Basser Seminar Series

Automated Birdsong Recognition

Speaker: Professor Stephen Marsland
Massey University, New Zealand

When: Wednesday 8 July, 2015, 2-3pm. Please note different time to usual.

Where: The University of Sydney, School of IT Building, SIT Boardroom (Room 124), Level 1

Add seminar to my diary

Abstract

Ecologists now routinely record birdsong for long periods of time, producing very large datasets that are virtually impossible to analyse by hand. Clearly, computer methods to recognise and classify these sounds are necessary. However, it is a hard problem, since the data are noisy, show large variety within individual birds, and often overlap. In this talk I will describe some of the motivations for this work, as well as our approach to the problem (which is very much work-in-progress) and includes use of wavelets, machine learning, and geometric shape analysis. Finally, I will describe the work that we are currently undertaking.

Speaker's biography

Stephen Marsland is professor of scientific computing at Massey University in New Zealand. He has a degree in mathematics from Oxford University and a PhD from Manchester University. Following postdocs in the US and Europe he moved to Massey in 2004. His research interests are in shape spaces, particularly Euler equations on diffeomorphism groups, machine learning and image analysis, complex systems, and ecological and health-based applications of these areas.