Researchers Use "Voice Recognition" Program To Count Bats
- Date:
- November 2, 1999
- Source:
- Ohio State University
- Summary:
- Researchers here have shown that computer technology can be used to help estimate how many bats are in an area, simply by analyzing recorded bat calls.
- Share:
COLUMBUS, Ohio -- Researchers here have shown that computer technology can be used to help estimate how many bats are in an area, simply by analyzing recorded bat calls.
Researchers used a computerized neural network that differentiates between the distinct vocal patterns of individual bats. The program helped estimate how many bats produced the calls that were recorded. This neural network is similar to computer voice recognition programs for humans.
The findings are especially promising for habitat managers, according to Stephen Burnett, study co-author and a graduate student in evolution, ecology and organismal biology at Ohio State University.
"Relying on a computer to separate the recorded voices is one way of determining the number of animals in an area without disturbing them in their natural habitat," he said.
Burnett presented the findings November 3 in Columbus at a meeting of the Acoustical Society of America. Burnett co-authored the study with Mitchell Masters, an associate professor of evolution, ecology and organismal biology at Ohio State.
Scientists who study animals typically spend a great deal of time in the field capturing and tagging the animals in order to roughly estimate how many of a given species are in a particular area. But now scientists can simply record vocal sounds and then transfer the recorded calls to the computer. The neural network analyzes and separates the calls, giving an estimate of how many bats made the calls.
Certain species of bats have sonar-like capabilities. That is, they emit sound waves -- or echolocation calls -- that bounce off nearby objects. While flying, echolocation allows these bats to detect obstacles and food.
Burnett and Masters recorded 1,449 echolocation calls from 24 big brown bats (Eptesicus fuscus) to test the neural network. All bats were housed at Ohio State's bat lab.
The researchers put each bat on a platform and recorded their calls. They then input the vocal data into the computer. The neural network evaluated each call based on 10 variables, including length, time and frequency.
The neural network gave the researchers a close estimate of the number of bats that produced a set of calls. The program estimated 29 bats in the group of 24 studied. "That's a reasonable estimate for the number of animals present," he said.
Like other mammals, bats produce calls through their larynx, or voice box. The call is like a fast, high-pitched whistle, inaudible to the human ear. "If we slow the echolocation call down, its sound is sort of like a chirp," Burnett said. "Humans can't hear any more than a click, because our ears can't hear that fast."
A grant from the National Institutes of Health division of deafness and communication disorders supported this research.
Ohio State's bat laboratory has a web site at: http://www.biosci.ohio-state.edu/~eeob/batlab/index.html
Story Source:
Materials provided by Ohio State University. Note: Content may be edited for style and length.
Cite This Page: