Machine Listening for Park Soundscape Quality Assessment

Abstract

The increasing importance attributed to soundscape quality in urban design generates a need for a system for automatic quality assessment that could be used for example in monitoring. In this work, the possibility for using machine listening techniques for this purpose is explored. The outlined approach detects the presence of particular sounds in a human-inspired way, and therefore allows to draw conclusions about how soundscapes are perceived. The system proposed in this paper consists of a partly recurrent artificial neural network modified to incorporate human attention mechanisms. The network is trained on sounds recorded in typical urban parks in the city of Antwerp, and thus becomes an auditory object creation and classification system particularly tuned to this context. The system is used to analyze a continuous sound level recording in different parks, resulting in a prediction of sounds that will most likely be noticed by a park visitor. Finally, it is shown that these indicators for noticed sounds allow to construct more powerful models for soundscape quality as reported in a survey with park visitors than indicators that are more regularly used in soundscape research.

A systematic review of prediction models for the experience of urban soundscapes

“…Consistently to the first layer of the descriptor hierarchy by Aletta et al, 2016 [3], [26] introduced the use of sound source identification as soundscape indicators accordingly to the inclusion criteria of the current study. Sound sources can either be evaluated by the participants to the study, or can be automatically assessed [31] and used as input layer for the models [45,32]. The contribution of the emotional impact of an environment can be identified through two attributes [27]: the physical surrounding and the implicit attributes of social aspects by including also explicit behavioural and implicit psychological factors.…”

“…The subjective evaluation of point 1) is usually collected through a rating task. However, this task is mainly a soundscape classification task [31,32] which does not address the perceptual identity of a soundscape, therefore, the studies modelling these as descriptors are excluded from the current review (see Sec. 2.2).…”

“…2.2). Other approaches rely on simulating the auditory stimuli response on individuals by moving the role covered by sound source presence into event saliency recognition [33,34,35,36,37,38], and by maintaining the cognitive and emotional factors on a higher level [32]. However, sound sources and event saliency recognition can still be considered as good indicators to fit the model.…”

“…In order to provide indicators and descriptors data for a soundscape model, in situ questionnaires, soundwalks and laboratory assessments are the most common strategies [39]. Out of the 24 references that will be presented later: 7 studies used in situ data collection approaches [27,40,28,41,21,42,43,32], 2 soundwalks [44,45], 9 laboratory experiments [46,47,48,49,50,51,52,53,54]. Six studies used a selected number of subjects assessing multiple soundscape data [21,55,56,57,58,59].…”

Estimation of the Perceived Time of Presence of Sources in Urban Acoustic Environments Using Deep Learning Techniques

“…In this paper, we do so by developing new indicators relying on source recognition models based on deep learning techniques, which has demonstrated state-of-the-art performance in many tasks studied in the machine listening community [27]. In the context of urban sound environments, machine listening has been successfully applied to sound event detection [28,29,30,31], sound scene classification [12,32,33,34,35] and soundscape quality evaluation [7,36,31]. A wide range of architectures exist, the most common of which are convolutional and recurrent neural networks [37].…”