| Article Access Statistics|
| Viewed||16314 |
| Printed||337 |
| Emailed||2 |
| PDF Downloaded||40 |
| Comments ||[Add] |
| Cited by others ||9 |
|Year : 2013
: 15 | Issue : 67 | Page
|Comparison of sound propagation and perception of three types of backup alarms with regards to worker safety
Véronique Vaillancourt1, Hugues Nélisse2, Chantal Laroche1, Christian Giguère1, Jérôme Boutin2, Pascal Laferrière1
1 School of Rehabilitation Sciences, Faculty of Health Sciences, University of Ottawa, Ottawa, Ontario, K1H 8M5, Canada
2 Research and Expertise Division, Institut de Recherche Robert-Sauvé en Santé et Sécurité Du Travail, Montréal, Québec, H3A 3C2, Canada
Click here for correspondence address
|Date of Web Publication||12-Nov-2013|
A technology of backup alarms based on the use of a broadband signal has recently gained popularity in many countries. In this study, the performance of this broadband technology is compared to that of a conventional tonal alarm and a multi-tone alarm from a worker-safety standpoint. Field measurements of sound pressure level patterns behind heavy vehicles were performed in real work environments and psychoacoustic measurements (sound detection thresholds, equal loudness, perceived urgency and sound localization) were carried out in the laboratory with human subjects. Compared with the conventional tonal alarm, the broadband alarm generates a much more uniform sound field behind vehicles, is easier to localize in space and is judged slighter louder at representative alarm levels. Slight advantages were found with the tonal alarm for sound detection and for perceived urgency at low levels, but these benefits observed in laboratory conditions would not overcome the detrimental effects associated with the large and abrupt variations in sound pressure levels (up to 15-20 dB within short distances) observed in the field behind vehicles for this alarm, which are significantly higher than those obtained with the broadband alarm. Performance with the multi-tone alarm generally fell between that of the tonal and broadband alarms on most measures.
Keywords: Backup alarms, detection threshold, occupational noise, perceived urgency, sound localization
|How to cite this article:|
Vaillancourt V, Nélisse H, Laroche C, Giguère C, Boutin J, Laferrière P. Comparison of sound propagation and perception of three types of backup alarms with regards to worker safety. Noise Health 2013;15:420-36
|How to cite this URL:|
Vaillancourt V, Nélisse H, Laroche C, Giguère C, Boutin J, Laferrière P. Comparison of sound propagation and perception of three types of backup alarms with regards to worker safety. Noise Health [serial online] 2013 [cited 2023 Sep 25];15:420-36. Available from: https://www.noiseandhealth.org/text.asp?2013/15/67/420/121249
| Introduction|| |
Audible backup alarms installed on mobile equipment are used to promptly alert nearby workers or passersby of safety risks associated with reverse vehicular operation and to provide the warning in the danger zone right behind the moving equipment. In the USA, for example, Section 1926.601 of the Occupational Safety and Health Administration's regulation on motor vehicles states that "no employer shall use any motor vehicle equipment having an obstructed view to the rear unless: (b)(4)(i) The vehicle has a reverse signal alarm audible above the surrounding noise level or: (b)(4)(ii) The vehicle is backed up only when an observer signals that it is safe to do so."  Standards such as SAE J994  and ISO 9533  provide guidance on the technical characteristics of backup alarms and the sound level adjustment procedures to be used in the field. Still, accidents and fatalities involving vehicles operating in reverse are reported every year. ,,,,,, Two important factors may affect the effectiveness of conventional backup alarms on worker safety: Non-uniform sound field generated by the alarms behind vehicles and sub-optimal acoustical characteristics. ,,,
A uniform distribution in alarm sound level is not necessary everywhere behind the vehicle, but it is crucial immediately in the danger zone, such distribution, free of abrupt spatial variations is important to promote homogeneous auditory situational awareness in this hazard area. In free space, sound decreases uniformly at a rate of 6 dB per doubling of distance in all directions from the source. In the field, however, the direct sound from the alarm is subject to diffraction and screening effects from the vehicle body and it also interacts with ground reflections and those from nearby obstacles to produce a more or less complex sound field behind the vehicle. As a result, the uniformity of the sound field is not guaranteed in the field, in particular for tonal alarms. ,,, As shown by Laroche et al.,  variations in sound pressure level is up to 15 dB are common for tonal alarms over very short distances of just a few centimeters when moving axially or laterally at the rear of the vehicle. Large differences in the loudness and urgency evoked by the alarm can stem from such abrupt variations in sound pressure levels, depending on the exact position of workers in the danger zone and confusions can arise in the perceived direction and distance of reversing vehicles that fall outside the visual field of nearby workers. The sound propagation behind the vehicles will also be affected by the alarm mounting location on the vehicle, thus impacting auditory perception of the alarm signal as shown recently by Heckman et al.  These authors have found that localization of the backup alarm was strongly affected by the mounting location, most notably when the alarm stimuli originated from either behind or in front of the listener. In addition to these sound propagation issues immediately behind the vehicles, conventional (tonal) backup alarm signals can travel over distances well beyond the danger zone, resulting in noise annoyance in workers, which are not at risk. Repeated exposure to false alarms (habituation) decreases the efficacy by reducing the degree of association between the alarm sound and a particular danger. ,,
Irrespective of sound field propagation issues, the inherent acoustical characteristics of the alarms (overall level, spectrum and tempo) also largely influence their efficacy in conveying good auditory situational awareness about mobile reversing equipment. Salient signals conveying a high degree of urgency and unequivocal information about the source of danger are required. , For example, a worker's ability to tell if reversing heavy vehicles are approaching from the front, back or either sides is critically important to promote a suitable reaction away from the vehicle upon detection. In theory, tonal sounds offer impoverished spatial cues compared to broadband signals. Although interaural time differences (ITDs) aid in localizing sounds below 1500 Hz and interaural intensity differences (IIDs) allow one to localize sounds higher than 3000 Hz, high frequency (>5000 Hz) spectral cues help resolve front/back (F/B) confusions and identify the position (elevation) of a source in the vertical plane. ,,,, Broadband sounds (BBS) are theoretically easier to localize as they offer a greater number of cues (IID, ITD and spectral cues) compared with sounds with a limited frequency spectrum such as conventional ambulance sirens and tonal backup alarms, which typically have a dominant frequency between 1000 and 4000 Hz,  a frequency region where few localization cues are available. The SAE J994  standard currently states that the dominant frequency of backup alarms should be between 700 and 2800 Hz. Confusions in judging the position of a source can yield increased response times in situations where a timely reaction is often critical to avoid a potentially dangerous hazard. From a work safety perspective, the effects of hearing protection devices (HPDs) and hearing loss should also be taken into consideration since both factors can be detrimental to the performance in auditory tasks, ,,, which may affect alarm detection, saliency, urgency and spatial localization.
In addition to safety aspects, the noise generated by backup alarms will propagate away from the vehicles and quite often, be a source of nuisance for residents living in close proximity.  In a report published by the National Academy of Engineering,  conventional tonal alarms were cited as one of the most important noise sources yielding behavioral and emotional consequences. They have also been identified as a major problem of construction night-time noise in many American States.  As mentioned earlier, annoyance from backup alarms affects workers who find them too loud and distracting and who sometimes will choose to deliberately sabotage the alarm device or turn it off. 
In recent years, a new type of vehicle backup alarm  has been drawing increasingly more interest from many industrial sectors. The new alarm, based on the use of broadband noise instead of the conventional typical tonal signal, is deemed to reduce environmental noise annoyance close to industrial settings and construction sites, to be more efficient for spatial localization and to provide a more uniform sound propagation behind vehicles. Promotional documents also claim that the BBS backup alarm is less susceptible to masking. While conceptually appealing, few published and peer-reviewed scientific studies have demonstrated the advantages and disadvantages of such an alarm to ensure worker safety, particularly in comparison to existing technologies. ,, Moreover, some standards (e.g. SAE J994  ) were initially developed to cover the characteristics of tonal-type alarms and test requirements were not specifically written or updated to deal with the acoustical properties of noise-type signals such as the new technology of broadband alarms.
Comparative studies on detection and audibility
As specified by ISO 9533,  differences between sound pressure levels measured while the vehicle is operating at high idle with the alarm off and those measured while the vehicle is in low idle (neutral) with the alarm on must exceed 0 dB at 7 measurement points behind the vehicle. Homer  reported such measurements for a tonal and a broadband alarm in a mining environment. The 0 dB signal-to-noise ratio (SNR) criterion was satisfied at all but one measurement point with the tonal alarm, compared to only three for the broadband alarm. The tonal alarm also propagated over greater distances than the broadband alarm, covering an area approximately 45% larger than the broadband alarm in a high noise environment. Both alarms did not achieve the 15 dB SNR audibility criterion specified in the ISO 7731  standard for general auditory danger signals. Indeed, the maximum alarm level was 114.0 dBA for the tonal alarm and 111.8 dBA for the broadband alarm, exceeding the background noise level (104.4 dBA) by only 9.6 dBA and 7.4 dBA, respectively. Finally, according to the author, similarities between the spectrum of the broadband alarm and that of the noise could make it more susceptible to masking in high noise environments than the tonal alarm. It should also be noted that maximum levels differed significantly between the broadband (111.8 dBA) and the tonal (114.0 dBA) alarms and that psychoacoustic measurements were not carried out to relate physical measurements to perceptual outcomes.
