supra-aural headset intended for military, police, and industry. Finally, the Sensear Smart Plug SP1 is an insert device designed for high noise environments such as industrial, commercial, or military settings. The Sensear SP1 is available with different ear tips; in this study, we tested the headset with silicon (SES) and foam (SEF) tips. The frequency response of these four headsets, measured on an acoustic manikin (KEMAR® Manikin GRAS Type 45BA), is shown in [Figure 3]. Responses are within 5 dB in the range of 500-2500 Hz across headsets, but much larger differences are found at higher and lower frequencies.
|Figure 3: Frequency response of the headsets measured on an acoustic manikin. The input is a pink noise electrical signal. The output from each device is scaled to an in-ear manikin level of 90 dBA for comparison purposes|
Click here to view
Four different measurement setups [Table 3], covering the array of techniques proposed in ISO 11904-2,  AS/NZS 1269-1,  and CSA Z107.56,  were used to measure the level of the different audio signals under the headsets [Figure 4].
|Figure 4: The different measurement setups used. (a) David Clark headset placed on the acoustic manikin. (b) SEF headset inserted in Type 2 artificial ear. (c) SEF headset inserted in Type 3.3 artifi cial ear. (d) Peltor™ MT32H01 headset held on Type 1 artificial ear; AE1.CH fitting method. (e) PLA headset placed on Type 1 artificial ear; AE1.CF fitting method. (f) David Clark headset placed on Type 1 artificial ear; AE1.MF fitting method. Tables 2-4 for headset, measurement setup and fitting descriptions|
Click here to view
|Table 3: Description of the measurement setups used. Figure 4 for illustrations|
Click here to view
The acoustic manikin (MAN) provides a full head, torso, pinna, and ear simulator IEC 60318-4  (formerly IEC 60711  or IEC 711 simulator) built according to worldwide averages of men and women dimensions. Placement of headsets on the manikin is applicable to all device types and is facilitated by the anthropomorphic features allowing an adjustment of all parameters relevant to fitting (e.g., depth of insertion, headband force, etc.). The pinnae come in two sizes ("small" and "large") and two levels of hardness (Shore-OO 55 and Shore-OO 35). The large size, typical of American and European males, was used in this study. The softer version (Shore-OO 35), recommended in IEC/TS 60318-7  and ITU Rec. P.57,  was used with all headsets. The harder version (Shore-OO 55) was also used with the supra-aural headsets (PLA, PEL) to determine the effects of pinna hardness on measured levels. These two adaptations of the manikin with large soft and large hard pinnae are referred to as MAN.SP and MAN.HP, respectively. All headsets (DVC, PLA, PEL, SES, and SEF) were tested on the right side of the manikin.
The Type 3.3 artificial ear (AE3.3) allows for testing with a flat cheek, large right soft pinna, and ear simulator (IEC 60318-4  ) providing a more compact test setup than the acoustic manikin. However, since it does not provide a full head size, special provisions must be applied for fitting the circum-aural (DVC) and supra-aural (PLA, PEL) headsets to ensure realistic headband pressure and placement over the simulated pinna. With the Type 3.3 resting on a table, the test ear cup or earpiece was placed on the artificial ear while the other side of the headset was placed under the table. A wooden block was fixed under the table to adjust the distance between both sides of the headset to be the same as the width of the head of the acoustic manikin (152 mm) in order to achieve an equivalent headband force. This fitting method, which simulates the natural headband force for an average human head, was found to provide the closest match to the acoustic manikin among different fitting options tested in the pilot study.  In the case of the insert headset (tips SES and SEF), fitting on Type 3.3 is the same as for the acoustic manikin.
The Type 2 artificial ear (AE2) comprises an ear simulator (IEC 60318-4  ) with external-ear extension. In this study, it was used to explore an even simpler setup than the Type 3.3 artificial ear, without cheek and pinna simulator, for use with insert headsets only. The ear simulator is the same as the one used with the acoustic manikin and the Type 3.3 artificial ear. However, due to the absence of a reference pinna with the Type 2, differences in insertion depth or placement may exist when fitting insert headsets in the external ear portion of Type 2 compared to the Type 3.3 or acoustic manikin. The AE2 setup was used for testing the insert headset with the two different tips (SES and SEF).
The Type 1 artificial ear (AE1), defined in IEC 60318-1  and IEC 60318-2,  contains an acoustic coupler with three connected cavities approximating the acoustical impedance of the human ear and different adaptors for the calibration of certain types of supra-aural and circum-aural audiometric earphones. Consequently, the Type 1 was used only with supra-aural (PEL, PLA) or circum-aural (DVC) headsets. As with the AE3.3 setup, care was exercised to provide realistic headband pressure and placement of these headsets. Headsets were fitted in different ways utilizing the conical adapter ring or a metal mounting plate adaptor supplied with the Type 1 artificial ear. The conical adaptor was used in conjunction with two different manners of applying pressure that could be used in the field: headset handheld over artificial ear (AE1.CH) and natural headband force (AE1.CF). The mounting plate adaptor was used only with the natural headband force fitting (AE1.MF). In all, three different fitting methods (AE1.CH, AE1.CF, and AE1.MF) were explored with the supra-aural headsets (PLA, PEL), and one fitting method (AE1.MF) was used for the circum-aural headset (DVC), as described in [Table 4].
|Table 4: Description of fitting methods on the Type 1 artificial ear for the supra-aural and circum-aural headsets|
Click here to view
All experiments were conducted under quiet conditions in laboratory settings. In soundproof facilities, participants were asked to position the headsets on four different test measurement setups and using applicable fitting methods, as described in [Table 5]. The experiment was approved by the Office of Research Ethics and Integrity of the University of Ottawa.
|Table 5: Combination of measurement setups and headsets tested with a number of fitting methods for each. Abbreviations for headsets and measurement devices are indicated in the brackets. There are three combinations of AE1 setups for the Plantronics and Peltor headsets for the three fitting methods used on the Type 1 artificial ear with supra-aural headsets, as per Table 4|
Click here to view
Once the headset was placed on a test setup, the recording of the four audio signals [Figure 2] was played from an iPod Touch player (Apple Inc., Cupertino, CA) connected to a Yamaha RX-A820 receiver/amplifier (Yamaha Corp., Hamamatsu, Japan). Prior to the experiment, the volume on the amplifier was adjusted and noted for each headset to produce an in-ear level of 90 dBA on the manikin with the ICRA1 noise, in order to simulate diffuse-field-related exposure levels of around 85 dBA. The volume of the amplifier was subsequently kept constant for a given headset for all the different test measurement setups and audio signals. The sound played through the headset was picked up by the microphone in the selected test measurement setup, then stored as a wave file (24 kHz, 16 bits) using a sound level meter (B&K 2250) with built-in audio recorder. All equipment was calibrated according to manufacturers' specifications, including any corrections supplied for the different ear simulators and calibration couplings accessories.
