Analysing Interlinked Frequency Dynamics of the Urban Acoustic Environment

Abstract

As sustainable metropolitan regions require more densely built-up areas, a comprehensive understanding of the urban acoustic environment (AE) is needed. However, comprehensive datasets of the urban AE and well-established research methods for the AE are scarce. Datasets of audio recordings tend to be large and require a lot of storage space as well as computationally expensive analyses. Thus, knowledge about the long-term urban AE is limited. In recent years, however, these limitations have been steadily overcome, allowing a more comprehensive analysis of the urban AE. In this respect, the objective of this work is to contribute to a better understanding of the time–frequency domain of the urban AE, analysing automatic audio recordings from nine urban settings over ten months. We compute median power spectra as well as normalised spectrograms for all settings. Additionally, we demonstrate the use of frequency correlation matrices (FCMs) as a novel approach to access large audio datasets. Our results show site-dependent patterns in frequency dynamics. Normalised spectrograms reveal that frequency bins with low power hold relevant information and that the AE changes considerably over a year. We demonstrate that this information can be captured by using FCMs, which also unravel communities of interlinked frequency dynamics for all settings.

Acoustic Characterization of Potential Quiet Areas in Dortmund, Germany

“…In the summer, dawn biophony is still visible in most strata but the mean amplitude difference between deep night and the dawn chorus peak reduces by 2 to 4 dB(A). Past studies in the neighboring city of Bochum identified increases in the spring normalized mean power spectrum between frequencies of 3-9kHz that subsequently reduces, supporting our findings [87]. At sport and recreation facilities, autumn acoustic environments are clearly very active from morning until night, most likely explained by the wellknown popularity of soccer in Dortmund, with practices running from morning until well in the evening under lights.…”

Acoustic Characterization of Potential Quiet Areas in Dortmund, Germany

“…In the summer, dawn biophony is still visible in most strata but the mean amplitude difference between deep night and the dawn chorus peak reduces by 2 to 4 dB(A). Past studies in the neighboring city of Bochum identified increases in the spring normalized mean power spectrum between frequencies of 3-9 kHz that subsequently reduces, supporting our findings [87]. At sports and recreation facilities, autumn acoustic environments are clearly very active from morning until night, most likely explained by the well-known popularity of soccer in Dortmund, with practices running from morning until well in the evening under lights.…”

The Urban Acoustic Environment as a Complex Network

“…1). In the urban acoustic environment, high correlations between frequency bins characterize the prevalence of particular sound sources and therefore provide valuable information to distinguish between the AEs of different urban settings [8]. We calculate the FCMs using Pearson correlation [7]:…”

“…FCMs, CNs and their respective ASPL were then calculated between all frequency bins for each respective day. The frequency range for all data was limited to 13 kHz, as we found no substantial power above 13 kHz in the urban environment [8].…”

“…One of the main advantages of CNs is their ability to capture the emergent properties and behaviors of the system as a whole, rather than focusing solely on individual components [6]. Here, we apply CNs to quantify frequency correlation matrices (FCMs), which have shown to be a promising tool to depict environment specific interrelationships between power spectra in high spatial and temporal resolution [7,8]. The aim of this work is to give an intuitive understanding on how FCM based CNs in the urban AE work.…”