A study on sound propagation behind heavy vehicles by the Australian Department for Transport, Energy and Infrastructure compared three different alarms: The broadband alarm, a focused tonal alarm and an intelligent tonal alarm with automatic gain adjustment. The first two met audibility criteria set forth in the ISO 9533  standard, but not the intelligent tonal alarm.  Likewise, some sound directivity was demonstrated for the broadband alarm and the focused tonal alarm, but not for the intelligent self-adjusting tonal alarm. Moreover, sound propagation was less hindered for the broadband alarm compared to the two other alarms when screened by various parts of the vehicle. In the same study, telephone surveys with eight different user company representatives revealed less security issues, greater identification of the reversing vehicle by workers, accrued attention to safety, equivalent detection of the alarm behind vehicles, reduced worker fatigue and a considerable reduction in noise complaints from neighbors, particularly at night, following the transition from a tonal to a broadband alarm. As methodological details of the survey were not documented, the validity and applicability of the reported benefits are difficult to ascertain.
For noise levels greater than 80-85 dBA, passive hearing protectors do not generally seem to hinder sound detection and speech perception and can even improve such abilities in listeners with normal hearing. ,,, They can however reduce speech discrimination in background noise levels lower than 80 dBA, even in normal hearing listeners, ,,, and particularly hinder sound detection  and speech perception  in individuals with hearing loss. For example, Robinson and Casali  showed that a tonal alarm could remain audible at fairly low SNRs (0 dB) in noise levels greater than 85 dBA when passive earmuffs are used, even in the presence of significant hearing loss (45-50 dB HL), indicating that workers with such hearing loss should hear the alarm if ISO 7731  is met (alarm level exceeding that of the background noise by at least 15 dB). Active level-dependent devices which protect against loud noises but allow sound transmission at lower levels did not seem to significantly improve masked thresholds in individuals with normal hearing , or those with hearing loss,  compared with passive protectors.
In another study, , masked thresholds for a conventional tonal alarm in 85 and 100 dBA noises were compared for normal hearing individuals participating in a parallel probability gauge monitoring task in an unoccluded condition (85-dBA noise only), with passive earmuffs and earplugs and while wearing an active noise reduction earmuff. In 100-dBA noise, masked thresholds were better for passive earplugs and the earmuff with active noise reduction was advantageous over the passive earmuffs. At 85 dBA, no difference was noted amongst the various HPDs, although an improvement in masked thresholds of about 2.6-4.3 dB was noted compared to unoccluded.
To the best of our knowledge, only one study on sound detection directly compared a tonal alarm (Preco 6003 unit) and a broadband alarm (BBS-97 unit), in an experiment involving both unprotected listening and use of passive and active HPDs.  In a fairly quiet background noise (52.3 dBA), minimum distance for detection in 12 normal hearing individuals was assessed in 8 listening conditions (no protection and with 7 different HPD types). Mean detection distance was less for a passive earmuff (1132 feet) and greatest for the unprotected condition (1652.3 feet). Apart from the passive earmuff and a passive earplug (characterized by greater attenuation values), HPDs did not significantly decrease detection distances compared with unprotected performance. Finally, a 221.5-feet advantage in detection distance was noted for the tonal alarm (1600.9 feet) compared to the broadband alarm (1379.4 feet). This difference is interpreted by the author as a significant advantage allowing the tonal alarm to be detected over greater distances, thereby allowing longer reaction times. Caution must be exerted in using such findings to claim that the tonal alarm is inherently safer. The detection distances found with either alarm in this study appear large enough to provide early warning to the workers. If the tonal alarm is heard over greater distances, but is ignored for lack of impending danger (habituation) or is associated with the alarm of a nearby vehicle (confusion), this extra amount in reaction time would not prove to be beneficial and the safety could be hindered rather than improved. Furthermore, the greater detection distance for the tonal alarm may simply result in an increased noise annoyance in the community. Many factors must therefore be taken into account when reaching conclusions on the superiority of a given alarm type.
Comparative studies on sound localization
It is generally agreed upon that frequency-rich signals are easier to localize than signals with limited spectral content and broadband signals have been shown to be advantageous over conventional sirens on emergency vehicles during localization tasks. ,,,
Relative to backup alarms, in a survey involving 1477 vehicles (313 equipped with a broadband alarm), 80% of respondents (site managers and workers) claimed to always correctly identify a reversing vehicle equipped with a broadband alarm, compared with only 10% for a tonal alarm and all said the former was less annoying.  In 2005, the National Institute for Occupational Health and Safety financed a study on sound localization in controlled conditions at Washington State University  using headphone rendering of binaural recordings made during two scenarios of a reversing vehicle; one which would result in a direct hit and the other, a near miss. Neither alarm (broadband or tonal) proved superior over the other. It should be noted that because generic head-related transfer function, based on the Knowles Electronics Manikin for Acoustic Research (KEMAR), were used instead of those specific to each individual, trials in which front-back confusions occurred were eliminated. However, such confusions are the most critical from a safety standpoint and are known to be highly dependent on the spectral content of the signal. 
Catchpole et al.,  investigated the sound localization of various signals using combinations of pure tones and noises. In a first part, 8 individuals with normal hearing were asked to judge if the signal (white noise, notched white noise from 1000 to 3000 Hz, notched white noise from 1000 to 10000 Hz, 2000-Hz pure tone, ascending frequency sweep from 1000 to 3000 Hz and descending frequency sweep from 3000 to 1000 Hz) came from the right or left. Localization was more accurate and faster with the noise signals compared to the tonal signals, with the greatest accuracy for the white noise and least accuracy for the 2000-Hz pure tone. In a second task, the same individuals were asked to both identify and localize correctly 6 signals (each of the three tonal signals individually and in combination with the notched white noise from 1000 to 3000 Hz). Localization/identification was more accurate and faster for the combination signals (tone + noise), indicating that one can increase the efficacy of a tonal alarm by adding noise. In a third part of the study, 18 individuals were asked to estimate, on a scale of 0-100, the perceived urgency evoked by three stimuli (noise + ascending frequency sweep, noise + descending frequency sweep and a 5-harmonic complex tonal signal). Overall, the ascending frequency sweep was judged more urgently than the descending frequency sweep, which was in turn perceived more urgent than the tonal signal. While adding noise to a tonal alarm could increase safety behind heavy vehicles by improving sound localization, noise nuisance would not necessarily be reduced.
Passive HPDs are usually detrimental to sound localization. ,,,,,,, To the best of our knowledge, few studies have specifically examined the effect of HPDs on the sound localization of conventional alarms compared to alternative alarm designs. ,,
In a first study, Alali and Casali  investigated the effect of 7 HPDs on the ability of normal hearing individuals to localize a conventional single-tone alarm and a modified spectrum alarm (with additional frequency components at 400 and 4000 Hz) in two levels of pink noise (60 and 90 dBA). The 15-s signals were presented through eight loudspeakers covering 360° in eight conditions of hearing protection: Either unprotected, using earplugs (two passive earplugs and two augmented earplugs) or using earmuffs (one passive, one active dichotic and one active diotic). Individuals were required to identify the perceived position of the sound source. Head movements were allowed and the alarm's level was increased within a given trial to simulate a vehicle approaching at a speed of 10 mph. For each of the dependent variables, only performance with the diotic earmuffs was found to differ significantly from all other listening conditions, a result that can be explained by the loss of binaural cues resulting from the use of a single microphone feeding input to both left and right earcups. Overall, localization was better in the low noise compared to the high noise environment and left/right (L/R) localization was better than F/B localization, with significantly greater F/B confusions in the high noise environment. Finally, results showed some advantages for the use of the modified alarm in high noise environments compared to the conventional alarm. However, results did not demonstrate the superiority of new hearing protection technologies over conventional passive devices. Localization accuracy was generally good without and with HPDs, with the exception of the diotic listening device. Similar results were obtained by Casali and Alali  when comparing the performance of an active earplug to unprotected performance in a different group of individuals. The specific contribution of head movements to such results was not studied. However, Noble,  demonstrated that sound localization accuracy in the horizontal plane for a band-limited noise centered on 1000 Hz could by significantly improved by allowing head movements when participants wore earmuffs, whereas near perfect sound localization was achieved without protection when participants were allowed head movements.
Summary of findings
The benefits of broadband alarms remain unclear. Most of the studies comparing broadband alarms to conventional alarms only addressed either noise annoyance or some aspect related to safety, such as detection distance, masked detection thresholds or sound localization, often in quiet or in artificial noises, with little attention to sound propagation issues behind vehicles in the field, making laboratory findings difficult to project to real workplaces. In addition, there were few studies on the perception of alarms for workers wearing HPDs - a common situation in most workplaces. Furthermore, the alarms under the study were often used as provided by the manufacturers without any control over some key acoustical parameters, such as output volume or perceptual features such as detection thresholds, which can be quite different for tonal and noise signals, making it difficult to interpret comparisons of absolute sound pressure levels across alarms and to relate outcomes to either the inherent properties of the alarm signals per se or the given characteristics of the alarm units tested.
| Objectives of the study|| |
This two-part study was intended to compare three types of backup alarms (the conventional tonal signal, a multi-tone signal and the broadband noise technology) on several aspects related to safety behind heavy vehicles for workers wearing hearing protection or not. More specifically, the aims of the study were to: (i) Verify conformity of the broadband alarm with the SAE J994  standard; (ii) determine if the multi-tone and broadband alarms provide a more uniform sound propagation pattern behind heavy vehicles than the conventional tonal alarm; and (iii) compare the performance of the three alarms on various psychoacoustic measures in individuals with normal hearing wearing or not HPDs while exposed to noises from typical workplaces where heavy vehicles are found. The first part, performed in the field, focused on objective measurements of the sound propagation immediately behind non-moving vehicles for various vehicles and terrain configurations. The second part, performed in a laboratory environment, was centered on subjective measurements over various psychoacoustic tasks (detection thresholds, loudness, perceived urgency and sound localization). It should be noted that the impact of the broadband noise technology on reducing noise annoyance is not addressed in this study.
| Methods|| |
Sound field behind vehicles
Three backup alarms were tested in this study: (i) A conventional tonal alarm from Grote Industries (http://www.grote.com/); (ii) a custom-made multi-tone alarm; and (iii) a broadband alarm from Brigade Electronics (http://www.brigade-electronics.com/). The multi-tone alarm was proposed by Laroche et al., as a more desirable design than the conventional tonal alarm to achieve a more uniform sound field behind the vehicle. It was included in this study as an alternative to the two commercial alarms.