For each participant, measurements were taken twice in succession (fit and complete re-fit) for each of the 21 headset-measurement setup combinations in [Table 5], resulting in 42 fits in total. The subjects fitted the headsets on the measurement setup based on manufacturers' instructions and prior experience fitting similar listening devices. The testing order for the measurement setups was counter balanced across participants. Within each setup, participants tested all applicable headsets in random order.
Sound field transformation
The data recorded with the sound level meter with each measurement setup were transferred into Matlab (The Mathworks Inc., Natick, MA) for further analysis. For each test audio signal, the middle 10 s was extracted for analysis, and the equivalent level in each third-octave frequency band was computed and exported into Excel (Microsoft Corporation). The levels were then A-weighted using the third-octave band factors in IEC 61672-1  and transformed to equivalent diffuse-field levels. The third-octave diffuse-field correction factors specified in ISO 11904-2  were used for measurements obtained with the acoustic manikin, and the Type 3.3 and Type 2 artificial ears, as specified in CSA Z107.56  since they share the same ear simulator (IEC 60318-4  ). Measurements with the Type 1 artificial ear were transformed using the wide-band artificial ear to eardrum transfer function proposed by Macrae  followed by the ISO 11904-2  diffuse-field transformation. The latter transformation published in 2004, instead of the original transmission gain proposed by Macrae  in 1995, was used in this study to ensure that the same eardrum to diffuse-field transformation is employed for all test measurement setups. The difference between the two transformations ranges from −1.2 to +1.0 dB across the set of third-octave bands from 200 to 5000 Hz.
In method comparison studies, two particular questions are relevant: 
The statistical properties of each individual method (e.g., variability within and between observers); and
The degree of agreement between methods (e.g., average bias between two methods being compared, range of differences between the two methods for single measurements carried out by the same observers).
While common statistical techniques have been used in the context of method comparison studies, such as comparison of means, and correlation or regression analysis, they are not appropriate or designed to answer such questions.  In this study, we used the Bland-Altman limits of agreement (LoA) approach  to compare the different methods of sound measurement from communication headsets. This approach, commonly used in clinical measurements, is particularly well suited when comparing a new or simplified method to an established method. 
Properties of each method
The A-weighted diffuse-field transformed levels for each participant and trial were used to compute descriptive statistics. For each measurement method, the mean of the two trials (fit-refit) was taken and averaged across the 12 participants (X-0) for each headset and audio signal combination. The between-subject standard deviation (sB ) of the participant fit-refit means was also computed to report variability across participants in each experimental condition. In addition, the within-subject standard deviation (sω) (i.e., the standard deviation of repeated measurements by the same participant) was calculated by averaging the variance of the two trials of the 12 participants and taking the square root.  The difference between a single measurement and the true value for a given method is expected to be less than 1.96 sω in 95% of cases. Another way of describing the measurement variability for each method is the repeatability coefficient , defined as:
This coefficient is useful to quantify how repeatable a method is upon successive measurements; two measurements obtained with the same method under identical experimental conditions are expected to vary by no more than the repeatability coefficient in 95% of cases.
Measurement agreement between methods
Bland-Altman's LoA approach provides a means of quantifying the 95% range of differences between single measurements performed by two different methods as follows: 
where d− is the bias or difference between the method means X-, and is the estimated variance of the between-subject differences by each method. The latter was calculated from equation 5.3 in Bland and Altman: 
where is the observed variance of the differences between the within-subject means of the two methods being compared, and s2xω and s2yω are the within-subject variances of the two replicated trials (fit-refit) for each method (x and y). In this study, the acoustic manikin with soft pinna (MAN.SP) serves as the gold standard or reference method for communication headsets sound measurements from which all other methods are compared. There are 16 possible comparisons between the AE1, AE2, AE3.3, or MAN.HP setups and the reference method MAN.SP over the different measurement setup-headset combinations tested in this study [Table 5].
One-way repeated measures ANOVAs were conducted in SPSS (IBM Corp., Armonk, NY) to determine the effect of the four audio signals (ICRA1, IEC, HP, and HR) on the bias level d− between each measurement setup under test and the acoustic manikin reference method. For each headset, this analysis was used to determine whether the LoA between methods could be combined over all four audio signals or computed separately for each different signal.
Validity of single number corrections
The difference between the A-weighted in-ear sound level (measured in test setup) and the A-weighted diffuse-field equivalent sound level (computed using the applicable third-octave band transformation) was calculated for each measurement setup, headset, and audio signal. The in-ear/diffuse-field difference level was then compared to the 8 dB single number correction proposed for the Type 1 artificial ear or to the 5 dB single number correction proposed for the acoustic manikin, Type 3.3, and Type 2 artificial ears. This procedure allowed us to quantify the increased measurement uncertainty, if any, introduced by the simplified single number transformation specified in CSA Z107.56  and AS/NZS 1269.1.  Given that the MAN, AE3.3 and AE2 setups use the same ear simulator (IEC 60318-4  ) and diffuse-field transformation function (ISO 11904-2  ), tests were carried out to determine whether the in-ear/diffuse-field difference levels from these three methods were not statistically different and could be combined. For the supra-aural PLA and PEL headsets, which were tested with the MAN and the AE3.3 setup, a dependent t-test was conducted to compare the two measurement setups for each audio signal. Similarly, for the two ear tips (SES, SEF) of the insert headset, which were tested on the MAN, AE3.3, and AE2 setups one-way repeated measures ANOVAs were conducted for the same purpose.
