What bird is singing?

Abstract: The goal of this work was to create a model for characterizing bird species through recordings of their song. A large set of twenty species with 10+ recordings in each species was considered and the algorithm used time-frequency models for the characterization. Three different approaches were used in an attempt to build the foundations for the final model in a bottom to top manner. The resulting model divided birdsong into syllables that were analyzed and compared in several time-frequency domains in order to characterize the species. The domains include the Spectrogram, the Doppler domain and the Ambiguity domain. The result varied depending on the complexity of the song and the quality of the recordings but for simple songs with recordings of good quality the results were very good.