The frequency content of the three alarms is illustrated in [Figure 1]. Sound pressures levels (SPL), measured at approximately 1 m in front of the alarms in a semi-anechoic chamber are shown as a function of frequency. The multi-tone alarm consists of three primary tones located between 1000 and 1300 Hz, whereas the conventional tonal alarm contains a single primary tone around 1250 Hz in addition to even and odd harmonics weaker by 30 dB or more. For the broadband alarm, the energy is distributed over a larger frequency span, most of the energy being found in the 700-4000 Hz range.
Field tests were then performed at three different sites, each with a different terrain configuration and vehicle. The first site was a limestone plant consisting of a hard soil with rocks, gravel and dirt and the tested vehicle was a large wheel loader. The second site was a limestone plant with a hard soil consisting of gravel and dirt and the tested vehicle was a 50 tons off-highway truck. The last site was a sawmill plant with hard soil and the chosen vehicle was a large wheel log loader. To study the sound propagation behind the vehicles, a test method inspired from standard ISO 9533  was developed. The ISO 9533 procedure was adapted and enhanced to be able to produce, in addition to what is required by the standard, contour maps of the sound field behind the vehicles for the different alarms. Type 1 ½-inch microphones from BSWA Technology Co. were used in conjunction with an Edirol audio recorder and National Instruments/Labview acquisition system for sound pressure measurements. For the positioning of the alarms on the vehicle, two mounting scenarios were considered: (i) A "realistic" mounting, which consisted of replicating the position of the alarm on the vehicle used at each site (the alarms were off-centered in all three cases tested); and (ii) an "ideal" mounting where the alarm was centered, unobstructed and facing outward.
Two sets of measurements were performed for each alarm at all three sites. In the first set, alarm level adjustments were performed by measuring the sound pressure levels at the seven microphone positions specified in the ISO 9533 standard [Figure 2]a. The alarm level was then manually adjusted so that a difference equal to or greater than 0 dB (SNR ≥ 0 dB) was obtained at all measurement points between the sound levels generated when the vehicle was operating at high idle without alarm present and those prevailing when the backup alarm was activated and the vehicle operated at low idle. The procedure was repeated for each alarm. It allowed examining if one alarm type would require higher levels than others to maintain the desired SNR ≥ 0 dB at all microphones. A reference microphone was located at a center position 1 meter behind the vehicle to monitor alarm levels.
|Figure 2: (a) Position of the seven microphones for the alarm level adjustment (as per ISO 9533). (b) Illustration of the scanning lines for sound mapping measurements|
Click here to view
In the second set, a microphone was mounted to a pole and digital audio recordings were performed while the alarm was activated by moving the microphone slowly at a constant speed along 9 axes and 2 curvilinear arches behind the vehicle [Figure 2]b. The alarm levels were set at the values found during the first set of measurements and the vehicle engine was stopped. A post-processing scheme was developed under MATLAB to obtain sound pressure levels along the various lines. Subsequently, an interpolation algorithm was used to produce sound pressure level contour maps behind the vehicle when the alarm is activated. For each site, such a procedure was used to investigate the uniformity of the sound field generated by the three alarms.
A total of 24 (6 men; 18 women) young adults (mean = 25.0 years; standard deviation = 2.3 years) with normal hearing (hearing thresholds ≤ 25 dB HL from 250 to 8000 Hz) took part in several psychoacoustic tasks (detection thresholds, equal loudness judgments, perceived urgency ratings and sound localization) in the laboratory, both with and without HPDs. The first three tasks were carried out in four background noises recorded in the field and played back in a sound simulation room [Figure 3]. The noises were selected from a larger set of recordings to cover a range of the overall sound levels as well as temporal and frequency contents. Noise 1 (limestone plant), at 80.5 dBA, is a buzzing sound containing significant low frequency components. Noise 2 (quicklime plant), at 83.3 dBA, is also a buzzing sound but shows more stable temporal characteristics. Noise 3 (sawmill plant), at 85.9 dBA, is characterized by some flapping, whispering and crackling noises associated to wood chips processing. Finally, Noise 4 (sawmill plant), at 89.6 dBA, is more like a buzzing sound where one can hear heavy vehicle engines and strident high frequency sound bites in the background. The spectra of the four noises are presented in [Figure 4]. The fourth psychoacoustic task was carried out in Noise 2 in a large audiometric room equipped with a sound localization system.
|Figure 3: Sound simulation room with one loudspeaker (S7) at one meter in front of the subject to present the alarms and five loudspeakers (S1-S5) and a subwoofer (S6) to create a quasi-diffuse noise field|
Click here to view
|Figure 4: One-third octave sound level spectra for the four background noises used in the study|
Click here to view
All subjects were tested with one type of hearing protector, in addition to open-ear testing. Half of the subject sample used the peltor optime 95 earmuff (noise reduction rating [NRR] = 21 dB) while the rest wore EAR UltraFit earplugs (NRR = 25 dB). The first three tasks were performed within an initial 90-min testing session, whereas a second session of similar duration was required for the sound localization task. Within each session, all measurements in the open-ear conditions preceded those with HPDs. Prior to testing, subjects were familiarized with the signals and tasks to be performed.
An adaptive method was used for all threshold measurements. Using a tablet PC and software specifically designed for this study, subjects were required to adjust the level of each alarm up and down using 2-dB steps until it was barely audible. Each trial consisted of the subject adjusting the three alarms to the threshold level, one at a time. Testing was performed twice in each of 5 conditions (in quiet and in the 4 background noises), firstly in open-ear conditions and then with the assigned HPD. Testing was also repeated in each condition to increase the reliability of the threshold estimation. A total of 60 thresholds were therefore measured (5 noise conditions × 3 alarms × 2 repetitions, with and without HPD). Initial presentation levels were varied across alarms and trials, but always remained at suprathreshold levels. From this initial level, participants were required to adjust the alarm level until it became inaudible, then up to barely audible in an ascending excursion. The testing order of the noises and alarms was counterbalanced across subjects.
Equal loudness judgments
Loudness refers to the subjective appraisal of a sound's intensity and it depends on attributes such as the sound pressure level and the frequency content of both the signal of interest and the background noise. The tonal alarm presented at a SNR = 0 dB (81, 83, 86 and 89 dBA in Noises 1 through 4) served as the reference alarm for equal-loudness measurements. In each of the four noises, participants were required to adjust the level of broadband and multi-tonal alarms, one at a time, until they were perceived to be as equally loud as the reference alarm. The subject could listen to the comparison or the reference alarm at will. Each trial therefore consisted of 2 alarm comparisons (broadband vs. tonal and multi-tone vs. tonal) for a total of 16 equal-loudness adjustments (2 alarm comparisons × 4 noises, with and without HPD). The testing order of the noises and comparison alarms was again counterbalanced across participants. Starting levels for broadband and multi-tone alarms were also randomized.
Equal loudness does not guarantee that the alarms will convey the same sense of urgency to the workers. Like loudness, perceived urgency not only depends upon the sound pressure level of the signal, but also on a number of other acoustical characteristics, ,,,, including the frequency content and temporal characteristics and one's familiarity with the signal acting as a warning signal. To investigate the degree of urgency evoked by the three backup alarms, they were presented at three different SNRs (−6 dB, 0 dB and +6 dB) while subjects had to rate alarm urgency on a scale of 0-100, with a rating of 0 indicating that the alarm was heard, but evoked no sense of urgency and 100 being most urgent. Nine urgency ratings (3 alarms × 3 SNRs) were performed in each of the 4 background noises, with and without HPD, for a total of 72 ratings. The testing order of the noises, alarms and SNR presentation levels was randomly selected by the testing software.
