NIST scientists adapt voice recognition to detect birds at airports

We all know about voice recognition biometrics, but typically it is human voices to which we are referring. Scientists at the National Institute of Standards and Technology (NIST), however, think the technology is for the birds — literally. They have developed an audio telescope that can identify different species of birds by their hoots and squawks, not in order to improve ornithological study but to help prevent unfortunate encounters between planes and birds. “Bird strikes are a significant cause of plane crashes, particularly for small craft,” says Vincent Stanford of NIST. According to the Federal Aviation Administration, bird strikes have caused $2 billion of damage to American aircraft since 1990.

Stanford’s research relies on a modified version of voice recognition software developed by Rockville, Maryland-based Intelligent Automation. Using an array of 192 microphones arranged on the ground in concentric circles nearly two metres across, the audio telescope is trained on a flock of birds after they are spotted using a radar or infrared detector (these latter technologies cannot distinguish between a massive Canada goose and a carrier pigeon). “It was tested with recorded bird calls and also in real-world environments with background noise like trains going past,” Stanford told New Scientist. “It turns out to be very successful at discriminating between different birds.”

So far, the system can only detect sounds from a distance of 100 meters. In order to be useful at auirports, NIST scientists (or the venture capitalists that buy the rights) will have to boost the capability to 2.5 kilometers. This should be possible if more microphones are added to the array, Stanford claims.

-read more in this New Scientist report