A method to collect representative samples of urban soundscapes
Invited paper
Octogone EA 4156
Tuesday 2 june, 2015, 15:00 - 15:20
0.6 Madrid (49)
Abstract:
Many studies of urban soundscapes have been performed with different aims ranging from auditory perception, to evaluation of noise annoyance, sound design for public places, speech learning, and urban studies, among others. A typical first step of most studies involves collecting soundscapes recordings in sound databases and/or making new recordings of soundscapes. While this initial collection of samples is crucial insofar as it will have a strong impact on the validity of the study, there is one major limitation: it is a selection and therefore it is the result of a choice from the scientist’s own assessment of what is representative of the urban soundscape that is studied. The mental representation of urban soundscapes corresponds to the shared knowledge in a given language, a given culture and a given place at a given period of time. Therefore, a representative sample of an urban soundscape would be an example that is as close as possible to this mental representation, i.e. very typical. This paper addresses this issue with a method based on online questionnaire surveys. In this study, we collected French linguistic data through a large-scale questionnaire survey to find the shared knowledge about soundscapes in three types of streets at different times during the day. From a linguistic analysis of the responses, we elaborated profiles to be used in the field to guide recording sessions. These profiles are lists of sound sources and acoustic descriptions that are different for each type of street. We then recorded sound samples of urban soundscapes in the field and annotated sound sources to evaluate the distance between the samples and the linguistic profiles. The proposed method as well as the results in the case of urban soundscapes in Toulouse (France) are likewise presented.