| Results|| |
[Figure 5] displays the A-weighted measured in-ear levels with the reference MAN.SP setup and the diffuse-field related levels computed using the third-octave conversion procedure in ISO 11904-2,  for all headsets and audio signals. The data are averaged over the 12 participants but shown separately for each trial (fit-refit). The in-ear manikin level is very close to 90 dBA for the ICRA1 signal for all headsets, as expected from the experimental protocol which used this audio signal for level adjustment purposes. For each headset, some differences are seen for the other signals due to their different spectra [Figure 1] compared to ICRA1. This is most noticeable for the IEC and HP signals which are richer in low-frequency content. Overall, there is very little difference in the mean level between the two trials for both the in-ear and the diffuse-field levels; however, the diffuse-field-related levels are lower than the measured in-ear levels by an amount ranging from about 2 to 11 dB across headsets and signals.
|Figure 5: Mean level and standard deviation of measurements taken using the reference acoustic manikin (MAN.SP) with each headset and audio signals (n = 12 participants) for both fits. Measured in-ear levels and diffuse-field related levels obtained with the third-octave correction factors specifi ed in ISO 11904-2 are shown|
Click here to view
Properties of each method
The descriptive statistics for the 21 headset-measurement setup combinations and four audio signals are summarized in [Table 6]. For each condition, the mean A-weighted diffuse-field equivalent levels (X-0), the between-subject standard deviation (sB ), and the within-subject standard deviation (sω) are presented. For the reference method (MAN.SP), the between-subject standard deviation (sB ) varied from 0.5 to 1.1 dB and the within-subject standard deviation (sω) varied from 0.3 to 1.2 dB. Results for the MAN.HP setup were very similar to MAN.SP for the two supra-aural devices tested. Results for the AE3.3 and AE2 setups were also in close agreement with MAN.SP and MAN.HP, except for the higher between-subject standard deviation (sB ) with the AE3.3 in the case of the supra-aural PEL headset (1.2-1.9 dB) and the notably smaller within-subject standard deviation ( sω) with the AE2 (0.1-0.4 dB) for both ear tips of the insert headset (SES and SEF).
|Table 6: Mean equivalent diffuse-field levels (X̅ ) in dBA, between-subject standard deviation (sB) in dB, and within-subject standard deviation (sω) in dB for the 21 headset-measurement setup combinations and the four audio signals. The repeatability coefficient is 2.77 sω|
Click here to view
The AE1 setup produced erratic results [Table 6], with descriptive statistics highly dependent on the fitting method [Table 4] and headset tested. The AE1.CH setup yielded the largest between-subject standard deviation ( sB ) from 3.0 to 6.7 dB among all measurement setups tested, and a within-subject standard deviation ( sω) often two or three times larger than the MAN and AE3.3 setups for the same headsets. For the AE1.CF setup, the within-subject standard deviation ( sω) is comparable to the MAN and AE3.3 setups, but the between-subject standard deviation ( sB ) with the supra-aural PEL headset are about three times higher than the MAN setups and almost twice higher than the AE3.3 setup. In the case of the AE1.MF setup, results were highly dependent on the headset; very low within-subject ( sω) (0.1-0.2 dB) and between-subject ( sB ) (0.2 dB) standard deviations were observed with the circum-aural DVC headset, while large between-subject ( sB ) (2.3-2.9 dB) and within-subject (sω) (3.5-4.6 dB) standard deviations were found for the supra-aural PLA headset.
Measurement agreement between methods
One-way repeated measures ANOVAs revealed that bias d− between the AE1, AE2, AE3.3, or MAN.HP setups and the reference acoustic manikin (MAN.SP) differed significantly (P < 0.05) between audio signals (ICRA1, IEC, HP, and HR) in 12 of the 16 possible setup-headset comparisons. Consequently, Bland-Altman's LoA analyses were conducted considering audio signal dependence in all method comparisons. [Figure 6] displays the LoA between the different measurement setups and the reference acoustic manikin (MAN.SP) for each headset and audio signal. These limits quantify the range within which 95% of the difference between methods will lie for single measurements. As shown in [Figure 6], the MAN.HP, AE3.3, and AE2 setups produced LoAs with much smaller bias and variability components than the AE1 setups.
|Figure 6: Bland-Altman limits of agreement comparing each measurement setup to the reference acoustic manikin (MAN.SP) for the 16 possible setup-headset comparisons and four audio signals. The limits of agreement comprise a bias (symbol) and a variability component (error bar) indicating the 95% interval of the difference between single measurements on the two setups. A negative bias indicates a lower diffuse-field related level with the measurement setup under test than the reference acoustic manikin|
Click here to view
For the MAN.HP, which was tested with the two supra-aural headsets, bias was slightly negative ranging from −0.3 to −0.6 dB for the PLA headset and from −0.1 to 0.0 dB for the PEL headset across the four audio signals, while variability was from 1.0 to 2.4 dB across these test conditions. For the AE3.3, bias was either slightly negative or slightly positive for the DVC, PLA, SES, and SEF headsets, ranging from −0.8 to +0.7 dB across headsets and audio signals, while variability was from 0.8 to 2.9 dB. For the PEL headset, a larger positive bias was observed from +1.1 to +1.5 dB together with an increased variability from 2.5 to 3.7 dB across audio signals. For the AE2, which was tested with the two ear tips of the insert headset, bias was comparable to that with the MAN.HP setup, from +0.1 to +0.6 dB across audio signals, and variability was consistent across audio signals at about 2.0 dB. Overall, measurements taken with the MAN.HP, AE3.3, and AE2 produced 95% LoA always overlapping with, and nearly centered on, the zero-difference value, with relatively small bias and variability components indicating a good agreement with the reference acoustic manikin (MAN.SP).
