A model for implementing soundscape maps in smart cities

Smart cities are required to engage with local communities by promoting a user-centred approach to deal with urban life issues and ultimately enhance people's quality of life. Soundscape promotes a similar approach, based on individuals' perception of acoustic environments. This paper aims to establish a model to implement soundscape maps for the monitoring and management of the acoustic environment and to demonstrate its feasibility. The final objective of the model is to generate visual maps related to perceptual attributes (e.g. 'calm', 'pleasant'), starting from audio recordings of everyday acoustic environments. The proposed model relies on three main stages: (1) sound sources recognition and profiling, (2) prediction of the soundscape's perceptual attributes and (3) implementation of soundscape maps. This research particularly explores the two latter phases, for which a set of sub-processes and methods is proposed and discussed. An accuracy analysis was performed with satisfactory results: the prediction models of the second stage explained up to the 57.5% of the attributes' variance; the cross-validation errors of the model were close to zero. These findings show that the proposed model is likely to produce representative maps of an individual's sonic perception in a given environment.Smart cities are required to engage with local communities by promoting a user-centred approach to deal with urban life issues and ultimately enhance people's quality of life. Soundscape promotes a similar approach, based on individuals' perception of acoustic environments. This paper aims to establish a model to implement soundscape maps for the monitoring and management of the acoustic environment and to demonstrate its feasibility. The final objective of the model is to generate visual maps related to perceptual attributes (e.g. 'calm', 'pleasant'), starting from audio recordings of everyday acoustic environments. The proposed model relies on three main stages: (1) sound sources recognition and profiling, (2) prediction of the soundscape's perceptual attributes and (3) implementation of soundscape maps. This research particularly explores the two latter phases, for which a set of sub-processes and methods is proposed and discussed. An accuracy analysis was performed with satisfactory results: the prediction models of the second stage explained up to the 57.5% of the attributes' variance; the cross-validation errors of the model were close to zero. These findings show that the proposed model is likely to produce representative maps of an individual's sonic perception in a given environment.A