Preventing shipping traffic collisions through identification and classification of the North Atlantic right whale bioacoustic signatures

Erceg, Mirjana

4. term

Medialogy, Master

2022

2022-06-10

79 pages

Whales play an essential role in sustaining the ecological balance of the world. Despite this, the North Atlantic right whale population is declining rapidly due to vessel collisions, commercial whaling, bycatch, entanglement, climate change and debris, with only about 400 left. Researchers have achieved very good results in identifying and classifying sounds using Data augmentation and Deep learning. Research presented here discusses the use of NARW's sounds to avoid vessel collisions by identifying nearby NARWs and classifying their bioacoustic signatures. As part of this project, sounds are converted into spectrograms that are fed into a Convolutional Neural Network. The data was augmented with time warping, frequency masking, and time masking, and Transfer learning was used for training. It has been found that the dataset was effectively trained with an accuracy of 88%. This model may ultimately lead to even better results in the future if improvements are made.

north atlantic right whale ; bioacoustics ; sound recognition ; pretrained model ; neural network ; sound classification ; whale calls ; cetaceans ; transfer learning ; specaugment ; data augmentation ; ship collision ; vessel collision ; conservation ; biodiversity ; ecosystem

Download
View record in AAU Student Projects