Achievement

IGERT trainee Forrest Briggs and his advisor Xiaoli Fern (Computer Science) have been working on audio classification of bird species in noisy field-collected recordings with multiple simultaneously vocalizing birds. They have used the recently proposed multi-instance multi-label framework for machine learning to address this little-studied problem. This work has potential applications including automatic acoustic population surveys of birds at a temporal resolution and duration that is infeasible using human observers. Furthermore, this work broadens the scope to include audio applications for multi-instance multi-label learning algorithms, which have previously only been used for text and images, Briggs, Fern, and collaborators are currently preparing an article on this work for the Journal of the Acoustical Society of America, and a condensed conference version which may be submitted to the International Conference on Machine Learning Applications, 2011.