In this paper a new method for automatic classification and recognition of bird sounds is presented. Our main idea is to study, how inharmonic and transient bird sounds can be recognized efficiently. The data consisted of sounds of eight bird species. Five species, the Mallard (Anas platyrhynchos), the Graylag Goose (Anser anser), the Corncrake (Crex crex), the River Warbler (Locustella fluviatilis), and the Magpie (Pica pica) have inharmonic sounds, whereas the remaining three reference species, the Quail (Coturnix coturnix), the Spotted Crake (Porzana porzana), and the Pygmy Owl (Glaucidium passerinum) have harmonic sounds. The wavelet analysis was selected due to its ability to preserve both frequency and temporal information, and its ability to analyse signals which contain discontinuities and sharp spikes. The feature vectors calculated with the proposed algorithm from the wavelet coefficients were used as the inputs of two neural networks, the self-organizing map (SOM) and the multilayer perceptron (MLP). The results were encouraging, for the unsupervised SOM network recognized 78% and the supervised MLP network 96% of the test sounds correctly.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below