Finally, the ability to judge the direction of backup alarm incidence was assessed through a source-identification task in the horizontal plane using a set of 12 loudspeakers arrayed uniformly over a 180° localization arc [Figure 5]. Subjects were seated in the center of the localization arc, at a distance of 1 meter from each loudspeaker. Three spatial configurations were tested, with the loudspeaker arc placed behind the subjects (to quantify L/R confusions) and to their right and left sides (to quantify F/B confusions). Subjects were not allowed head movements in order to assess sound source-identification in the most challenging situation that could occur in the workplace, where workers may not be free or have the reflex or time to use head movements to resolve confusions.
|Figure 5: Loudspeaker array for sound localization judgments in three configurations, with speakers to the left or right to assess front/back localization and speakers behind to assess left/right localization|
Click here to view
The alarm signals were adjusted to simulate increasing sound pressure levels associated with a vehicle reversing at a speed of 4.4 m/s (10 mph). Testing was performed in one of the selected background noises (Noise 2 at 80 dBA) presented directly overhead. The SNR gradually increasing from −6 to 0 dB over 3 s to simulate the sound pressure level rise of the alarm at a fixed worker's position as the vehicle is backing up in uniform noise. In this scenario, the signal ends 2 s before projected impact (vehicle within a 8.8 m distance), a choice based on the 2 s reaction time required to adequately react and move away from the trajectory of an oncoming vehicle upon hearing an alarm signal according to standard SAE J1741.  Following familiarization, a given alarm signal was randomly presented in a bloc of 24 trials (twice from each of the 12 loudspeakers) and subjects were required to verbally identify the loudspeaker thought to have emitted the sound. The task consisted of 432 sound localization judgments (24 trials ×3 spatial configurations ×3 alarms, with and without HPD). The testing order of the alarms and spatial configurations was counterbalanced across subjects.
| Results|| |
Conformity with SAE J994
Standard SAE J994  defines a set of performance requirements for backup alarms designed to be installed on off-road vehicles. Originally and for a long time, only tonal signals were used in backup alarms. The acoustical tests described in the standard were then designed to ensure that such tonal characteristics were dominant in the emitted noise spectrum, within a specified bandwidth. These laboratory tests mainly consists in measuring the frequency spectrum of the alarm in a semi-anechoic room using a microphone facing the alarm at 1.2 m. Absorbing materials should be placed on the ground between the microphone and the alarm to eliminate potential reflections. The standard then states that "the predominant sound frequency of the alarm shall be defined as the frequency that produces the highest A-weighted sound pressure level. The acceptable frequency range is 700-2800 Hz." The measured spectra for the three alarm tested are presented in [Figure 1]. If we examine the spectra, one clearly sees that the tonal and the multi-tone alarms show well-defined maxima between 1000 and 2000 Hz, easily meeting the required frequency range of 700-2800 Hz. In the case of the broadband alarm, the energy is distributed over a large frequency span, most of the energy being found in the 700-4000 Hz range and there are no clear, well-defined peaks or maxima. However, strictly speaking, the standard defines the predominant sound frequency as the frequency that produces the highest A-weighted sound pressure level. Therefore, a literal application of this definition leads to conclude that the broadband alarm meets the SAE J994 requirements as the highest A-weighted sound pressure level is found to be between 700 and 2800 Hz.
Alarm level adjustment per ISO 9533
Results obtained for the "SNR ≥ 0 dB" procedure are summarized in [Table 1]. For each alarm, site and mounting configuration, the mean and standard deviation of the SNR is presented as well as the sound pressure levels at the reference microphone. It is observed that higher levels had to be used for the tonal alarm compared to the multi-tone and broadband ones. Furthermore, higher mean SNR and standard deviations are obtained for the tonal alarm, suggesting more sound level variations behind the vehicle when using this signal.
|Table 1: Mean (standard deviation) values of the signal-to-noise ratio (expressed in dB) over the 7 microphone locations and sound pressure levels (in dBA) at the 1m reference microphone for the two mounting conditions of the alarms|
Click here to view
Sound pressure level maps
Maps of the overall sound pressure levels behind the vehicle are presented in [Figure 6] for site 1 (top row: "Ideal" mounting; bottom row: "Realistic" mounting) where each change of color corresponds to a 3 dB step in SPL. Variations of the sound pressure level in excess of 15 dB over a short range of less than 1 m can be observed for the tonal alarm due to the effect of acoustic interferences. Not surprisingly, this interference effect is quite pronounced when only one strong tonal component dominates. However, it smoothes out considerably when adding two additional tonal components to the signal, as it is the case for the multi-tone alarm. Finally, a relatively more uniform and monotonically decreasing sound field was obtained for the broadband alarm.
|Figure 6: Sound pressure levels behind the vehicle (expressed in dBA) for site 1. Results for the "ideal" mounting condition are presented in the top row while those for the "realistic" mounting condition are shown in the bottom row|
Click here to view
Another way of looking at the variation of the sound field is shown in [Figure 7] where the SPL is plotted as a function of the distance from the vehicle along a straight line centered right behind the vehicle. Not only is a much smoother decreasing sound field without sharp variations observed with the broadband alarm, but lower levels are also obtained with this alarm when the alarms are set in accordance with the level adjustment per ISO 9533 procedure presented in an earlier section [Table 1]. The SPL maps also show that the alarm type has a greater impact on sound field variations behind the vehicle than the actual mounting of the alarms on the vehicles, at least for the mounting conditions, vehicles and terrain configurations examined in this study. Finally, it should be mentioned that similar results to those presented for site 1 were obtained at the two other sites.
|Figure 7: Sound pressure levels (expressed in dBA) as a function of the distance from the vehicle along a centered line right behind the vehicle for site 1 ("ideal" mounting condition). Alarm levels are adjusted as per ISO 9533 for a minimum signal-to-noise ratio of 0 dB at the 7 microphone positions|
Click here to view
Most data on sound detection, equal loudness, perceived urgency and sound localization obtained in this study did not satisfy the assumption of sphericity required for performing univariate statistical analyzes. Furthermore, a mixed linear model with an unstructured covariance matrix could not be used for the analysis since these data are highly dimensional. Multivariate methods of analyses based on an ANOVA type statistic proposed by Ahmad et al.,  were then used as needed. SAS/STAT version 9.2 (from SAS-Statistical Analysis System) was used to perform the analyses in conjunction with interactive matrix language programming by a statistician.
Detection results obtained in the four selected background noises, with and without HPDs, are expressed as the average SNR at threshold in [Figure 8]. Since measurement was repeated in each testing condition, individual thresholds were averaged across both measurements. Over all testing conditions, average A-weighted detection thresholds varied from −13 to −24 dB SNR.
|Figure 8: Mean A-weighted signal-to-noise ratio at threshold (in dB) in the four selected background noises, with and without hearing protectors. The upper (lower) panel refers to the group of subjects wearing earmuffs (earplugs). Error bars designate ±1 standard error of the mean|
Click here to view
As can be seen, error bars often intersect each other, thereby suggesting similar thresholds with and without HPD, particularly when using earplugs, whereas larger differences between open-ear and protected values are noted with earmuffs.
The experimental plan consists of a mixed design with one inter-subject factor (HPD type ‒ earmuffs or earplugs) and repeated measurements on three intra-subject factors: (1) Type of alarm (tonal, multi-tone and broadband); (2) noise (4 noises); and (3) HPD use (with and without). The subject factor was also considered into the analysis.
Using Dempster's (1960) statistic adjusted for non-sphericity, no significant effect on detection thresholds was found for the inter-subject factor of HPD type (F [2.33; 55.92] = 0.736). Data from both groups were therefore combined to carry out the remaining analyses based on ANOVA-type statistics with an unstructured covariance matrix. For intra-subject factors, overall significant main effects were found for alarm type (χ2 [1.431] = 12.978, P < 0.001), noise (χ2 [0.481] = 14.277, P < 0.001) and HPD use (χ2  = 11.347, P = 0.001), as well as significant interactions between alarm type and noise (χ2 [2.083] = 14.408, P < 0.001), alarm type and HPD use (χ2 [1.683] = 5.331, P = 0.008) and between all three intra-subject factors (χ2 [5.583] = 2.385, P = 0.03), at α = 0.05.
[Table 2] summarizes the key statistically significant findings for each significant effect on detection thresholds after adjustment for interacting factors. For the "alarm type" factor of the three-way interaction, the significant differences obtained in Noises 1 and 2 were on the order of 1-3.8 dB, whereas those in Noises 3 and 4 reached 7 dB (2.3-7 dB), with generally higher thresholds for the broadband signal over the tonal alarm in Noises 3 and 4. More specifically, a 5-7 dB advantage of the tonal alarm over broadband in high-frequency rich noises (Noises 3-4) when HPDs are worn is found. For all other noises and conditions of HPD use, the maximum difference between tonal and broadband alarms is 3-4 dB.
When significant differences are noted in the case of the "HPD" factor, threshold levels are generally significantly lower with HPDs.
For equal loudness judgments, participants were required to adjust the level of broadband and multi-tonal alarms to be perceived as equally loud as the tonal alarm (reference) presented at 0 dB SNR in each of the noises. Average differences between the adjusted levels of each alarm and the tonal alarm are found in [Figure 9]. Positive and negative differences are indicative of adjustments to greater and lower levels, respectively than that of the tonal alarm to achieve equal loudness.
|Figure 9: Equal loudness (expressed in dB) with respect to the tonal alarm in the four selected background noises and with/ without hearing protectors. Error bars designate ±1 standard error of the mean|
Click here to view
Although error bars intersect in many cases, suggesting similar results obtained with and without HPDs, differences are often noted, particularly for the broadband alarm in participants wearing earplugs.
The experimental plan consists of a mixed design with one inter-subject factor (HPD type - earmuffs or earplugs) and repeated measurements on three intra-subject factors: (1) Type of alarm (tonal, multi-tone and broadband); (2) noise (4 noises); and (3) HPD use (with and without). The subject factor was also considered into the analysis.
