Audiovisual characteristics, patterns, and facial-expression responses in an under-viaduct green space

Under-viaduct green spaces (UVGSs) combine natural landscape elements with persistent transport-related visual and acoustic cues, yet their audiovisual characteristics and experiential implications remain insufficiently understood.

This exploratory study examined a UVGS in Shanghai using three recorded audiovisual scenarios representing different landscape and traffic conditions. Forty-nine participants viewed the stimuli in a controlled laboratory setting. Visual features were quantified through semantic segmentation, acoustic conditions through field sound-level measurements and 1-s sound-source prominence coding, and facial-expression responses through DeepFace-derived scores sampled at 1-s intervals. Hierarchical clustering was used to identify recurring audiovisual patterns.

The three scenarios differed clearly in visual composition, overall sound level, and sound-source composition, but exploratory component-wise Friedman tests showed no significant scenario-level differences in any of the seven relative DeepFace output components after Holm adjustment. Six audiovisual patterns were identified, with descriptive variation in facial-expression scores across the patterns. Because pattern occurrence was strongly associated with scenario and temporal position, these differences were interpreted descriptively rather than as independent pattern effects.

These scores represent relative facial-expression outputs rather than direct measures of subjective emotional states. Overall, the findings show that audiovisual experience in UVGSs is better understood by considering visual context, sound-source composition, and their co-occurrence rather than isolated environmental indicators.