The Application of the Acoustic Complexity Indices (ACI) to Ecoacoustic Event Detection and Identification (EEDI) Modeling

Abstract

In programs of acoustic survey, the amount of data collected and the lack of automatic routines for their classification and interpretation can represent a serious obstacle to achieving quick results. To overcome these obstacles, we are proposing an ecosemiotic model of data mining, ecoacoustic event detection and identification (EEDI), that uses a combination of the acoustic complexity indices (ACItf, ACIft, and ACIfte) and automatically extracts the ecoacoustic events of interest from the sound files. These events may be indicators of environmental functioning at the scale of individual vocal species (e.g., behavior, phenology, and dynamics), the acoustic community (e.g., dawn and dusk chorus), the sound marks (e.g., the flag species of a community), or the soundscape (e.g., sonotope types). The EEDI model is represented by three procedural steps: 1) selecting acoustic data according to environmental variables, 2) detecting the events by creating an ecoacoustic event space (EES) produced by plotting ACIft and its evenness (ACIfte), 3) identifying events according to the level of correlation between the acoustic signature (ACItf) of the detected events and an ad hoc library of previously identified events. The EEDI procedure can be extensively used in basic and applied research. In particular, EEDI may be used in long-term monitoring programs to assess the effect of climate change on individual vocal species behavior (fishes, frogs, birds, mammals, and arthropods), population, and acoustic community dynamics. The EEDI model can be also used to investigate acoustic human intrusion in natural systems and the effect in urban areas.