Subjective evaluation in acoustics through listening testsdifference testing in sound insulation, room acoustics and soundscapes

Resumen

In the different areas of acoustics, various measures and indicators are used nowadays to assess the acoustic behaviour of environments and their impact on the lifes of human beings. Although the metrics used in the different areas have been described over the decades attempting to take into account the human perception of sound, in recent years a paradigm shift is becoming more evident, focusing on developing better parameters, as well as improving existing ones, in order to ensure that they are as representative as possible of the subjective perception of sound. To this end, subjective tests are being carried out in which samples of the population are consulted about their subjective perception of different sound stimuli. The purpose of these tests, which can be carried out in real environments (in situ surveys) or in a laboratory (listening tests), is to find out from their responses how the human auditory system behaves in order to assess the adequacy of current indicators and to make improvements when necessary. The main obstacle for these tests to satisfactorily achieve this goal is the high heterogeneity of methodologies used to perform them, which leads to different investigations being subject to different sources of bias and analysis methods, hindering the comparison of their results. This fact reduces the relevance of their conclusions and makes it difficult to carry out meta-analyses based on the results of different tests, which would enable to obtain improvements in the indicators, making them representative of all the acoustic situations assessed by the different perceptual tests. The main objective of this thesis is to provide methodological contributions that facilitate the definition of homogeneous, accurate and minimally biased methodologies with highly informative analysis tools for the performance of perceptual tests in laboratory environments (listening tests) in the areas of sound insulation, room acoustics and soundscapes. These methodological contributions were identified after a detailed study of the types of listening tests currently employed in each of these areas, evaluating their advantages, disadvantages and main sources of bias, as well as studying in depth the state of the art in other areas of sensory evaluation where these types of tests have a greater background and where the quality of the different methodologies has, to date, been more thoroughly studied. Accordingly, specific methodological contributions were described for the design, performance and analysis of listening tests particularly appropriate for the most relevant purposes in each of the areas. These contributions, based on comparative difference testing methods, not only focused on a methodological homogenization in each of the three areas, but also on a transversal homogenization, which could facilitate the description of a common methodological framework based on difference testing that would allow the use of highly informative analysis methods barely employed so far in acoustics such as the Thurstonian models. These contributions were evaluated through the performance of three listening tests, one in each of the study areas. In the area of sound insulation, a listening test following an attribute-related 2-AC difference testing protocol was carried out on a large sample of 119 participants with the aim of assessing the annoyance perceived by them in the presence of different types of urban noise under different façade sound insulation conditions. The results of this test allowed not only to evaluate the representativeness of the current Single-Number Quantities of sound insulation (SNQs), but also to propose an improvement of these through the application of an automatic optimization algorithm based on Sequential Quadratic Programming (SQP), using as input parameters the results of the test analyzed by means of Thurstonian models. The application of the optimization algorithm enabled to obtain a reference spectrum for the calculation of a more representative SNQ of urban sounds (Lopt-urban) than any of the existing ones. On the other hand, in the area of room acoustics, a listening test was carried out in order to evaluate the operational power of seven overall difference testing protocols, some of them commonly used for listening tests in this area and others potentially applicable to room acoustics but, to date, usually applied in other fields of sensory evaluation. This listening test, in which a large sample of 134 participants evaluated, using these protocols, the perceptual difference between different confusable auralizations of the same room, allowed a comparison of the operational power of the various protocols using Thurstonian models. This comparison allowed to identify and recommend the protocol with the best operational power and the least influence of the various sources of bias. Finally, in the area of soundscapes, two novel listening test tasks were developed, also based on difference testing, for the determination of perceptual similarity between different soundscapes. The usefulness of these tasks was tested by means of a listening test conducted on a small population of 17 participants who compared urban soundscapes with different architectural features. These soundscapes were also objectively classified using automatic clustering algorithms (i.e., K-means and hierarchical clustering) based on percentiles of the psychoacoustic indicators Loudness (N5, N50 and N95), Sharpness (S5, S50 and S95), Roughness (R10, R50 and R95) and Fluctuation Strength (FS10, FS50 and FS95). Subsequently, the results of this clustering were compared with those obtained by the listening test, finding a good correlation, in most cases, and highlighting not only the applicability of the listening test tasks developed but also the usefulness of psychoacoustic indicators to assess the perceptual similarity of urban soundscapes with similar architectural characteristics.