The AE1, which was tested with circum-aural and supra-aural headsets, produced much larger bias and variability components that were highly dependent on the fitting methods [Table 4] and headsets, as shown in [Figure 6]. For the AE1.CH setup, bias was similar across headsets from +8.6 to +12.5 dB while the variability was largest with the PEL headset from 10.5 to 12.6 dB, across audio signals. For the AE1.CF, bias was also similar across headsets from +2.5 to +5.2 dB, and the variability was again largest with the PEL headset from 5.1 to 6.7 dB, across audio signals. The AE1.MF setup produced noticeably different results for the two supra-aural headsets. For the PEL headset, bias ranged from +1.3 to +2.3 dB and the variability was about 4.2 dB, across audio signals. For the PLA headset, this setup produced the highest bias from +13.0 to +17.3 dB with a variability ranging from 6.7 to 8.6 dB, across audio signals. In addition to the very large bias and variability observed for the AE1 setup, in many cases the 95% LoA also did not even overlap with the zero-difference value (e.g., AE1.MF with the DVC and PLA, AE1.CH with the PLA and for one audio signal with the PEL, AE1.CF with the PLA). These results indicate a very poor agreement with the reference acoustic manikin (MAN.SP) for all fitting options of the AE1 setup.
Validity of single number corrections
The dependent t-tests and one-way repeated measures ANOVAs conducted on the A-weighted in-ear/diffuse-field difference levels for the MAN.SP, AE3.3, and AE2 setups revealed a mixture of significant (P < 0.05) and insignificant differences across audio signals, headsets, and setups. The difference between measurement setups (range: 0.0 to 0.6 dB) was much smaller than the difference between headsets (range: 0.0 to 5.2 dB) and audio signals (range: 0.0 to 5.5 dB). Consequently, the data were pooled over the MAN.SP, AE3.3, and AE2 setups to highlight the A-weighted in-ear/diffuse-field difference levels overall audio signals and headsets when the third-octave conversion procedure is used as per ISO 11904-2.  These results are shown in [Figure 7]a. The mean A-weighted in-ear/diffuse-field difference ranged from 2.2 dB (SES headset with HP audio signal) to 11.1 dB (PLA headsets with the HR audio signal). In 14 of the 20 differences across audio signals and headsets, the A-weighted in-ear/diffuse-field difference was higher than the proposed 5 dB single number correction proposed in AS/NZS 1269-1  and CSA Z107.56.  The IEC and HR signals produced larger differences than the ICRA1 and HP audio signals. Standard deviations were relatively stable from 0.6 to 1.7 dB across audio signals and headsets.
|Figure 7: (a) Mean A-weighted in-ear/diffuse-field differences for the pooled MAN.SP, AE3.3, AE2 measurement levels using the third-octave correction factors specified in ISO 11904-2 for all audio signals and headsets. The 5 dB single number correction is shown as a dashed line. The between-subject standard deviation of the level differences is also shown. (b) Mean A-weighted inear/ diffuse-fi eld differences for the AE1.MF, AE1.CH, and AE1. CF setups using the third-octave correction factors based on Macrae and ISO 11904-2 for all audio signals and applicable headsets. The 8 dB single number correction is shown as a dashed line. The between-subject standard deviation of the level differences is also shown|
Click here to view
Mean A-weighted in-ear/diffuse-field differences for the AE1.CH, AE1.CF, and AE1.MF setups are shown in [Figure 7]b. The difference ranged from 5.3 dB (PLA headset on AE1.MF with IEC audio signal) to 8.0 dB (DVC headset on AE1.MF with HP audio signal). In all 28 differences across audio signals and headsets, the A-weighted in-ear/diffuse-field difference was smaller or equal to the 8 dB single number correction. The HP and HR signals produced slightly larger differences than the ICRA1 and IEC audio signals. Standard deviations were not consistent, ranging from 0.3 to 9.5 dB across audio signals and headsets; they were notably smaller for the DVC headset from 0.3 to 0.4 dB and larger for the PEL headset from 7.5 to 9.5 dB.
| Discussion|| |
Due to complicated field logistics and data transformation steps, specialized measurement methods, such as the use of MIRE and acoustic manikin techniques, can be technically challenging and are not often used by occupational health and safety professionals and other stakeholders in hearing loss prevention to assess sound exposure from communication headsets. Standard ISO 11904-2,  for example, describes the use of an acoustic manikin measurement setup with full head and torso simulator together with third-octave band sound conversion procedures to relate in-ear manikin measurements to the equivalent free or diffuse sound field. More compact measurement setups and simpler conversion procedures have been proposed in some standards (CSA Z107.56  and AS/NZS 1269-1  ) and were evaluated in this study. Measurements on four test setups (manikin and Type 3.3, Type 2, and Type 1 artificial ears) with different methods of headset fitting were carried out in conjunction with third-octave band and single number conversion procedures. In all, data were collected over 21 different headset-measurement setup combinations and four different audio signals. Twelve participants experienced in the placement of hearing devices on real ears and/or artificial fixtures fitted the communication headsets twice on each measurement setup.
The KEMAR acoustic manikin (GRAS Type 45BA) with soft pinna (MAN.SP) served as the gold standard or reference method for testing the other measurement setups. The manikin and its variants are recognized for anthropomorphic testing in the fields of hearing conservation, telecommunication, and noise abatement, as well as sound recording and sound quality. In compliance with the requirements of ITU-T P.58,  the manikin is built based on worldwide averages of human head and torso dimensions in order to provide acoustic diffraction and pick-up characteristics resembling those encountered by the median human head and torso. The manikin is designed to be used with one or two installed ear simulators IEC 60318-4  and two artificial pinnae. It can accommodate all types of headsets and is compatible with ISO 11904-2  for acoustic measurements from sources close to the ears.
Properties of each method
Prior to comparisons against the reference acoustic manikin, the statistical properties of each individual method were assessed to determine the various sources of measurement variability. The results, described in [Table 6], revealed that the acoustic manikin with soft (MAN.SP) and hard (MAN.HP) pinnae, and the AE3.3 and AE2 setups, all produced relatively low levels of measurement variability characterized by between-subject (sB ) and within-subject (sω) standard deviations of at most 1.9 dB and 1.2 dB, respectively, over the different headsets and audio signals. In contrast, the AE1 setup produced measurement variability which was highly dependent on the fitting method used and headset tested, with between-subject (sB ) and within-subject (sω) standard deviations as high as 6.7 dB and 4.6 dB, respectively, in some conditions.
