Automatic Recognition of Birds through Audio Spectral Analysis

In this paper identification of birds through their sounds is discussed. Bird species identification is gaining importance in field of ecological conservation and ornithology. In India, there are many critically endangered bird species like Forest Owlet, Great Indian bustard, Indian Vulture etc., which are on the verge of extinction. Despite the fact that state bird of Maharashtra is Forest Owlet, evaluation of its population is in its preliminary stage only. Here we present a technique for automatic identification of this owlet and hence provide an aid for population census. From five unknown bird songs we identify a particular bird (Forest Owlet) through frequency domain analysis. Here, four different frequency domain analysis technique, viz., Mean Square Error (MSE) approach, Correlation analysis based on frequency shift and symmetry property, Wiener Filter theory and Mel Frequency Cepstral Coefficients (MFCC) approach are used. This paper present the comparison of these methods when implemented in MATLAB. Recorded bird calls from xento-canto website have been used in the above analysis.