A multi-variate analysis of variance was first used to compare the 16 average data points obtained in both groups of HPD users (earmuffs vs. earplugs) and no effect of HPD type was noted (χ2 [7.788] = 7.710, P = 0.440). Data from both groups were then combined to carry out the remaining analyses. For intra-subject factors, overall significant effects were found (χ2 [3.417] = 36.157, P < 0.001), particularly significant main effects of alarm type (χ2 [1.0] = 18.424, P < 0.001), noise (χ2 [2.536] = 11.771, P = 0.005) and HPD use (χ2 [1.0] = 11.451, P = 0.001) and significant interactions between alarm type and noise (χ2 [3.160] = 11.726, P = 0.010) and alarm type and HPD use (χ2 [1.0] = 4.102, P = 0.043), at α = 0.05. No interaction was revealed between all three intra-subject factors.
[Table 3] summarizes the key statistically significant findings for each significant effect on perceived equal loudness after adjustment for interacting factors. It is found that level differences for equal loudness with the reference is always positive for the multi-tone alarm and negative for the broadband alarm, irrespective of noise and condition of HPD use. In other words, higher levels are required for the multi-tone alarm (0.3-1.4 dB without HPDs; 0.9-2.8 dB with HPDs) to match the loudness of the tonal alarm, whereas lower levels are needed for the broadband alarm (2.0-3.4 without HPDs; 0.2-0.9 with HPDs). At equal sound levels, loudness is therefore greater for the broadband alarm than the tonal alarm, an effect most significant without HPDs.
|Table 3: Key statistically significant findings for perceived equal loudness|
Click here to view
Average perceived urgency ratings on a scale of 0-100 in response to the alarms presented at 3 different SNR in each of the four selected background noises, both with and without HPDs, are displayed in [Figure 10]. Given the lack of intersecting error bars in many cases, HPDs seem to have a significant effect on perceived urgency in both participant groups. It is also clear that presentation levels significantly impact on perceived urgency.
|Figure 10: Perceived urgency rating (0-100 scale) in the four selected background noises, at three signal-to-noise ratio, with and without hearing protectors. Error bars designate ±1 standard error of the mean|
Click here to view
The experimental plan consists of a mixed design with one inter-subject factor (HPD type ‒ earmuffs and earplugs) and repeated measurements on four intra-subject factors: (1) Type of alarm (tonal, multi-tone and broadband); (2) noise (4 noises); (3) HPD use (with and without); and (4) presentation level (−6, 0 and + 6 dB SNR). The subject factor was also considered into the analysis.
The total variances of the variance-covariance matrices were found to be different in both groups (24 244.4 with Box є of 0.095 for earplugs and 20 745.1 with Box є of 0.293 for earmuffs), and thus statistical analyses were performed separately for each group.
In the group using earmuffs, overall significant effects were found (χ2 [1.757] = 15.880, P < 0.001), particularly significant main effects of alarm type (χ2 [2.071] = 13.025, P = 0.002), noise (χ2 [2.360] = 10.535, P = 0.008), HPD use (χ2 [1.0] = 10.053, P = 0.002) and presentation level (χ2 [1.026] = 11.725, P = 0.001), significant second-order interactions between alarm type and noise (χ2 [5.327] = 20.389, P = 0.001), alarm type and presentation level (χ2 [3.550] = 15.900, P = 0.002), HPD use and presentation level (χ2 [1.999] = 11.705, P = 0.003) and a third-order interaction between alarm type, HPD use and presentation levels (χ2 [5.065] = 14.442, P = 0.014).
In the group using earplugs, overall significant effects were found (χ2 [2.124] = 17.423, P < 0.001), particularly significant main effects of noise (χ2 [1.951] = 11.961, P = 0.002), HPD use (χ2 [1.0] = 7.393, P = 0.007) and presentation level (χ2 [1.017] = 11.466, P = 0.001) and a significant 2 nd order interaction between alarm type and HPD use (χ2 [1.689] = 9.026, P = 0.008). The interaction between HPD use and presentation levels (χ2 [1.438] = 8.156, P = 0.009) was also taken into consideration since the P value was very close to the adjusted value of 0.0085 to establish statistical significance.
The following tables summarize the key statistically significant findings for each significant effect on perceived urgency, after adjustment for interacting factors, separately for the groups using earmuffs [Table 4] and earplugs [Table 5]. It should be noted that in both cases (earmuffs and earplugs), few differences reached the threshold for statistical significance due to high data variability. The most relevant finding is based on the presentation levels, with perceived urgency being rated consistently higher at the highest presentation level (+6 dB SNR) compared with the lowest presentation level (−6 dB SNR), with and without both HPD types.
|Table 4: Key statistically significant findings for perceived urgency for the group of subjects wearing earmuffs|
Click here to view
|Table 5: Key statistically significant findings for perceived urgency for the group of subjects wearing earplugs|
Click here to view
During the sound localization experiment, participants were required to identify the source emitting the alarm signals under three speaker array configurations [Figure 5]. The percentage of L/R (for speakers at the back) and F/B (for speakers to each side) confusions are reported herein. Confusions occur when an individual identifies a sound as coming from a source located in a quadrant opposite to that of the actual sound source. The experimental apparatus used in this study can be split up into two 90°-quadrants, as delimited by speakers 1-6 and speakers 7-12 [Figure 5]. It should be noted that in this study, confusions between speakers 6 and 7 consist of a L/R confusion when speakers are at the back, but not a F/B confusion when speakers are placed on either side, a decision based on previous findings. Indeed, data from a previous study using the same apparatus  show that the mean angular error for individuals with normal hearing in the side conditions (11°) is very similar to the 15° angular separation between both speakers. Hence, it is impossible to determine if mistaking speakers 6 and 7 is truly a F/B error or simply a failure to discriminate 2 adjacent sound sources. In contrast, normative data reveal a much smaller mean angular error when speakers are at the back (6°), therefore the same reasoning could not be applied in this condition.
In each listening condition, individual confusion percentages were obtained by dividing the number of confusions by the number of trials (24) and multiplying by 100. [Figure 11] summarizes average results for localization tasks carried out with the three alarms, in each of the three listening conditions (speaker array configuration), both with and without HPDs, in each of the two subject groups (earmuffs and earplugs).
|Figure 11: Percentage of confusion with and without hearing protectors for the three speaker arrangements (behind: left/right confusions; left and right: front/back confusions). Error bars designate ±1 standard error of the mean|
Click here to view
As can be seen, error bars rarely intersect in the earmuff group, suggesting different performances with and without hearing protection, whereas more similar performances are obtained with the earplug group. Results appear to be similar for both groups in the open-ear condition, a finding which was expected. With HPDs, the percentage of confusions appears to be greater for earmuff than earplug users.
The experimental plan consists of a mixed design with one inter-subject factor (HPD type - earmuffs and earplugs) and repeated measurements on three intra-subject factors: (1) Alarm type (tonal, multi-tone and broadband); (2) listening condition (speakers behind, speakers to the right and speakers to the left); and (3) HPD use (with and without). The subject factor was also considered into the analysis.
The total variances of the variance-covariance matrices were found to be different in both groups (22.8 with Box є of 0.616 for earmuffs and 11.1 with Box Ί of 0.956 for earplugs),, and thus statistical analyses were performed separately for each group.
In the group using earmuffs, overall significant effects were found (χ2 [1.584] = 14.938, P < 0.001), particularly significant main effects of alarm type (χ2 [1.190] = 11.752, P = 0.001), listening condition (χ2 [1.112] = 12.295, P = 0.001) and HPD use (χ2 [1.0] = 9.001, P = 0.003), significant second-order interactions between alarm type and listening condition (χ2 [2.295] = 16.016, P = 0 < 001), alarm type and HPD use (χ2 [2.315] = 8.536, P = 0.020) and listening condition and HPD use (χ2 [1.976] = 8.922, P = 0.011) and a third-order interaction between alarm type, listening condition and HPD use (χ2 [2.217] = 15.081, P = 0.001).
In the group using earplugs, overall significant effects were found (χ2 [1.438] = 14.486, P < 0.001), particularly significant main effects of alarm type (χ2 [1.162] = 12.221, P = 0.001) and listening condition (χ2 [1.055] = 12.293, P < 0.001) and a significant second-order interaction between alarm type and listening condition (χ2 [1.743] = 14.955, P < 0.001). Of interest, no significant main effect of HPD use (with and without) was found in earplug users (χ2 [1.0] = 3.540, P = 0.06).
The following tables summarize the key statistically significant findings for each significant effect on sound localization performances, after adjustment for interacting factors, separately for the groups wearing earmuffs [Table 6] and earplugs [Table 7]. In the tables, results reported for the F/B confusions are averaged for the left and right speaker configurations as similar results were obtained in both test conditions. In both groups, the tonal alarm is significantly harder to localize than the mutli-tone and broadband alarms, the broadband alarm yielding the best performance. Furthermore, as expected, localization across all alarm types was best for speakers behind (assessing L/R localization) compared to speakers to the side (assessing F/B localization). Finally while earplugs did not have a significant effect on localization performance, earmuffs significantly increased F/B confusions for all alarm types and L/R confusions for the tonal alarm.
|Table 6: Key statistically significant findings for sound localization (confusions) for the group of subjects wearing earmuffs|
Click here to view
|Table 7: Key statistically significant findings for sound localization (confusions) for the group of subjects wearing earplugs|
Click here to view
| Discussion|| |
Conformity with SAE J994 and sound field behind vehicles
As stated previously, the broadband alarm meets the SAE J994  acoustical requirements if a literal interpretation of the definition of the predominant frequency is used. This definition, which states that "the predominant sound frequency of the alarm shall be defined as the frequency that produces the highest A-weighted sound pressure level", was initially introduced when only tonal alarms were available. However, such vague definition of predominant frequency can lead to various outcomes and interpretations when dealing with broadband signals. It would therefore be recommended to revisit the acoustical requirements and analysis methods proposed in the standard to better cover the variety of backup alarm signals commercially available.
