Somewhere in Puerto Rico, a small yellow frog is chirping into a microphone attached to an iPod. Several kilometers away, a computer is listening. Within a minute, that song will be posted online, and the species of the frog will be identified — all without scientists lifting a finger.
This wildlife recording studio is part of a new project to study biodiversity using automated hardware and software. ARBIMON, which stands for automated remote biodiversity monitoring network, was developed by Mitchell Aide and Carlos Corrada-Bravo from the University of Puerto Rico, who report their new work this week in the journal PeerJ. They teamed up to apply 21st century technology to the problem of species monitoring, combining readily available parts with advanced machine-learning algorithms to analyze thousands of hours of wildlife audio in real time.
The key was to remove a bit of the human element and replace it with computers. “The main contribution has been the software side of it,” Aide explained. “Lots of people are wandering around with external hard drives full of recordings and have no way of analyzing them or managing them.”
The heart of the ARBIMON recording unit is nothing more than an inexpensive microphone attached to an iPod. Wired to an antenna that can transmit the data to a base station as far as 40 kilometers away, the whole setup is powered by a solar panel and a car battery, tucked away from the elements inside a waterproof case. From that base station, the data is sent over the internet to Puerto Rico, where ARBIMON’s servers go to work.
In under a minute, machine-learning algorithms have analyzed the audio files, scanning the frequencies for patterns indicative of a specific species. So far the team has used the technology to single out calls from several frogs, a couple birds, a monkey, and two yet-to-be identified insects.
This wildlife recording studio is part of a new project to study biodiversity using automated hardware and software. ARBIMON, which stands for automated remote biodiversity monitoring network, was developed by Mitchell Aide and Carlos Corrada-Bravo from the University of Puerto Rico, who report their new work this week in the journal PeerJ. They teamed up to apply 21st century technology to the problem of species monitoring, combining readily available parts with advanced machine-learning algorithms to analyze thousands of hours of wildlife audio in real time.
The key was to remove a bit of the human element and replace it with computers. “The main contribution has been the software side of it,” Aide explained. “Lots of people are wandering around with external hard drives full of recordings and have no way of analyzing them or managing them.”
The heart of the ARBIMON recording unit is nothing more than an inexpensive microphone attached to an iPod. Wired to an antenna that can transmit the data to a base station as far as 40 kilometers away, the whole setup is powered by a solar panel and a car battery, tucked away from the elements inside a waterproof case. From that base station, the data is sent over the internet to Puerto Rico, where ARBIMON’s servers go to work.
In under a minute, machine-learning algorithms have analyzed the audio files, scanning the frequencies for patterns indicative of a specific species. So far the team has used the technology to single out calls from several frogs, a couple birds, a monkey, and two yet-to-be identified insects.
- More Here
No comments:
Post a Comment