Today I presented a summary of our project from Bonn University and the Animal Sound Archive Berlin as a poster at the Listening Into The Wild Workshop at Queen Mary University.
Authors:Daniel Wolff, Rolf Bardeli(2), Martina Koch(3), Klaus-Henry Tauchert(4), Karl-Heinz Frommolt(5) Frank Kurth(6) and Michael Clausen(7)
(1) City UniversityLondon,
(2) Fraunhofer Institute Intelligent Analysis and Information Systems IAIS, German
(3) Humboldt Universität Berlin
(4) Sächsische Vogelschutzwarte Neschwitz
(5) Animal Sound Archive Berlin, Germany
(6) Fraunhofer Institut für Kommunikation,
Informationsverarbeitung und Ergonomie (FKIE)
(7) Friedrich-Wilhelms Universität Bonn, Institut für Informatik
In 2006, a project was set up aiming at a transfer of established techniques in speech recognition and music information retrieval to problems typically met in bioacoustics. Its focus was set on the automated detection of particular animal species in natural environments. The Animal Sound Archive at Humboldt University, Berlin delivered the precisely annotated data necessary for the development of recognition algorithms. Furthermore, the acquisition of a great amount of acoustic monitoring data was performed by experts from this institution. Afterwards, at the University of Bonn, detection algorithms were developed for specific endangered bird species, utilising newly developed methods for denoising in nature recordings and periodicity analysis. We here sketch the algorithms used for birdsong detection as well as a general overview of the project including special requirements for unsupervised monitoring.
Download the poster here.