In the next Machine Learning Group seminar, we have a talk by Dan Stowell who is an EPSRC Research Fellow at Queen Mary University of London.

Date & Time: Friday 26th September, 13:00-14:00

Location: Room A112 (College Building)

Title: Machine learning for bird sounds: at large scale and fine detail

Abstract: Bird sound recordings embed a wealth of information: species, individual, social grouping, and more. If we could extract this information reliably through machine learning, we could provide evidence at big-data scale to answer questions in conservation monitoring, animal behaviour, and comparative linguistics. I will discuss two areas of my recent work. First, bird species classification, for which I will describe an “unsupervised feature learning” method that led to the strongest audio-only classifier in a recent large-scale 501-species evaluation challenge. And second, methods to infer the structure of the “conversation” in multi-party bird sound recordings.

Bio: Dan Stowell is a researcher applying machine learning to sound. He has worked on voice, music, birdsong and environmental soundscapes. He was recently awarded an EPSRC Early Career research fellowship, based at QMUL, where he is developing new techniques in structured machine listening for soundscapes with multiple birds. Website: http://www.mcld.co.uk/