Identification of high-frequency periodic acoustic fish tags with deep learning

Innovasea produces fish tags and receivers to track the presence and motion of fish and marine mammals while underwater. Fish tracking (acoustic telemetry) is used by researchers worldwide to determine the abundance and habits of marine life, make decisions about fishing seasons and allowed catches, and help protect marine mammals. Innovasea has developed a novel high-frequency tag technology that is suitable for very small fish and generates more precise trajectories. However, the new smaller fish tags send no explicit identification information so signals from a specific fish tag are isolated from background noise and other fish tags based on the period and/or pattern of the signals. To obtain useful fish tracking trajectories, Innovasea currently applies manual processing which requires expert knowledge.
In this project we will apply advanced deep learning techniques to large manually processed training sets provided by Innovasea to eliminate the manual preprocessing steps. The project is scoped with an initial phase to test feasibility of the concept and subsequent phases for development with an eventual aim of transitioning the best performing prototype system to a fully realized system for filtering Innovasea data.

Faculty Supervisor:

Stan Matwin

Student:

Santosh Kumar Medisetty;Oliver Kirsebom

Partner:

InnovaSea Marine Systems Canada Inc

Discipline:

Computer science

Sector:

Manufacturing

University:

Dalhousie University