AsianScientist (Jan. 16, 2018) – Researchers may someday identify birds based on their songs and calls with a simple click of the mouse, thanks to new technology described in the International Journal of Computer Applications in Technology.
Birds play an important role in the ecosystem as both predator and prey. They control insect populations, pollinate and disperse seeds for many plants, as well as release nutrients into the land and sea in the form of guano. Bird populations have thus been monitored in many countries as an indicator of environmental change.
While it is possible to manually carry out a spectrographic analysis of bird calls and songs, this technique is laborious and requires experts who have a good ear for the sounds that birds make. Such an approach is also subjective when it comes to distinguishing between birds with very similar calls and songs.
In this study, a team of scientists led by Dr. Arti Bang and Dr. Priti Rege of the College of Engineering in Pune, India, developed an automated system that extracts syllables from 10-millisecond audio frames. Their method builds upon techniques used to analyze music recordings based on features such as tempo, key signature and genre. By minimizing the background noise in the audio recordings, the researchers were able to further increase the accuracy of their method.
Using their algorithm, the researchers were able to successfully classify ten bird species native to India. They observed that bird songs and calls are made up of syllables, and that each call and song is unique to a given species.
The article can be found at: Bang & Rege (2017) Evaluation of Various Feature Sets and Feature Selection Towards Automatic Recognition of Bird Species.
———
Source: Inderscience Publishers; Photo: Pixabay.
Disclaimer: This article does not necessarily reflect the views of AsianScientist or its staff.
