2019 Fiscal Year Annual Research Report
Establishing an open science platform of marine ecosystem by soundscape information retrieval
| Project/Area Number |
19K21554
|
|---|
| Allocation Type | Multi-year Fund |
|---|
| Research Institution | Japan Agency for Marine-Earth Science and Technology |
|---|
Principal Investigator |
Lin TzuHao 国立研究開発法人海洋研究開発機構, 地球環境部門(海洋生物環境影響研究センター), ポストドクトラル研究員 (00824377)
|
|---|
| Project Period (FY) |
2019-04-01 – 2020-03-31
|
| Keywords | Soundscape / Information retrieval / Ecoacoustics / Machine learning / Deep sea / Underwater sound |
|---|
| Outline of Annual Research Achievements |
Soundscapes, which are composited by sounds of biological, environmental, and anthropogenic sources, have been considered as a remote sensing platform to monitor marine biodiversity and anthropogenic activities. This study aims to apply marine soundscapes in the evaluation of marine ecosystem health. Our goal is to facilitate the assessment of the spatial-temporal dynamics of marine ecosystems via an international monitoring network of marine soundscapes. There are three primary tasks in this project: (1)Developing open tools of soundscape information retrieval (SIR) to separate biotic and abiotic sounds from marine soundscapes. (2)Evaluating the feasibility of soundscape monitoring in deep-sea environments. (3)Establishing an open science platform of marine soundscapes.
In this project, we developed an open toolbox of soundscape information retrieval, called Soundscape Viewer, for MATLAB and Python. The toolbox was applied in the analysis of long-duration recordings to observe the ecosystem dynamics of algal reefs, continental shelf environments, coral reefs, river estuaries, and deep-sea environments. Our results demonstrate that the machine learning-based source separation can transform single-channel audio into multiple ecological dimensions, from the change of acoustic environment, biodiversity, and anthropogenic interference. Based on international collaboration, this project enables the "Ocean Biodiversity Listening Project" and will generate a large amount of acoustic data for researchers who interest in using soundscape information for ecosystem monitoring.
|
