SIGNATURE WORK
CONFERENCE & EXHIBITION 2022

A close relationship exists between birds and humans regarding their production and life. As a subdiscipline in zoology, ornithology has received long and extensive attention from researchers at home and abroad. Research on birds has not only contributed to the development of ornithology but has also had a meaningful impact on the development of the entire field of biology. Bird species identification from audio is a challenging task in bioacoustics. Efforts have been made to classify species from around the globe. However, a large-scale bird sound classifier for China Region still needs to be in place. In this work, we proposed our pipeline to classify birds within the China region. There have been accumulatively reported over 1000 bird species spotted in China. Based on the China Bird Watching Annual Report-China Bird List Version 8.0 released in 2020, we have preliminarily expanded the scope to around 1480 species. With significant efforts to collect field audio data for these species, we still lack data from a certain amount of species due to their extreme rarity. By employing deep learning methods based on various acoustic features and innovative algorithms, our proposed pipeline is able to identify the majority of the Chinese bird population with an accuracy of over 80%.