资源环境科技发展态势分析平台

It is increasingly acknowledged by landscape architects and urban planners that the soundscape contributes significantly to the perception of urban public open spaces. Describing and classifying this impact, however, remains a challenge. This article presents a hierarchical method for classification that distinguishes between backgrounded and foregrounded, disruptive and supportive, and finally calming and stimulating soundscapes. This four-class classification is applied to a growing collection of immersive audio-visual recordings of sound environments from around the world that could be explored using virtual reality playback. To validate the proposed methodology, an experiment involving 40 participants and 50 soundscape stimuli collected in urban public open spaces worldwide was conducted. The experiment showed that (1) the virtual reality headset reproduction based on affordable spatial audio with 360-degree video recordings was perceived as ecologically valid in terms of realism and immersion; (2) the proposed classification method results in well-separated classes; (3) membership to these classes could be explained by physical parameters, both regarding sound and vision. Moreover, models based on a limited number of acoustical indicators were constructed that could correctly classify a soundscape in each of the four proposed categories, with an accuracy exceeding 88% on an independent dataset.