This study proposes soundscape recognition models by clustering people based on differences in sound source perceptions. We investigated the effect of sound source identification differences on urban soundscape perception by categorizing people’s environmental sound recognition in outdoor environments. Virtual reality technology employing audio-visual stimuli collected in various urban environments replicated actual environments. Fifty participants’ subjective responses regarding sound source identification, perceived affective quality (8 typical (ISO scale) and 116 extensive attributes (Swedish rating scale)), and overall quality were surveyed. Their categorizations by sound source identification were divided into three clusters: Cluster 1–Attentive to traffic noise and other noises, Cluster 2–Less attentive to the sound environment, and Cluster 3–Attentive to natural and human sounds. Even in identical spaces, participants identified different sound sources, as each cluster focused on different sounds. The soundscape perceptual components were derived differently for each cluster; Cluster 2 extracted additional perception dimensions, i.e., tranquil and relaxed soundscapes. The results showed that each sound source that received an attentive reaction had a positive effect on soundscape perception, showing that appropriate human activities can be encouraged to improve relaxation via soundscape enhancements. The overall quality assessment by cluster revealed similar results, but the resulting indicators’ effects varied. The study’s different soundscape recognition models for each cluster, based on the relationship between soundscape indicators and descriptors, present a new perspective for interpreting urban soundscape perception and can also be used effectively in urban planning design.

本研究通过基于声源感知差异对人进行聚类来提出声景识别模型。我们通过对室外环境中人们的环境声音识别进行分类，研究了声源识别差异对城市声景感知的影响。虚拟现实技术采用在各种城市环境中收集的视听刺激来复制实际环境。调查了 50 名参与者关于声源识别、感知情感质量（8 个典型（ISO 量表）和 116 个广泛属性（瑞典评级量表））和整体质量的主观反应。他们根据声源识别的分类分为三个集群：集群 1——关注交通噪音和其他噪音，集群 2——不太关注声音环境，和集群 3——关注自然和人类的声音。即使在相同的空间中，参与者也识别出不同的声源，因为每个集群专注于不同的声音。每个集群的声景感知成分的推导方式不同；聚类 2 提取了额外的感知维度，即宁静和放松的音景。结果表明，每个收到注意力反应的声源都对声景感知产生积极影响，表明可以鼓励适当的人类活动通过声景增强来改善放松。按集群进行的整体质量评估显示出相似的结果，但由此产生的指标效果各不相同。该研究针对每个聚类的不同音景识别模型，基于音景指标和描述符之间的关系，