High-resolution machine listening for wildlife sounds

Supervisor: Dr Dan Stowell

Research group(s): Centre for Digital Music

Acoustic monitoring of birds and animals provides us with valuable new evidence about their populations, their behaviour, and more. Currently, the state of the art is to apply deep learning to spectrogram representations of data. We believe we can do much better than this, using mathematical methods for signal processing - such as high-resolution pitch tracking, chirplet representations, Gaussian processes, dereverberation. In this project you will use datasets both public and from collaborators, and develop new methods that allow for high-resolution analysis of wildlife sounds. This might be to analyse the exact pitch modulations in birdsong, or to infer the number of individuals in a complex recording, for example. This is application-motivated but requires mathematical development of current signal processing and machine learning methods.