Software performance for the automated identification of bird vocalisations: the case of two closely related species

Autonomous recording units now facilitate the large collection of audio recordings. However, analysis amounts acoustic data remains a challenge. The time required for manually searching bird vocalisations may be equivalent or greater to duration This major constraint can significantly reduced through use software developed automated identification in We have compared performance four (CallSeeker, Kaleidoscope Pro, Raven and Song Scope) Convolutional Neural Network (CNN) using recordings containing calls Bicknell’s Thrush Gray-Cheeked Thrush, as well other species whose characteristics overlap with those our target species. evaluated all on basis two main criteria, their ability detect classify them correctly by Software ranged from 30 90% terms call detection (recall) 27 99% correct classification (precision). CNNs offer promising solution long-standing problem detecting animal noisy soundscapes, while eliminating tedious manual step configuring algorithms maximise performance.