Of note, Annex B in ISO 11904-2  reports a standard uncertainty of 0.5 dB for the repeatability of the mean of three fits and measurements with supra-aural earphones on the acoustic manikin. Using the within-subject standard deviation data from this study, a comparable estimate (sω/√3) is obtained ranging from 0.2 to 0.7 dB over the two supra-aural devices (PLA, PEL) and four audio signals tested. [Table 6] may thus prove useful in documenting between-subject and within-subject measurement variability associated with the different test setups for the purpose of estimating measurement uncertainty for specific testing conditions in the field. Other sources of uncertainty to consider include the microphone calibration and sound level meter uncertainty, the deviation of the test fixture from human subjects, the fluctuations in the test signal level, the variations in environmental conditions, and the rounding error. 
Agreement between methods
Bland-Altman's LoA was computed for the 16 possible comparisons between the reference acoustic manikin (MAN.SP) and the other measurement setups. Since one-way repeated measures ANOVAs revealed an effect of the audio signal type on the bias component, Bland-Altman analyses were conducted independently for each audio signal and reported in [Figure 6].
The AE1 method provided by far the largest bias among all test setups under comparison against the reference acoustic manikin [Figure 6]. In addition, a much larger variability component for the LoA was observed with the AE1. Such a poor agreement of the AE1 with the reference acoustic manikin prompted more investigation into the third-octave diffuse sound field conversion procedure used, which was based on the work of Macrae.  As an alternative, use of the free-field transformation specified in Annex A of ANSI/ASA S3.6  was investigated. In this standard, the free-field correction is based on the difference between the free-field equivalent sensitivity level and the coupler sensitivity level for two types of supra-aural audiometric earphones (Telephonics TDH 39, Telephonics TDH 49/50) and for a circum-aural headset (Sennheiser HDA 200). The average of the TDH 39 and TDH 49/50 free-field correction values was applied to measurement data obtained with the PEL and PLA supra-aural headsets on the AE1 setup. Similarly, the Sennheiser free-field equivalent correction was applied to the measurements taken with the DVC circum-aural headset. These new A-weighted free-field-related estimates with the AE1 were then compared to the reference acoustic manikin (MAN.SP) measurements corrected for the free-field transformation specified in ISO 11904-2.  While the variability component of the LoA hardly changed for the PLA, PEL, and DVC, bias increased for the PLA and PEL headsets when using the AE1 with the transformation specified in ANSI/ASA S3.6  instead of the original transformation based on Macrae.  Bias was slightly closer to zero for the DVC headset when using the ANSI/ASA S3.6  transformation specific for the Sennheiser audiometric headset. Overall, however, the results obtained with the AE1 were largely different from the reference acoustic manikin method using either sound field transformation procedure and provided much larger LoA for both bias and variability components than all other measurement methods under comparison against the reference acoustic manikin. This is not surprising since the transformation proposed in ANSI/ASA S3.6  and the wide-band artificial ear to eardrum transfer function implicit in the method developed by Macrae  are very specific to certain types of audiometric earphones and cushion characteristics and not to general-purpose headsets such as the PLA, PEL, and DVC used in this study. Given the lack of agreement between the AE1 setup and the reference acoustic manikin (MAN.SP), as demonstrated by the overly large LoA [Figure 6], and the large between-subject and within-subject measurement variability obtained with the AE1 [Table 6], our results imply the non-suitability of the Type 1 artificial ear for sound measurements under communication headsets.
The AE3.3 and AE2 setups were also explored in this study and compared to the reference acoustic manikin [Figure 6]. Bias for these two measurement devices was small (typically <1 dB) and in most cases negligible. The variability component of the LoA also proved to be relatively small (about 2-3 dB). Since the AE3.3 and AE2 measurement setups use the same ear simulator (IEC 60318-4  ) as the reference manikin (MAN.SP), the third-octave diffuse field correction specified in ISO 11904-2  was used for all these setups, as specified in CSA Z107.56.  The good agreement between the AE3.3 or AE2 setups and the reference acoustic manikin were therefore not surprising and confirmed their usability in the context of sound exposure measurement under communication headsets.
Finally, in addition to the use of the reference acoustic manikin with the soft pinna (MAN.SP), measurements were also taken on the acoustic manikin with a harder pinna (MAN.HP) to determine the effects of pinna hardness on measured levels for the supra-aural headsets (PLA, PEL). The variability component of the LoA was small and similar with both headsets (about −0.6-0.0 dB); however, the bias was slightly negative for measurements taken with the PLA headset while it was essentially nil for the PEL headset [Figure 6]. The markedly more compliant headband of the PLA headset produced less pressure on the pinna than the stiffer PEL headset and therefore may have contributed to a small placement effect when fitted on the hard versus softer pinna. The mean negative bias introduced by the hard pinna on the PLA headset was however <0.5 dB over the four audio signals. Nevertheless, the use of the softer pinna (Shore-OO 35) is recommended in IEC/TS 60318-7  and ITU Rec. P.57  for acoustic manikin measurements.
Single number corrections
As an alternative to frequency-dependent transformations, diffuse-field levels can be estimated using single number corrections of 8 dB for the Type 1 artificial ear or 5 dB for the acoustic manikin and Type 3.3 and Type 2 artificial ears, applied directly to the A-weighted in-ear measurements, as proposed in some national standards (AS/NZS 1269-1,  CSA Z107.56  ). However, the increased measurement uncertainty associated with this simplified transformation has not been established.
