Author: Ines Hessler - Senior Data Manager
The Fisheries and Oceans Canada Ocean and Freshwater Science Contribution Program (OFSCP) is an initiative that supports scientific research and science-related activities that align with the departmental interest areas such as fisheries, aquatic ecosystems and marine navigation (click here for more information).
In September 2018, our proposed OFSCP project entitled Big data analysis and management of Canadian marine acoustic data sets (DAMMA), has successfully been funded for 3 years and will run until 2021. This funding allows us to progress individual aspects of our national partnerships, establish direct collaborations among underwater acoustics researchers, computer scientists, and data managers, as well as to promote the development of underwater acoustic data management best practice and Machine Learning based data analysis and visualization tools. As part of the project we will be planning and running a number of national workshops addressing topics such as
- Multidimensional visualization of acoustic fields from sources in the 3D basins
- Machine Learning techniques for underwater acoustic data analysis
- Detection and classification of underwater sounds using Deep Neural Networks
- Acoustic Big Data Management
The first workshop run as part of the DAMMA project took place November 12 and 13, 2018 in Montreal, and addressed the topic ‘Multidimensional visualization of acoustic fields from sources in a 3D basin’. During this workshop, we could significantly progress an ongoing MERIDIAN sub-project in which we will develop a web-based and interactive visualization tool for sound fields from different frequencies in the Estuary and Gulf of St. Lawrence. This so-called Ocean Soundscape Atlas is one the main MERIDIAN deliverables and will help to better illustrate and understand the interaction and impact of anthropogenic and natural noise on marine life and the propagation of sound. It will be a fantastic tool to inform researchers and decision-makers, and will likely contribute to an improved environmental management of Canadian waters. The Ocean Soundscape Atlas project is a collaboration between the Université du Québec à Rimouski, Dalhousie University, St. Lawrence Global Observatory, Simon Fraser University and Fisheries and Oceans Canada.
Currently, the MERIDIAN data analytics team is working on the development of Machine Learning based algorithms and software packages that use marine observations to identify and classify patterns of underwater sounds. Many aspects of the computational work necessary to support, enrich, and promote acoustics scientific research is, however, either unknown or poorly understood by the larger acoustics community. Specifically, ocean scientists often do not have access to new technologies and state-of-the-art analytical methods that can be applied to their ongoing underwater acoustics research. So, we are hoping to equip the underwater acoustic community with an extended set of computer science based tools to support them to exploit their complex and multi-faceted datasets in a comprehensive manner while promoting the collaboration between the ocean and computer science community. As part of the DAMMA project we are planning a workshop series on Machine Learning based tools to detect and classify sounds. Our specific interest lies in exploring the possibilities Deep Neural Networks provide for detecting and classifying underwater sounds and to what extend we can humanize such detectors/classifiers. The workshop series will kick off later in the year and we currently busy mapping out the details.