AI-based soundscape analysis: Jointly identifying sound sources and predicting annoyancea)

Sound source classification for soundscape analysis using fast third-octave bands data from an urban acoustic sensor networka)

The exploration of the soundscape relies strongly on the characterization of the sound sources in the sound environment. Novel sound source classifiers, called pre-trained audio neural networks (PANNs), are capable of predicting the presence of more than 500 diverse sound sources. Nevertheless, PANNs models use fine Mel spectro-temporal representations as input, whereas sensors of an urban noise monitoring network often record fast third-octaves data, which have significantly lower spectro-temporal resolution. In a previous study, we developed a transcoder to transform fast third-octaves into the fine Mel spectro-temporal representation used as input of PANNs. In this paper, we demonstrate that employing PANNs with fast third-octaves data, processed through this transcoder, does not strongly degrade the classifier's performance in predicting the perceived time of presence of sound sources. Through a qualitative analysis of a large-scale fast third-octave dataset, we also illustrate the potential of this tool in opening new perspectives and applications for monitoring the soundscapes of cities.

Towards an interdisciplinary formalization of soundscapes

Soundscapes have been studied by researchers from various disciplines, each with different perspectives, approaches, and terminologies. Consequently, the research field determines the actual concept of a specific soundscape with the associated components and also affects the definition itself. This complicates interdisciplinary communication and comparison of results, especially when research areas are involved which are not directly focused on soundscapes. For this reason, we present a formalization that aims to be independent of the concepts from the various disciplines, with the goal of being able to capture the heterogeneous data structure in one layered model. Our model consists of time-dependent sound sources and geodata that influence the acoustic composition of a soundscape represented by our sensor function. Using a case study, we present the application of our formalization by classifying land use types. For this we analyze soundscapes in the form of recordings from different devices at 23 different locations using three-dimensional convolutional neural networks and frequency correlation matrices. In our results, we present that soundscapes can be grouped into classes, but the given land use categories do not have to correspond to them.