In this study, the alarms were adjusted to levels ranging from 97 to 109 dBA at 1 m to meet ISO 9533  requirements (SNR > 0 dB at seven positions behind the vehicles) for Type B and C alarms typically used in noisy workplaces. To meet such requirements, levels were 5-10 dB higher for the tonal alarm compared to the broadband alarm. Such results stem directly from considerable level variations noted from one microphone to the other [Table 1] for the tonal alarm. In addition, sound pressure level contour maps reveal important variations in excess of 15 dB over short distances of less than 1 m for the tonal alarm and smaller variations (~7-8 dB) for the multi-tone alarm [Figure 6] and [Figure 7]. On the other hand, a quite homogeneous field, monotonically decreasing with distance, is observed for the broadband alarm. Similar results were obtained for the different mounting conditions tested and from one site to another. High SPL variations can increase the risk of accidents for workers in the danger zone behind vehicles. When a vehicle is approaching, an increase in SPL is expected by the worker. In a highly non-uniform sound field such as the one produced by the tonal alarm, a worker could sense a decrease in alarm level even though the vehicle is approaching and erroneously interpret this situation as a sign that the vehicle is moving away instead of backing up toward oneself. Moreover, when in areas of low alarm levels due to destructive acoustic interferences, workers may incorrectly estimate the distance of the approaching vehicle or misjudge the urgency of the signal or they may altogether fail to detect the alarm. Alarms are often set to exceedingly high output levels to overcome some of these problems, in turn causing undesirable behaviors such as disconnecting or modifying the alarm. High output alarms can also be an important source of annoyance for people working or living outside the danger zone, but in close proximity to heavy vehicle traffic.
Across all conditions tested, average sound detection thresholds varied from −13 to −24 dB SNR, indicating that alarms remain audible when adjusted at levels significantly below that of the background noise [Figure 8]. Accordingly, backup alarms should be clearly audible when adjusted based on ISO 9533 recommendations (SNR ≥ 0 dB). Indeed, according to various authors , auditory warning signals should be adjusted to levels 12-25 dB above the masked detection threshold for optimal use. As per ISO 9533, alarm levels are evaluated using a single background noise scenario with the vehicle running at full idle. Realistically, however, a vehicle can operate at various idling speeds and other engine regimes, with other noise sources in proximity of workers also contributing to the overall noisy background; hence, lower or higher alarm levels may be required depending on the situation. Alarms, which automatically adjust their level according to the background noise, may therefore be warranted in workplaces where various situations prevail.
Detection threshold results also show that the tonal alarm presents a certain advantage over the broadband signal, with up to 5-7 dB lower average thresholds in high-frequency rich noises (Noises 3 and 4) when hearing protection is worn. Across all other testing conditions, such an advantage of the tonal alarm over the broadband alarm reached a maximum of 3-4 dB. From a practical standpoint, this laboratory advantage in sound detection for the tonal alarm is offset by the considerable sound pressure level variations (in excess of 15 dB) noted over short distances in the sound propagation patterns in the field. Given the much more uniform sound distribution of the broadband alarm, the lower detection thresholds for the tonal alarm during laboratory measures would not represent a real advantage over the broadband signal for alarm detection by workers in the field. The same could be said for the multi-tone alarm; despite lower detection thresholds compared to the broadband alarm in some conditions (in Noises 3 and 4), its propagation behind heavy vehicles is also less uniform than that for the broadband signal.
With regards to loudness, the broadband alarm could be adjusted, on average, at levels 2-4 dB lower without hearing protection and 1 dB lower with protection to be judged as equally loud as a reference tonal alarm set at a SNR of 0 dB in various background noises ranging from 81 to 89 dBA [Figure 9]. In other words, when adjusted to identical levels well above masked detection thresholds in high noise levels, the broadband alarm seems louder than the tonal alarm while the multi-tone seems softer than both the broadband and tonal alarms. Equal loudness results obtained in this study are consistent with those reported by Scharf and Fishken  for pure tones compared with broadband noises. Despite being detected at higher levels in background noises than the tonal alarm, the broadband signal's loudness grows more rapidly as a function of level due to its larger frequency bandwidth.
The single most important factor acting upon perceived urgency is the presentation level (or SNR) of the alarms, as shown in [Figure 10], with perceived urgency being rated consistently higher at the highest presentation level (+6 dB SNR) compared with the lowest presentation level (−6 dB SNR), with and without both HPD types. Based on the results, the average slope of the function relating the increase in perceived urgency (scale of 0-100 units) with increases in SNR (dB) in fixed noises was approximately 5 units/dB without HPDs and 4 units/dB with HPDs. These estimated slopes allow determining the required level adjustment necessary to offset a difference in perceived urgency. The tonal alarm was judged to be slightly more urgent than the broadband alarm in 19 out of the possible 24 comparisons between both alarm types (of which only 2 reached statistical significance at lower levels in Noise 3), with differences ranging from 0 to 20 units. High data variability precluded most differences from reaching statistical significance. The maximum advantage of the tonal alarm over the broadband alarm reached 20 units and can therefore be equaled to a 4-dB advantage in presentation level. Stated otherwise, the tonal alarm's level could be reduced up to 4 dB to convey the same degree of urgency as the broadband alarm. However, as previously conjectured for detection thresholds, such an advantage for the tonal alarm in conveying perceived urgency during laboratory measures is not sufficiently important to offset the larger sound pressure level variations (in excess of 15 dB) measured in the field and would therefore not represent a real benefit in the field. For example, the sound pressure level variations of 10-15 dB measured over short distances behind the vehicles for the tonal alarm would correspond to large variations of 50-75 units in perceived urgency, which would overshadow the smaller laboratory advantage found over the broadband alarm.
It must be noted that the laboratory measures of perceived urgency performed in this study only took into account some of the acoustical characteristics (alarm type, noise background and hearing protection) that can influence the perceived urgency of alarms in the field. Even though the participants were not exposed to real dangers, they were well informed of the study context and hence there was no confusion as to the meaning of the stimuli (backup alarms) being presented. Therefore, the study does not directly address the effective urgency conveyed by the alarms outside the laboratory setting, where workers or passersby must not only perceive the signal, but also quickly recognize its particular meaning. When questioned on factors influencing their urgency ratings, many participants replied that they attributed greater urgency to the tonal alarm since it is a more familiar warning signal than the broadband alarm. Based on previous work with warning signals, it can be anticipated that familiarization with the broadband alarm could impact its perceived degree of urgency. ,,
Overall, localization performance is superior with the broadband alarm compared with the other two alarm types. Localization is also significantly better in the L/R dimension (speakers behind) than in the F/B dimension (speakers to the side) for all alarms, a result congruent with current literature.  The percentage of confusions is relatively low in the L/R dimension, the worst condition being that of the tonal alarm combined with the use of earmuffs (with about 20% confusions). As expected, differences in performance across the three alarms are greatest in the F/B dimension. For tonal and multi-tonal alarms, participants were confounded between front and back one out of three times (33%) to one out of two times (50%), which in the latter case corresponds to chance performance. With the broadband signal, confusions were less frequent without HPDs (about 10%) and with earplugs (about 18%), but increased significantly with earmuffs (to 40%). Such results highlight the hindering effect of earmuffs on sound localization compared to performance unprotected or with earplugs. Again, results are consistent with those of the literature and support findings of better F/B localization performance with broadband signals containing high-frequency spectral information. , Subjects were not allowed head movements in order to assess sound source-identification in the most challenging situation that could occur in the workplace, where workers may not be free or have the reflex or time to use head movements to resolve confusions. Other factors that cannot be ignored when addressing safety issues include, among others: acoustical reflections and diffractions during sound propagation, temporal, spectral and spatial variations of the ambient noise, workers' hearing status and reaction time required for head movements.
No alarm presented advantages over the others in all aspects investigated within this study. While the broadband alarm can be set to lower levels than the tonal alarm to achieve equal loudness and is easier to localize in space, it may suffer from a larger masking effect with background noises exhibiting greater energy in the high-frequency region (Noises 3 and 4). Whereas, the tonal alarm was generally easier to detect in workplace noise and was judged to be slightly more urgent than the broadband alarm at lower presentation levels in the laboratory, such advantages are more than offset by the considerable sound pressure level variations noted over short distances in the sound propagation patterns in the field. Results described above are applicable in background noises similar to those used in this study, ranging in levels from 81 to 89 dBA and should not necessarily be generalized to other noises with different acoustical characteristics (level, spectrum and temporal fluctuations).
| Conclusions|| |
Comparisons of sound field measurements and psychoacoustic data on three types of backup alarms are presented in this paper. Alarms with broad frequency content appear to present some advantages over conventional tonal alarms, including: (i) A much more uniform sound propagation pattern behind vehicles; (ii) lower alarm output levels to meet the requirements set forth in ISO 9533; (iii) higher urgency ratings at high SNR without HPDs and; (iv) better sound localization performance. However, some disadvantages were also noted. Firstly, higher SNR are required for detection of the broadband alarm, at least in noises rich in high-frequency content. Secondly, detection thresholds and urgency ratings appear to be more severely affected by the use of HPDs for the broadband alarm than the tonal alarm. Overall, however, the most salient finding is that the broadband alarm yields a more uniform sound field behind heavy vehicles than the conventional tonal alarm and that this advantage overshadows smaller alarm differences found in the laboratory.
