Decoding the Deep: Applying artificial intelligence to marine soundscapes

Project Details

Description

This project won internal 2018/19 Themes Research and Innovation Fund (~30K).

Layperson's description

The ocean has always been a noisy place, with naturally-occurring sounds from waves, weather, and animals. These create an ‘underwater soundscape’ representing the unique acoustic conditions of a particular habitat. However, as human activities expand across the ocean, soundscapes are changing to become dominated by anthropogenic noise - for example, noise from vessel traffic, oil and gas exploration, dredging, construction, and military activities. There is growing awareness of man-made noise, and how it may adversely affect marine species. Many species of marine mammals and fish have been shown to respond to human noise, which reduces the ability of animals to communicate, find food, avoid predators, and sense their environment. This can result in animals leaving noisy areas, altering their behaviour or communication calls, or failing to feed and reproduce effectively. In the worst case scenario, some intense forms of anthropogenic noise can even result in physiological damage, affecting the internal organs or hearing structures of animals, which may result in death. Such impacts are of particular concern to acoustically-specialised and endangered species, such as whales and dolphins. Consequently, underwater soundscapes are increasingly being considered in marine spatial planning and habitat-quality assessments.

The University of Portsmouth is collaborating with world-record sailor Alex Alley, who will be collecting underwater acoustic data during his widely-publicised attempt to break the solo, non-stop, around-the-world record. With the assistance of marine instrumentation company RS Aqua, a custom-made foil containing an underwater microphone (‘hydrophone’) has been attached to Alley’s sailing yacht. The hydrophone will collect acoustic recordings throughout his voyage, from the English Channel to the central Atlantic to around the Antarctic. This dataset represents the first global underwater soundscape. Covering some of the noisiest and quietest areas of the ocean, it offers a unique opportunity to identify how human activities and noise
pollution have impacted the marine environment. It is important to assess how noisy
soundscapes are in order to best advise management. The ultimate aim of managing acoustic habitats is to keep quiet areas quiet whilst making noisy areas less noisy. To accomplish this there is a need to understand what sound sources occur in that habitat, classify whether they are naturally-occurring or man-made, and quantify their respective contributions to the overall soundscape.

Having such a large dataset is both advantageous and challenging, in that there is a considerable amount of acoustic data to review and analyse. An additional challenge in assessing soundscapes is identifying sounds in extremely variable data. Therefore, analysing such a large and important dataset requires innovative analytical techniques.
This project aims to develop innovative computational methods for automatic classification of sounds in acoustic data. This will be accomplished by combining the expertise of acousticians, biologists, cosmologists and computational scientists to build detection algorithms. Through integration of artificial intelligence and machine-learning techniques, detection algorithms will be developed that are able to classify sound sources in acoustic data. By identifying the sound sources present in an environment, the resulting ‘soundscape’ can be examined for species monitoring, mitigation of human impacts, and habitat-quality assessments.
Short titleTRIF Soundscape
StatusFinished
Effective start/end date1/04/1931/07/20

Keywords

  • TRIF

Fingerprint

Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.