Project AviPulse: Students develop birdcall identifier to boost conservation

By Sahana Ghosh

Kolkata, Jul 5 (IANS): A group of eco-conscious engineers, entrepreneurs and birders has leveraged technology to come to the rescue of our feathered friends by designing a bird sound identifier -- software which they say aims at conservation with public participation and could boost census studies and numbering of endangered species in India.

The sound processing tool is part of 'AviPulse', a non-profit voluntary project based out of Pune that targets creation of quality datasets on the bird population, migration patterns, anatomy, physiology, conservation status and other parameters.

"We have written an algorithm for 90 species of birds. Bird calls are processed and broken down into single syllables and then fed into the classifier algorithm which then detects the particular species with 98.7 per cent accuracy," Bhavin Chandarana, the project's co-founder and an IIT-Madras alumnus, told IANS.

The multidisciplinary effort is born out of IIT-Madras, Rhode Island School of Design and the Univesity of California-Berkley (UCB).

Its inception was during Chandarana and his batchmates' stint as students at IIT-Madras, around two years ago. Despite their regular jobs now, the bunch has striven on with dedication to devote time to the cause which they profess to be their "passion".

One of the core team members is Pallavi Hujband, a signal processing engineer and birding enthusiast, who contributed to the database of bird calls. She says it is generally "very difficult to identify bird calls as a single bird can have songs or calls."

"Bird calls can be for food, alarm, while songs are for mating. A single bird can have a variety of calls, songs or alerts so not all signal processing algorithms work in most of the cases. Many papers have been published on this but our algorithm is quite better than many others," Hujband told IANS.

Hujband believes the entire project helps in two aspects of conservation: generating awareness and enhancing public participation.

"The key to conservation is accurate identification of species. There are different applications for bird identification but none are specific to India and they do not take bird calls into consideration," said Chandarana elaborating on the significance of the "reliable" sound processing and recognition tool amid a multitude of existing 'wildlife conservation' apps.

Raunak Bhinge, a former IIT-Madras student who is currently a graduate student researcher in machine learning from UCB, is one of the other co-founders. The group collaborated with students from IITs at Powai, Kharagpur and Mandi.

"The sound processing tool could improve census studies and numbering endangered species," Bhinge said.

Besides auditory identification, the project includes an image processing tool for bird identification from images, a bird visualization tool to interactively identify a bird based on observations made in the field and field trips and on-site conservation activities.

Continued ...