In general, HPDs can improve alarm detection thresholds in noisy environments, at least for individuals with normal hearing. They can however negatively impact the perception of loudness and urgency, by lowering the ratings. As such, careful selection of HPDs is warranted in order to prevent instances of overprotection. Moreover, earmuffs were shown to significantly compromise sound localization compared to earplugs.
Nowadays, a number of backup alarms devices are equipped with sensors measuring the ambient noise levels in order to be able to produce an audible signal proportional to these levels- a self-adjusting backup alarm. The ISO 9533  standard was recently modified to incorporate a procedure to evaluate if such devices effectively generate alarm sound levels higher than the ambient noise (SNR > 0). However, in light of the results presented in this paper, work is needed to study how self-adjusting backup alarms are perceived (psychoacoustic dimensions) and to determine the performance of the adjustment procedure on the field during real operating conditions.
Finally, the findings and general trends presented in the paper must be interpreted with caution. Additional data on a greater number of subjects, more representative tests conditions (head movements, multiple noise sources, hearing loss, etc.) and more comprehensive analyses are required to draw firm conclusions from such findings. While this study did not reveal particular contraindications to the use of the broadband alarm on moving vehicles and even found some significant advantages in sound localization and sound propagation pattern nearby vehicles, results need to be more thoroughly analyzed from a work safety standpoint, taking into consideration other factors such as environmental annoyance, habituation and familiarity of the alarms. The conveyed meaning of the broadband signal, as perceived by workers and the general population and its recognition as an alarm signal as such are particularly important to consider when deploying this new type of alarm.
| References|| |
|1.||Occupational Safety and Health Administration (OSHA). Occupational safety and health standards: Motor vehicles (29 CFR, Part 1926.601). Washington, DC: Office of the Federal Register; 2000. |
|2.||SAE. Alarm - Backup ¯ Electric Laboratory Performance Testing. Warrendale, PA, USA: Society of Automotive Engineering, SAE J994; 2009. |
|3.||ISO. Earth-moving Machinery-Machine-mounted Forward and Reverse Audible Warning Alarm--Sound Test Method. Geneva, Switzerland: International Standards Organization, ISO 9533; 2010. |
|4.||Blouin S. Compendium of knowledge on personnel detection devices during vehicle backup maneuvers on construction sites. Montreal, Canada: Research Report B-067/IRSST [In French]; 2005. |
|5.||Laroche C, Denis S. Compendium of knowledge on acoustic signaling and forklift trucks. Internal Report. Montreal, Canada: IRSST [In French]; 2000. 64 p. |
|6.||Laroche C, Hétu R, L'Espérance A. Backup alarms can kill! Travail et Santé (Work and Health) [In French] 1991;7:9-13. |
|7.||Laroche C, Ross MJ, Lefebvre L, Larocque R. Determination of the optimal acoustic characteristics of backup alarms. Montreal. Canada: Research Report R-117/IRSST [In French]; 1995. |
|8.||Murray W, Mills J, Moore P. Reversing Accidents in UK. Transport Fleets 1996-97. UK: Transport and Logistics Research Unit, University of Huddersfield; 1998. |
|9.||NIOSH. The Worker Health Chartbook 2004. Cincinnati, OH, USA: NIOSH Publication 2004-146; 2004. |
|10.||Purswell JP, Purswell JL. The effectiveness of audible backup alarms as indicated by OSHA accident investigation records. In: Bittner AC, Champney PC, Morrissey SJ, editors. Advances in Occupational Ergonomics and Safety. Amsterdam, The Netherlands: ISO Press; 2001. p. 444-50. |
|11.||Laroche C, Lefebvre L. Improvement in the acoustic characteristics of reverse alarms used on vehicles. Proceedings of the 13 th Triennal Congress of the International Ergonomics Association. Tempere, Finland: Finnish Institute of Occupational Health; 1997. |
|12.||Laroche C, Lefebvre L. Determination of optimal acoustic features for reverse alarms: Field measurements and the design of a sound propagation model. Ergonomics 1998;41:1203-21. |
|13.||Laroche C. Investigation of an accident involving the reverse alarm on a heavy vehicle. Proceedings IEA 2006. Maastricht, Netherlands: Elsevier; 2006. |
|14.||Heckman GM, Kim RS, Khan FS, Bare C, Yamaguchi GT. Auditory localization of backup alarms: The effects of alarm mounting location. Warrendale, PA: SAE Technical Paper 2011-01-0086; 2011. |
|15.||Bliss JP, Gilson RD, Deaton JE. Human probability matching behaviour in response to alarms of varying reliability. Ergonomics 1995;38:2300-12. |
|16.||Bliss JP, Dunn MC. Behavioural implications of alarm mistrust as a function of task workload. Ergonomics 2000;43:1283-300. |
|17.||Morgan HP, Peppin RJ. "Noiseless" and safer back-up alarms. Proceedings of the NOISE-CON 2008. Dearborn, MI, USA: Curran Associates, Inc; 2008. |
|18.||Schexnayder CJ, Ernzen J. Effective noise control during nighttime construction. Making Work Zones Work Better Workshops Series. USA: Federal Highway Administration, US Department of Transportation; 2004. |
|19.||Noble WG. Earmuffs, exploratory head movements, and horizontal and vertical sound localization. J Aud Res 1981;21:1-12. |
|20.||Blauert J. Spatial Hearing: The Psychophysics of Human Sound Localization. Cambridge, MA, USA: MIT Press; 1997. |
|21.||Carlile S, King AJ. From outer ear to virtual space. Curr Biol 1993;3:446-8. |
|22.||Hartmann WM. How we localize sound. Phys Today 1999;52:24-9. |
|23.||Middlebrooks JC, Green DM. Sound localization by human listeners. Annu Rev Psychol 1991;42:135-59. |
|24.||Berger EH. Hearing protection devices. In: Berger EH, Layne M, editors. The Noise Manual. 5 th ed. Fairfax, VA, USA: AHIA Press; 2003. p. 379-454. |
|25.||Bolia RS, D'Angelo WR, Mishler PJ, Morris LJ. Effects of hearing protectors on auditory localization in azimuth and elevation. Hum Factors 2001;43:122-8. |
|26.||Simpson BD, Bolia RS, McKinley RL, Brungart DS. The impact of hearing protection on sound localization and orienting behavior. Hum Factors 2005;47:188-98. |
|27.||Tran Quoc H, Hétu R. Acoustic planning and signaling in industrial workplaces: Design criteria of acoustic warning signals. Can Acoust [In French] 1996;24:3-17. |
|28.||Burgess M, McCarty M. Review of Alternatives to "Beeper" Alarms for Construction Equipment. Canberra, Australia: Acoustics & Vibration Unit, School of Aerospace, Civil & Mechanical Engineering, UNSW; 2009. p. 68. [Report No.: AVU 0129]. |
|29.||Committee on Technology for a Quieter America. Technology for a Quieter America. Washington, DC, USA: National Academy of Engineering, The National Academies Press; 2010. |
|30.||Haas EC, Edworthy J. The perceived urgency and detection time of multitone auditory signals. In: Stanton NA, Edworthy J, editors. Human Factors in Auditory Warnings. Ashgate, England: Gower Technical; 1998. |
|31.||Withington DJ. Reversing goes broadband. Quarry Manage 2004;31:27-34. |
|32.||Homer JP. Audible warning devices used in the mining industry. Proceedings of the Noise-Con 2008. Dearborn, MI, USA: Curran Associates, Inc; 2008. |
|33.||ISO. Ergonomics - Danger Signals for Public and Work Areas - Auditory Danger Signals. Geneva, Switzerland: International Standards Organization, ISO 7731;2003. |
|34.||Bassett Consulting Engineers. Broadband Auditory Warning Alarms. Technical Report for SA Department for Transport, Energy and Infrastructure, doc AA0981-A9B01RP; 2009. |
|35.||Casali JG, Robinson GS, Dabney EC, Gauger D. Effect of electronic ANR and conventional hearing protectors on vehicle backup alarm detection in noise. Hum Factors 2004;46:1-10. |
|36.||Howell K, Martin AM. An investigation of the effects of hearing protectors on vocal communication in noise. J Sound Vib 1975;41:181-96. |
|37.||Berger EH. Hearing protection devices. In: Berger EH, Royster LH, Royster JD, Driscoll DP, Layne M, editors. The Noise Manual, 5 th ed. Fairfax, VA: American Industrial Hygiene Association; 2003. |
|38.||Kryter KD. Effects of ear protective devices on the intelligibility of speech in noise. J Acoust Soc Am 1946;18:413-7. |
|39.||Suter AH. Communication and Job Performance in Noise: A Review. ASHA Monographs Number 28. Rockville, MD: Amer Speech Language; 1992. |
|40.||Robinson GS, Casali JG. Audibility of reverse alarms under hearing protectors for normal and hearing-impaired listeners. Ergonomics 1995;38:2281-99. |
