|Publication Type:||Journal Article|
|Year of Publication:||2019|
|Authors:||VanSchaik, Zhao, Gasc, Omrani, Pijanowski|
Soundscape ecology evaluates biodiversity and environmental disturbances by investigating the interaction among soundscape components (biological, geophysical, and human-produced sounds) using data collected with autonomous recording units. Current analyses consider the acoustic properties of frequency and amplitude resulting in varied metrics, but rarely focus on the discrimination of soundscape components. Computational musicologists analyze similar data but consider a third acoustic property, timbre.
Here, we investigated the effectiveness of spectral timbral analysis to distinguish among dominant soundscape components. This process included manually labeling and extracting spectral timbral features for each recording. Then, we tested classification accuracy with linear and quadratic discriminant analyses on combinations of spectral timbral features.
Different spectral timbral feature groups distinguished between biological, geophysical, and manmade sounds in a single field recording. Furthermore, as we tested different combinations of spectral timbral features that resulted in both high and very low accuracy results, we found that they could be ordered to “sift” out field recordings by individual dominant soundscape component.
By using timbre as a new acoustic property in soundscape analyses, we could classify dominant soundscape components effectively. We propose further investigation into a sifting scheme that may allow researchers to focus on more specific research questions such as understanding changes in biodiversity, discriminating by taxonomic class, or to inspect weather-related events.
|Short Title:||Ecological Informatics|
Contributions of MIR to soundscape ecology. Part 2: Spectral timbral analysis for discriminating soundscape components