For the results pooled over the MAN.SP, AE3.3, and AE2 setups [Figure 7]a], the mean A-weighted in-ear/diffuse-field difference levels ranged from 2.2 to 11.1 dB across headsets and audio signals. As such, the corresponding deviation from the proposed 5 dB single number correction ranged from -2.8 to +6.1 dB. As shown in [Figure 7]a the IEC and the HR audio signals, both richer in high frequencies [Figure 1], provided higher A-weighted in-ear/diffuse-field difference levels than the two other audio signals. Since more weight is given to the higher frequencies with the third-octave transformation, such results are expected. Likewise, the higher difference levels shown for the two supra-aural headsets (PLA, PEL) in [Figure 7]a are related to the poorer low frequency response of these headsets compared to the other headsets [Figure 3], which again increased the importance of the high frequency acoustic energy when converting the measured in-ear levels to the diffuse sound field.
For the AE1 setup, the mean A-weighted in-ear/diffuse-field difference levels ranged from 5.3 to 8.0 dB across headsets, audio signals, and fitting methods [Figure 7]b; thus, the corresponding deviation from the proposed 8 dB single number correction ranged from +0.0 to +2.4 dB. While the deviation from the proposed single number correction appears smaller for the AE1 setup than for the MAN.SP, AE3.3, and AE2 setups, the between-subject standard deviation of the A-weighted in-ear/diffuse-field difference levels is much larger for the AE1 than for the other setups, up to 9.5 dB in some conditions [Figure 7]b]. The latter may indicate that fitting differences among participants introduced a strong frequency-dependent variability in the transmission of audio signals to the AE1 microphone.
In all, while simpler to use, single number correction values are by definition frequency-independent and therefore cannot account for the spectral differences in the different audio signals that can be received through communication headsets, nor the frequency response of different headsets that can be used. The difference between measured levels and the equivalent diffuse-field levels was expectedly dependent on the frequency responses of the headset and the spectra of the audio signal, in accordance with previous work.  Even though the 5 dB and 8 dB single number corrections are covered in the range of level differences [Figure 7], a large measurement uncertainty is introduced when using a single number conversion instead of frequency-dependent correction factors such as provided in ISO 11904-2. 
Implications for field use
As discussed above, owing to the poor agreement with the reference acoustic manikin [Figure 6] and the large measurement variability [Table 6], use of a Type 1 artificial ear is not warranted for the measurement of the sound exposure from communication headsets at any stage of an assessment. In contrast, Type 2 and Type 3.3 artificial ears are in good agreement with the reference acoustic manikin and produce comparable between-subject and within-subject measurement variability. As such, these two measurement setups offer an alternative to the full acoustic manikin, especially in field situations or environments where a more compact setup is needed or desired. The Type 2 artificial ear was tested with one insert headset and found suitable with both foam and silicone ear tips. The Type 3.3 artificial ear was tested with one circum-aural, two supra-aural, and one insert headsets (foam and silicone tips).
In [Figure 6], the LoA shown for the different measurement setups were derived considering audio signal as a factor with fixed effect, and therefore they strictly apply to the specific audio signals under study. While the four signals chosen for this study exemplify the acoustical characteristics of different audio signals transmitted over noiseless and noisy communication channels, they do not cover the wide range of possible existing signals in a workplace setting. In practice, the exact nature of the audio signal may differ from the ones chosen in this study, or be a priori unknown. In order to provide more general LoA to guide field use, the Bland and Altman  approach was extended assuming that the four audio signals used represented a random sample (with a random effect) from all possible audio signals. [Table 7] summarizes the LoA generalized over audio signals for the measurement setups (AE2, AE3.3, and MAN.HP) that are deemed to be suitable alternatives to the reference acoustic manikin (MAN.SP). These limits could serve as a guide for individuals who are faced with taking measurements under communication headsets in the field. The bias term is not significantly different from a null bias (0 dB) in 5 of the 9 comparisons and is within ±0.5 dB in all cases except for the +1.3 dB bias found for the supra-aural PEL headset with the AE3.3. The variability component of the LoA varies from 1.2 to 3.4 dB across setups and headsets.
|Table 7: Generalized limits of agreement (dB) for the Type 3.3 artificial ear, Type 2 artifi cial ear and acoustic manikin with|
hard pinna, with audio signals represented as a factor with random effects. The acoustic manikin with soft pinna is the
reference method. The P value measures the significance of the bias term using a Chi-square analysis
Click here to view
It is important to note that the LoA results in [Table 7] express the 95% range of differences for single measurements performed on the two measurement setups being compared. In practice, multiple measurements may be taken and averaged in the field. This will reduce the range of the expected difference between the chosen measurement setup (AE3.3, AE2, and MAN.HP) and the reference acoustic manikin method (MAN.SP). With an increasingly large number of replicated measurements and subjects, the LoA gradually reduces to the residual bias term in [Table 7].
The generalized LoAs in [Table 7] are computed assuming one-third octave band factors as per ISO 11904-2  to relate in-ear measurements to diffuse-field sound levels. A large increase in measurement uncertainty was found when using a frequency-independent single number correction, owing to differences in audio signal spectra and headset frequency responses. For the Type 2 and Type 3.3 artificial ears and acoustic manikin, the mean A-weighted diffuse-field levels derived by use of one-third octave band factors and single number correction differed from each other over a range from −2.8 to 6.1 dB across the headsets and audio signals used in this study [Figure 7]a]. This, in effect, amounts to a large measurement bias that depends on the specific headset and audio signal tested, and which cannot be controlled through measurement averaging. From a practical standpoint, the use of a single number correction to approximate the diffuse-field transfer function of the human ear thus appears to introduce an unacceptably large source of measurement error.
Finally, this study focussed only on methods to assess the sound exposure from the built-in audio channel of the communication headsets. In noisy work environments, the surrounding background noise may also reach the worker and contribute to the overall exposure, making this pathway dependent on the noise level, and the amount of attenuation provided by the headset. Artificial ears are not well suited to estimate this exposure component due to possible sound transmission flanking pathways arising from the low acoustic isolation of these setups that are not designed for sound attenuation measurements, as cautioned in CSA Z107.56.  An alternative is to use standardized methods applicable to hearing protectors (e.g., ISO 4869-2  ) to estimate the effective A-weighted sound level of the background noise. The overall sound exposure level when headsets are used is then computed from the energy sum of the audio signal and background noise components, both weighted by their relative duration in a typical 8 h workday.
| Conclusion|| |
This research compared different measurement tools described in national and ISO for the assessment of noise exposure under communication headsets. Measurements on four test setups (manikin and Type 3.3, Type 2, and Type 1 artificial ears) with different methods of headset fitting were carried out in conjunction with third-octave band and single number conversion procedures.
