Musicological indices for soundscape ecological analysis.

Soundscape ecologists have collected sound recordings from large-scale studies that are difficult to analyze with traditional approaches and tools. Natural soundscapes are complex and contain a diverse mixture of biological, geophysical, and anthropogenic sources that span similar frequency bands and often lack a discernible fundamental frequency. Selecting features that are responsive to signals without fundamental frequencies and that are capable of classification for multi-layer signals, or polyphonic textures, is a challenging task in soundscape ecology. Spectral timbral features in various combinations have been shown to discriminate in music classification problems, and lend support to our hypothesis; timbral features in soundscape analysis may detect and identify patterns that are inherently related to order-specific communication in frequency bands shared by biological, geophysical, and anthropogenic sounds. Combined timbral feature extractions provides a new level of information about acoustic activity within a soundscape. Current soundscape metrics assess biodiversity, functional diversity, and acoustic complexity, but may be missing crucial information if we compare musical acoustic analysis techniques used to identify genres and structures in music. This new method provides a relational approach to understanding sound event interactions within soundscapes, refining quantifiable soundscape data, and improving the resolution with which it is analyzed.