|41.||Giguère C, Laroche C, Vaillancourt V. Research on modeling the effects of personal hearing protection and communication devices on speech intelligibility in noise. Defence R&D Canada, DRDC Contract Report. DRDC Toronto CR 2011-101, Toronto, Canada; 2011. p. 70. |
|42.||Casali JG, Wright WH. Do amplitude-sensitive hearing protectors improve detectability of vehicle backup alarms in noise. Human Factors and Ergonomics Society Annual Meeting Proceedings. Human Factors and Ergonomics Society; 1995. p. 994-8. |
|43.||Lovejoy SM. Determination of Backup Alarm Masked Threshold in Construction Noise. Master of Science Thesis. Blacksburg, Virginia, USA: Industrial and Systems Engineering, Faculty of the Virginia Polytechnic Institute and State University; 2008. |
|44.||Christian E. The Detection of Warning Signal While Wearing Active Noise Reduction and Passive Hearing Protection Devices. Master of Science thesis. Blacksburg, Virginia, USA: Industrial and Systems Engineering, Faculty of the Virginia Polytechnic Institute and State University; 1999. |
|45.||Alali KA. Azimuthal Localization and Detection of Vehicular Backup Alarms Under Electronic and Non-electronic Hearing Protection Devices in Noisy and Quiet Environments. Ph.D. thesis. Blacksburg, Virginia, USA: Virginia Polytechnic Institute and State University; 2011. |
|46.||Withington DJ. Paterson SE. Safer sirens. Fire Eng J 1998;1:1-5. |
|47.||Withington DJ. The quest for better ambulance sirens. Ambulance UK 1996;11:20-1. |
|48.||Withington DJ. Localisable alarms. In: Stanton NA, Edworthy J, editors. Human Factors in Auditory Warnings. England: Ashgate Publishing Ltd.; 1999. |
|49.||Withington DJ. The use of directional sound to improve the safety of auditory warnings. XIV Triennial Congress of the International Ergonomics Association & 44 th Annual Meeting of the Human Factors and Ergonomics Society. San Diego, CA, USA: Human Factors and Ergonomics Society; 2000. |
|50.||Lakatos S, Miller G. Psychoacoustic evaluation of listener localization accuracy for broadband and conventional reversing alarms. Proceedings of the 16 th International Congress on Sound and Vibration. Krakow, Poland: International Institute of Acoustics and Vibration 2009. |
|51.||Catchpole KR, McKeown JD, Withington DJ. Localizable auditory warning pulses. Ergonomics 2004;47:748-71. |
|52.||Abel SM, Hay VH. Sound localization. The interaction of aging, hearing loss and hearing protection. Scand Audiol 1996;25:3-12. |
|53.||Atherley GR, Noble WG. Effect of ear-defenders (ear-muffs) on the localization of sound. Br J Ind Med 1970;27:260-5. |
|54.||Berger EH, Casali JG. Hearing protection devices. In: Crocker MJ, editor. Encyclopedia of Acoustics, 1 st ed. New York, NY: John Wiley and Sons; 1997. p. 967-81. |
|55.||Noble WG, Russell G. Theoretical and practical implications of the effects of hearing protection devices on localization ability. Acta Otolaryngol 1972;74:29-36. |
|56.||Noble W, Murray N, Waugh R. The effect of various hearing protectors on sound localization in the horizontal and vertical planes. Am Ind Hyg Assoc J 1990;51:370-7. |
|57.||Alali KA, Casali JG. The challenge of localizing vehicle backup alarms: Effects of passive and electronic hearing protectors, ambient noise level, and backup alarm spectral content. Noise Health 2011;13:99-112. |
|58.||Casali JG, Alali KA. Etymotic EB-15 (Lo Position) BlastPLG Evaluation: Backup Alarm Localization Appended Experiment. Blacksburg, Virginia, USA: Virginia Tech; 2010. [Audio Lab Report No. 6/9/10-2-HP, ISE Dept. Report No. 201002]. |
|59.||Edworthy J, Stanton N. A user-centred approach to the design and evaluation of auditory warning signals:1. Methodology. Ergonomics 1995;38:2262-80. |
|60.||Edworthy J, Loxley S, Dennis I. Improving auditory warning design: Relationship between warning sound parameters and perceived urgency. Hum Factors 1991;33:205-31. |
|61.||Haas EC, Casali JG. Perceived urgency of and response time to multi-tone and frequency-modulated warning signals in broadband noise. Ergonomics 1995;38:2313-26. |
|62.||Hellier E, Edworthy J. Quantifying the perceived urgency of auditory warnings. Can Acoust 1989;17:3-11. |
|63.||SAE. Discriminating Back-up Alarm System. Warrendale, PA, USA : Society of Automotive Engineering, SAE J1741; 1999. |
|64.||Ahmad MR, Werner C, Brunner E. Analysis of high-dimensional repeated measures designs: The one sample case. Comput Stat Data Anal 2008;53:416-27. |
|65.||Box GE. Non-normality and tests on variances. Biometrika 1953;40:318. |
|66.||Strivastava MS. Some tests concerning the covariance matrix in high-dimensional data. J Jpn Stat Soc 2005;35:251-72. |
|67.||Vaillancourt V, Laroche C, Giguère C, Beaulieu MA, Legault JP. Evaluation of auditory functions for royal canadian mounted police officers. J Am Acad Audiol 2011;22:313-31. |
|68.||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. |
|69.||Scharf B, Fishken D. Binaural summation of loudness: Reconsidered. J Exp Psychol 1970;86:374-9. |
|70.||Burt JL, Bartolome DS, Burdette DW, Comstock JR Jr. A psychophysiological evaluation of the perceived urgency of auditory warning signals. Ergonomics 1995;38:2327-40. |
|71.||Guillaume A, Pellieux L, Chastres V, Drake C. Judging the urgency of nonvocal auditory warning signals: Perceptual and cognitive processes. J Exp Psychol Appl 2003;9:196-212. |
|72.||Petocz A, Keller PE, Stevens CJ. Auditory warnings, signal-referent relations, and natural indicators: Re-thinking theory and application. J Exp Psychol Appl 2008;14:165-78. |
|73.||Butler RA. The bandwidth effect on monaural and binaural localization. Hear Res 1986;21:67-73. |
|74.||Makous JC, Middlebrooks JC. Two-dimensional sound localization by human listeners. J Acoust Soc Am 1990;87:2188-200. |
505, Boul. de Maisonneuve Ouest, Montréal, QC, H3A 3C2
Source of Support: The work was funded by the Institut de Recherche
Robert-Sauvé en Santé et Sécurité du Travail as stated in the
acknowledgement section of the paper., Conflict of Interest: None
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7], [Figure 8], [Figure 9], [Figure 10], [Figure 11]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5], [Table 6], [Table 7]
|This article has been cited by|
||Influence of sound spatial reproduction method on the detectability of reversing alarms in laboratory conditions
| ||Olivier Valentin, Pierre Grandjean, Clément Girin, Philippe-Aubert Gauthier, Alain Berry, Étienne Parizet |
| ||Acta Acustica. 2023; 7: 9 |
|[Pubmed] | [DOI]|
||Facilitating or disturbing? An investigation about the effects of auditory frequencies on prefrontal cortex activation and postural sway
| ||Valeria Belluscio, Giulia Cartocci, Tommaso Terbojevich, Paolo Di Feo, Bianca Maria Serena Inguscio, Marco Ferrari, Valentina Quaresima, Giuseppe Vannozzi |
| ||Frontiers in Neuroscience. 2023; 17 |
|[Pubmed] | [DOI]|
||Experimental assessment of the effect of wearing hearing protectors on the audibility of railway warning signals for normal hearing and hearing impaired listeners
| ||Jean-Pierre Arz, Nicolas Grimault, Ossen El Sawaf |
| ||International Journal of Occupational Safety and Ergonomics. 2021; : 1 |
|[Pubmed] | [DOI]|
||Effect of Hearing and Head Protection on the Localization of Tonal and Broadband Reverse Alarms
| ||Chantal Laroche, Christian Giguère, Véronique Vaillancourt, Claudia Marleau, Marie-France Cadieux, Karina Laprise-Girard, Emily Gula, Véronique Carroll, Manuelle Bibeau, Hugues Nélisse |
| ||Human Factors: The Journal of the Human Factors and Ergonomics Society. 2021; : 0018720821 |
|[Pubmed] | [DOI]|
||Towards autonomous cloud-based close call data management for construction equipment safety
| ||Olga Golovina, Jochen Teizer, Karsten W. Johansen, Markus König |
| ||Automation in Construction. 2021; 132: 103962 |
|[Pubmed] | [DOI]|
||Acoustic characterization of tonal and broadband backup alarms in laboratory and field conditions
| ||Olivier Robin, Tamara Krpic, Hugues Nélisse, Alain Berry |
| ||Applied Acoustics. 2020; 163: 107228 |
|[Pubmed] | [DOI]|
||Improving the prevention of fall from height on construction sites through the combination of technologies
| ||María del Carmen Rey-Merchán, Jesús M. Gómez-de-Gabriel, Juan-Antonio Fernández-Madrigal, Antonio López-Arquillos |
| ||International Journal of Occupational Safety and Ergonomics. 2020; : 1 |
|[Pubmed] | [DOI]|
||Industrial Workers Perception of Reverse Motion Warning Devices on Forklifts
| ||Yousif Abulhassan, David Wilbanks, Richard Kilpatrick, Tyler Howell |
| ||Proceedings of the Human Factors and Ergonomics Society Annual Meeting. 2019; 63(1): 1861 |
|[Pubmed] | [DOI]|
||Did you hear that? The role of stimulus similarity and uncertainty in auditory change deafness
| ||Kelly Dickerson,Jeremy R. Gaston |
| ||Frontiers in Psychology. 2014; 5 |
|[Pubmed] | [DOI]|