Results indicated the following:
The Type 1 artificial ear is not suited for sound measurements under communication headsets given the poor agreement between this test setup and the acoustic manikin technique specified in ISO 11904-2  and the poor measurement repeatability;
The Type 2 and Type 3.3 artificial ears yielded a good agreement with the acoustic manikin technique and provided comparable measurement repeatability, making them suitable alternatives for sound measurements under communication headsets when compact test setups are needed; and
The use of single number corrections was found to introduce a large measurement uncertainty as they do not account for the spectral difference in the audio signals nor the frequency response of the headsets. Use of the third-octave transformation as per ISO 11904-2  is preferred.
All in all, the results indicated that only measurement setups (Type 2, Type 3.3) based on ear simulator IEC 60318-4  achieved a good agreement with the ISO 11904-2  manikin technique. While widely available, the Type 1 artificial ear used for audiometric earphone calibration was not found to be suited to sound exposure measurements from the audio channel of communication headsets, and one has to look elsewhere for a practical survey method accessible to a wide range of professionals. The Canadian standard CSA Z107.56  , for example, also specifies a simple calculation method requiring only a sound level meter or noise dosimeter and computational steps relating overall sound exposure to the external background noise, the noise reduction of the device, and the effective listening SNR. However, more research is needed to fully characterize the effectiveness of this indirect method compared to established direct measurement methods such as the manikin and/or MIRE techniques.
Financial support and sponsorship
This project was funded by a research grant provided by the Workplace Safety and Insurance Board (Ontario, Canada).
Conflicts of interest
There are no conflicts of interest.
| References|| |
Nelson DI, Nelson RY, Concha-Barrientos M, Fingerhut M. The global burden of occupational noise-induced hearing loss. Am J Ind Med 2005;48:446-58.
National Institute for Occupational Safety and Health Education and Information Division. NIOSH′s Hearing Loss Prevention Unit Featured at Health and Safety Conference. Atlanta, Georgia, USA: Center for Disease Control and Prevention; May, 2007. Available from: http://www.cdc.gov/niosh/updates/upd-05-31-07.html
. [Last accessed on 2016 Feb 19]
Tak S, Davis RR, Calvert GM. Exposure to hazardous workplace noise and use of hearing protection devices among US workers - NHANES, 1999-2004. Am J Ind Med 2009;52:358-71.
Parent-Thirion A, Macías EF, Hurley J, Vermeylen G. Fourth European Working Conditions Survey. Dublin: European Foundation for the Improvement of Living and Working Conditions; 2007.
Peretti A, Pedrielli F, Baiamonte M, Mauli F, Farina A. Headphone Noise: Occupational Noise Exposure Assessment for Communication Personnel. Proceedings of Euronoise, 19-21 May, 2003. Naples, Italy: Verlag; 2003.
Giguère C, Behar A, Dajani HD, Kelsall T, Keith SE. Direct and indirect methods for measurement of occupational sound exposure from communication headsets. Noise Control Eng J 2012;60:630-44.
Nassrallah F, Giguère C, Dajani HR. Measurement Methods of Noise Exposure from Communication Headsets Used in Various Occupational Settings. Proceedings of the 11 th
International Congress on Noise as a Public Health Problem; 1-5 June, 2014, Nara, Japan; 2014.
Dajani H, Kunov H, Seshagiri B. Real-time method for the measurement of noise exposure from communication headsets. Appl Acoust 1996;49:209-24.
Williams W, Presbury J. Observations of noise exposure through the use of headphones by radio announcers. Noise Health 2003;5:69-73.
Barron and Associates Consulting Acoustical Engineers. A Study of Workplace and Headset Noise Exposure of Radio Operators; Project No. 194.781; 1979.
Barron and Associates Consulting Acoustical Engineers. A Study of Individual Radio Operator Noise Exposure at Gander Flight Service Station, Halifax Coast Guard Marine Radio, Vancouver Coast Guard Marine Radio. Report prepared for Transport Canada. Project No. 194.8002; 1981.
Forshaw SE, Cruchley JI, Crabtree RB, Mendlowitz SL, Elcombe DD. A study of hearing loss among high- and medium- frequency radio operators. Downsview, Canada: Department of National Defence, Defence and Civil Institute of Environmental Medicine, Behavioural Sciences Division, Report No. 82-R-35; 1982.
Crabtree RB. Hercules Audio Enhancement: Project Report. Toronto, Canada: Department of National Defence, Defence Research and Development; 2002.
Chiusano SV, Lees PS, Breysse PN. An occupational noise exposure assessment for headset-wearing communications workers. Appl Occup Environ Hyg 1995;10:476-81.
Patel JA, Broughton K. Assessment of the noise exposure of call centre operators. Ann Occup Hyg 2002;46:653-61.
Planeau V. Noise hazards associated with the call centre industry. Vandoeuvre, France: Institut National de Recherche et de Sécurité (INRS); 2005.
Smagowska B. Noise at workplaces in the call center. Arch Acoust 2010;35:253-64.
Trompette N, Chatillon J. Survey of noise exposure and background noise in call centers using headphones. J Occup Environ Hyg 2012;9:381-6.
Kunov H, Giguère C, Simpson R. Method for measuring noise exposure from communications headsets. Final Report to Labour Canada. Toronto, Canada: University of Toronto, Institute of Biomedical Engineering, Report No. IBME-89/08 (HK-HEADSET); 1989.
Achutan C, Kardous CA. Evaluation of potential noise hazards to mechanics and 911 dispatchers at a fire. Health Hazard Evaluation Report. Anchorage, AK, USA: Department of Health and Human Services, Center for Disease Control and Prevention, National Institute for Occupational Safety and Health, Report No. HETA 2007-0235-3064; 2008.
