Furnariidae Species Classification Using Extreme Learning Machines and Spectral Information

Automatic bird species classification and identification are issues that have aroused interest in recent years. The main goals involve more exhaustive environmental monitoring and natural resources managing. One of the more relevant characteristics of calling birds is the vocalisation because this allows to recognise species or identify new ones, to know its natural history and macro-systematic relations, among others. In this work, some spectral-based features and extreme learning machines (ELM) are used to perform bird species classification. The experiments were carried on using 25 species of the family Furnariidae that inhabit the Paranaense Littoral region of Argentina (South America) and were validated in a cross-validation scheme. The results show that ELM classifier obtains high classification rates, more than 90% in accuracy, and the proposed features overperform the baseline features.