Preliminary Evidence for Global Properties in Human Listeners During Natural Auditory Scene Perception

Area of Science:

• Auditory perception
• Scene analysis
• Acoustic feature processing

Background:

• Auditory and visual scene analysis theories emphasize object identification.
• Global processing in visual scenes is established, but less explored in auditory perception.
• Previous auditory research has lacked ecological validity with naturalistic scenes.

Purpose of the Study:

• Investigate global vs. local processing in auditory scene perception.
• Determine contributions of high-level global and low-level acoustic information.
• Introduce a new, high-quality collection of auditory scenes for ecological validity.

Main Methods:

• Participants rated 8 global properties of novel auditory scenes.
• Acoustic analysis identified low-level features predicting these ratings.
• Exploratory factor analyses (EFAs) and regression models were used.
• Deep neural network models compared human ratings with computational embeddings.

Main Results:

• Acoustic measures formed a seven-factor structure (57% variance).
• Global scene properties formed a two-factor structure (64% variance).
• Each global property was significantly predicted by acoustic variables (R²=0.33-0.87).
• Human ratings aligned more with global scene analysis models.

Conclusions:

• Evidence supports global perception of auditory scenes.
• Acoustic features contribute to global auditory scene perception.
• Findings suggest parallels with ventral stream processing in vision.
• The new scene collection facilitates future research on auditory perception.