Contribution
Loadings of Acoustical Metrics on Soundscape Items and their relation to Soundscape Dimensions across different datasets
* Presenting author
Abstract:
Studies in soundscape research vary regarding their performances when modeling the two main components Pleasantness and Eventfulness (ISO standard), Comfort and Content (suggested for indoor soundscapes), or Valence and Arousal (used prior to the standard) respectively, with one or the other dimension being easier to capture. Because the standard claims to be applicable all kinds of soundscapes and the underlying eight soundscape items (pleasant, annoying, eventful, uneventful, calm, chaotic, vibrant, and monotonous) are mapped to the dimensions using two static formulas, reasons for this variation must be found elsewhere. Given that the type of sound environment (e.g., outdoor vs. indoor) strongly affects the main soundscape dimensions, we also may assume that it affects the loadings of (objective) acoustic predictors on the single soundscape items. To test this hypothesis, we compared the magnitude of the relationship between selected (psycho-) acoustic variables and the eight soundscape items by estimating the performance of machine learning models for datasets of different soundscape studies using suitable performance metrics. We discuss the results, potential underlying mechanisms, and implications for soundscape research.