ANSI/ASA. Measurement of Occupational Noise Exposure. Washington, DC, USA: American National Standards Institute/Acoustical Society of America, ANSI/ASA S12.19; 1996 R2011.
ISO. Acoustics: Determination of Occupational Noise Exposure: Engineering Method. Geneva, Switzerland: International Organization for Standardization, ISO 9612; 2009.
ISO. Acoustics: Determination of Sound Immission from Sound Sources Placed Close to the Ear: Part 1: Technique Using a Microphone in a Real Ear (MIRE Technique). Geneva, Switzerland: International Organization for Standardization, ISO 11904-1; 2002.
ISO. Acoustics: Determination of Sound Immission from Sound Sources Placed Close to the Ear: Part 2: Technique Using a Manikin (Manikin Technique). Geneva, Switzerland: International Organization for Standardization, ISO 11904-2; 2004.
AS/NZS. Occupational Noise Management Part 1: Measurement and Assessment of Noise Immission and Exposure. Sydney, Australia: Standards Australia and Standards New Zealand, AS/NZS 1269-1; 2005.
Occupational Safety and Health Administration (OSHA). Evaluating Noise Exposure of Workers Wearing Sound-generating Headsets. Technical Manual (Ch. 5, Sec. III, Appendix F). Washington, DC, USA: Office of the Federal Register; 1999.
IEC. Electroacoustics: Simulators of Human Head and Ear - Part 1: Ear Simulator for the Measurement of Supra-aural and Circumaural Earphones. Geneva, Switzerland: International Electrotechnical Commission, IEC 60318-1; 2009.
ITU-T. Series P: Terminals and Subjective and Objective Assessment Methods. Geneva, Switzerland: International Telecommunication Union Standardization Sector, ITU-T P.57; 2009.
IEC. Electroacoustics: Simulators of Human Head and Ear - Part 4: Occluded-ear Simulator for the Measurement of Earphones Coupled to the Ear by Means of Ear Inserts. Geneva, Switzerland: International Electrotechnical Commission, IEC 60318-4; 2010.
CSA. Measurement of Noise Exposure. Mississauga, Canada: Canadian Standards Association, CSA Z107.56; 2013.
Nassrallah F, Giguère C. Communication headsets in the workplace: Accessibility to noise exposure measurement in Canada. Can Acoust 2014;42:15-21.
Nassrallah F, Giguère C, Dajani HR. Comparison of Direct Measurement Methods for Headset Noise Exposure in the Workplace. Proceedings of the 21 st
International Congress on Acoustics; 2-7 June, 2013, Montreal, Canada; 2013.
Dreschler WA, Verschuure H, Ludvigsen C, Westermann S. ICRA noises: Artificial noise signals with speech-like spectral and temporal properties for hearing instrument assessment. International collegium for rehabilitative audiology. Audiology 2001;40:148-57.
IEC. Sound System Equipment - Part 1: General. Geneva, Switzerland: International Electrotechnical Commission, IEC 60268-1; 1985.
Vermiglio AJ. The American English hearing in noise test. Int J Audiol 2008;47:386-7.
Laroche C., Giguère C., Vaillancourt V., and Soli S. Development and validation of hearing standards for Canadian Coast Guard seagoing personnel and C&P and land-based personnel - Phase II. Final Report to Department of Fisheries and Oceans under Contract No. F0753-0000009; 2005.
Zheng Y, Giguère C, Laroche C, Sabourin C, Gagné A, Elyea M. A psychoacoustical model for specifying the level and spectrum of acoustic warning signals in the workplace. J Occup Environ Hyg 2007;4:87-98.
Rhebergen KS, Versfeld NJ. A Speech Intelligibility Index-based approach to predict the speech reception threshold for sentences in fluctuating noise for normal-hearing listeners. J Acoust Soc Am 2005;117(4 Pt 1):2181-92.
ANSI/ASA. Methods for Calculation of the Speech Intelligibility Index. Washington, DC, USA: American National Standards Institute/Acoustical Society of America, ANSI/ASA S3.5; 1997 R2012.
ITU-T. Head and Torso Simulator for Telephonometry. Geneva, Switzerland: International Telecommunication Union Standardization Sector, ITU-T Rec; 2013. p. 58.
IEC/TS. Electroacoustics: Simulators of Human Head and Ear - Part 7: Head and Torso Simulator for Acoustic Measurement of Hearing Aids. Geneval, Switzerland: International Electrotechnical Commission/Technical Specifications, IEC/TS 60318-7; 2011.
IEC. Electroacoustics: Simulators of Human Head and Ear - Part 2: An Interim Acoustic Coupler for the Calibration of Audiometric Earphones in the Extended High-frequency Range. Geneva, Switzerland: International Electrotechnical Commission, IEC 60318-2; 1998.
IEC. Occluded-ear Simulator for the Measurement of Earphones Coupled to the Ear by Ear Inserts. Geneva, Switzerland: International Electrotechnical Commission, IEC 60711; 1981.
IEC. Electroacoustics - Sound Level Meters - Part 1: Specifications. Geneva, Switzerland: International Electrotechnical Commission, IEC 61672-1; 2013.
Macrae JH. Hearing conservation standards for occupational noise exposure of workers from headphones or insert earphones. Aust J Audiol 1995;17:107-14.
Bland JM, Altman DG. Measurement in medicine: The analysis of method comparison studies. J R Stat Soc 1983;32:307-17.
Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res 1999;8:135-60.
Bland JM, Altman DG. Measurement error. BMJ 1996;312:1654.
ANSI/ASA. Specification for Audiometers. Washington, DC, USA: American National Standards Institute/Acoustical Society of America, ANSI/ASA S3.6; 2010.
ISO. Acoustics - Hearing protectors - Part 2: Estimation of Effective A-weighted Sound Pressure Levels when Hearing Protectors are Worn. Geneva, Switzerland: International Organization for Standardization, ISO 4869-2; 1994.
University of Ottawa, 451 Smyth Road, Ottawa, Ontario, K1H 8M5
Source of Support: None, Conflict of Interest: None
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5], [Table 6], [Table 7]