Two-stage spectral space and the perceptual properties of sound textures

Textural sounds can be perceived in the natural environment such as wind, waterflows, and footsteps. Recent studies have shown that the perception of auditory textures can be described and synthesized by the multiple classes of time-averaged statistics or the linear spectra and energy spectra of input sounds. The findings lead to a possibility that the explicit perceptual property of a textural sound, such as heaviness and complexity, could be predictable from the two-stage spectra. In the present study, numerous rating data were collected for 17 different perceptual properties with 325 real-world sounds, and the relationship between the rating and the two-stage spectral characteristics was investigated. The analysis showed that the ratings for each property were strongly and systematically correlated with specific frequency bands in the two-stage spectral space. The subsequent experiment demonstrated further that manipulation of power at critical frequency bands significantly alters the perceived property of natural sounds in the predicted direction. The results suggest that the perceptual impression of sound texture is strongly dependent on the power distribution of first- and second-order acoustic filters in the early auditory system.