Summary

Noisy industrial environments interfere with the audibility of alarms and other audible warning devices and thus compromise the safety of people working in the presence of moving or dangerous machinery.

In many cases where accidents have occurred, the vehicles involved had reverse alarms. Generally, the effectiveness of sound alarms depends on the sound environment, and they are far from infallible. The exact reasons are difficult to determine, because perception and localization of sound in a complex sound environment are phenomena that are, for the most part, little known and little studied. Although studies on perception and localization of alarms have been carried out in terms of occupational health and safety, it is difficult to do such research with human subjects on worksites.

The objective of this study was to examine the possibility of using current sound field reproduction methods to re-create sound environments in the laboratory that are representative of work environments, thus avoiding the many problems inherent in field tests. The use of such techniques to conduct trials with human subjects in the near future is under consideration. An example of such trials would be to study the localization and audibility of reverse alarms in simulated sound environments under controlled conditions, in order to carry out detailed parametric studies with human subjects.

The project was divided into two main parts. First, two representative sound environments (one indoors and one outdoors) were measured with an 85-microphone array, to record the soundscape and the spatial distribution of sound so as to reconstruct the arrival directions of all the sounds that make up the sound environment. To that end, two sites were visited: an open-pit chalk quarry (Graymont, Bedford) and a manufacture and assembly workshop (Agrigratte, St Jacques). At each site, sound environments with several types of sound sources were captured, which resulted in several hours of recordings.

Secondly, this time in the laboratory, sound environments were reproduced using various algorithms that were compared. These reproductions were done using the Wave Field Synthesis (WFS) system of the Groupe Acoustique de l’Université de Sherbrooke (GAUS), made up of 96 loudspeakers and 4 subwoofers. This system makes it possible to use classic techniques of sound field reproduction, and to develop (as was the case for this study) new algorithms or tools according to needs. Before performing the measurements and physical evaluations of the sound environments reproduced in the laboratory, theoretical simulations were also performed to compare the algorithms, to determine the best parameters and to visualize the results for three simple cases (a) two sources in free field; (b) one source in movement; and (c) a diffuse environment. These simulations showed that the two best algorithms were (1) inverse problem approaches with spherical reproduction sources, and (2) the LASSO method with spherical reproduction sources. To objectively assess the reproduction, various metrics were used and are described in this report: reference microphone spectra, acoustic pressure levels on the microphone array, spatial acoustic maps in time and, in the case of simulations, instantaneous sound pressure fields. It was proven that the proposed methods can reproduce the sound environments measured with the microphone array. Moreover, various demonstrations were carried out for listening during the GAUS visits. The goal of these demonstrations is to provide examples conducive to listening to illustrate, in practice, the full potential of the technologies and methods tested in the context of a study related to sound perception in the occupational health and safety field.

Several conclusions emerge from the study. Firstly, using the inverse method with spherical reproduction sources currently provides the best results with respect to objective assessment. Secondly, this approach, while physically correct and valuable, can sometimes give the impression of excessive diffusion (on the basis of informal listening). The LASSO algorithm was tested to potentially mitigate this problem by simultaneously limiting the number of active reproduction sources. Despite the fact that the LASSO is more demanding than the inverse problem approach in terms of calculation resources, a number of gains were observed: substantial capacity to tighten the re-created spatial imagery and the possibility of working with fewer microphones for recording. The LASSO algorithm could greatly simplify the exercise of future recordings on sites that are often congested and very busy.

The main finding is that the GAUS WFS system can be used to reproduce sound environments representative of the workplace for future trials using human subjects. One avenue of research could be to use other physical or psychophysical metrics, to carry out the physical assessment (without a human subject) of the reproduction quality or to characterize the targeted and reproduced environments. These are gateways that may lead to future technological research. This research project led to the creation and validation of an acoustic simulation platform for sound environments in the workplace. It is now available to carry out broad research campaigns dealing with the perception and localization of alarms and other audible warning devices, which are important for occupational safety in the workplace, using human subjects in various positions (standing, sitting, in movement, and even with the listening plane set at different heights). Many findings and practical recommendations may emerge to enhance perceptibility and localization of warning devices, because studies with controlled environmental parameters and sources can now be carried out safely with workers.