Seminar: "Studying biodiversity change by using soundscape information retrieval techniques" by Dr. Tzu-Hao Lin
“Studying biodiversity change by using soundscape information retrieval techniques” by Dr. Tzu-Hao Lin, Research Center for Information Technology Innovation, Academia Sinica
Biodiversity monitoring is an important task in conservation management. Due to the limitation of sampling intensity, biodiversity change remains difficult to detect. In recent years, soundscape monitoring networks have been deployed in terrestrial and marine environments to facilitate the field assessment of biodiversity. With large amount of acoustic recordings, the remote sensing of geophysical events, biodiversity, and human activities becomes possible; although constructing a sound recognition database remains a challenge and maybe site and time specific. To facilitate the soundscape-based biodiversity monitoring without a recognition database, we have developed three tools for soundscape information retrieval (SIR): (1) visualization of long-duration recordings using long-term spectral analysis, (2) blind source separation using periodicity-coded nonnegative matrix factorization, (3) measurement of bioacoustic diversity based on clustering. In this talk, I will demonstrate applications of SIR techniques in the biodiversity monitoring in forest and coral habitats. I will also discuss the potential opportunity to advance SIR by deep learning. In the future, the integration of SIR and a soundscape monitoring network will help us build an audio library. Besides, it will also be possible to study the interactions between wildlife and habitat through various soundscape components.