%0 Conference Proceedings
%A Bergler, Christian
%A Schmitt, Manuel
%A Maier, Andreas
%A Smeele, S.
%A Barth, Volker
%A Noth, Elmar
%+ Department of Human Behavior Ecology and Culture, Max Planck Institute for Evolutionary Anthropology, Max Planck Society
%T ORCA-CLEAN: A deep denoising toolkit for killer whale communication : 
%G eng
%U https://hdl.handle.net/21.11116/0000-0007-FE01-A
%R 10.21437/Interspeech.2020-1316
%D 2020
%B Interspeech 2020
%Z date of event: 2020-10-25 - 2020-10-28
%C wolrdwide
%X In bioacoustics, passive acoustic monitoring of animals living<br>in the wild, both on land and underwater, leads to large data<br>archives characterized by a strong imbalance between recorded<br>animal sounds and ambient noises. Bioacoustic datasets suffer<br>extremely from such large noise-variety, caused by a multitude<br>of external influences and changing environmental conditions<br>over years. This leads to significant deficiencies/problems concerning the analysis and interpretation of animal vocalizations<br>by biologists and machine-learning algorithms. To counteract<br>such huge noise diversity, it is essential to develop a denoising<br>procedure enabling automated, efficient, and robust data enhancement. However, a fundamental problem is the lack<br>of clean/denoised ground-truth samples. The current work<br>is the first presenting a fully-automated deep denoising approach for bioacoustics, not requiring any clean ground-truth,<br>together with one of the largest data archives recorded on<br>killer whales (Orcinus Orca) – the Orchive. Therefor, an approach, originally developed for image restoration, known as<br>Noise2Noise (N2N), was transferred to the field of bioacoustics, and extended by using automatic machine-generated binary masks as additional network attention mechanism. Besides<br>a significant cross-domain signal enhancement, our previous<br>results regarding supervised orca/noise segmentation and orca<br>call type identification were outperformed by applying ORCACLEAN as additional data preprocessing/enhancement step
%K Killer Whale; Denoising; Call Type; Deep Learning; Orca

%B Interspeech 2020
%P 1136 - 1140
%I International Speech Communication Association
%@ 1990-9772
%U https://github.com/ChristianBergler/ORCA-CLEAN