The definition of 'soundscape' emphasises the perceptual construct of sounds; thus, the mechanism of sound perceptions becomes vital for soundscape evaluations. Using a qualitative approach, this study explored the aspects and processes of sound perceptions and built a perceptual soundscape structure from the insight of sociology. The interview was conducted between January and March 2018, in four urban public spaces. Data reached saturation after 23 participants were interviewed based on the grounded theory approach. Four perceptual aspects of sounds were identified from the semantic coding analysis: sound classification, sound features, psychological reactions, and soundscape preferences. These aspects form a three-level process of perceiving soundscapes: sound classifications, sound appraisals (including sound features and psychological reactions), and finally, judgment (soundscape preferences). Overall, four aspects categorised into three levels of perception make up the soundscape structure. Soundscape preferences are at the most profound level of perception and are informed by the previous three aspects. Soundscape preferences are expressed through descriptive words and narrative 'image'. The 'image' reflects people participating in different activities according to their social background. Social relationships influence soundscape preferences through people's sound requirements for various activities. The perceptual structure of soundscapes may provide guidance for future soundscape research and soundscape questionnaire design.