Funding Institutions:
FAPESP / Microsoft Research Virtual Institute
NavScales: Navigating through scales in space, time and knowledge domains (Grant #13/02219-0)
People involved:
Leandro Tacioli
Claudia Bauzer Medeiros
Luís Felipe Toledo
WASIS (Wildlife Animal Sound Identification System) is a public-domain software that recognizes animal species based on their sounds. From a partnership between Laboratory of Information Systems (LIS) and Fonoteca Neotropical Jacques Vielliard (FNJV) of the Institute of Biology of the University of Campinas (UNICAMP), the main goal of this project is to design a tool which supports multiple algorithms to help scientists and general public on the identification of species.
And why is it important? Besides the curiosity itself of knowing which species is calling, we can possibly identify invasive species in a certain area, help on establish migratory patterns from sounds of different locations during a specific period of time, as well as support long duration recording analysis.
The software architecture was designed to support multiple audio feature techniques that extract meaningful information of animal sounds, and classification algorithms that use these extracted data to match against respective audio information stored in the software data repository. The main purpose of these algorithms is to design a classification scheme that can best predict the classes/labels for unseen data (an audio file that we want to identify), process similar to the human brain ability to differentiate among a wide range of sounds and to assign them to previously heard sounds.
Download: WASIS 1.5