Dr. Dan Stowell, Queen Mary, University of London
D. Stowell, S. Musevic, J. Bonada and M. D. Plumbley, Improved multiple birdsong tracking with distribution derivative method and Markov renewal process clustering . International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2013), May 2013
Automatic analysis of bird sound recordings is important for various reasons, such as ecosystem monitoring and tracking migration patterns. But bird sounds are complex and highly variable, and so fully automatic analysis isn't yet possible. * An important problem is that bird sound recordings usually contain multiple birds singing, so it is often hard to tell "who" is singing when. * We introduced a method which takes the individual "tweets" detected in a sound recording, and joins them together into sequences, helping separate one bird from another. The method finds an exact solution to a particular type of probabilistic model.