Computers may be better at telling the difference between dog barks than humans, according to a new study by Hungarian researchers.
In a series of tests, researcher Csaba Molnár, from Eötvös Loránd University in Hungary, and his team found that a computer was able to differentiate the acoustic features of barks and classify them according to different contexts and individual dogs.
More than 6,000 barks from 14 Hungarian sheep dogs in six different situations were analyzed by the computer. The barks were tape recorded and then digitized on a computer, which used software to study their differences.
The dogs were tested under stranger, fight, walk, alone, ball and play scenarios, which the computer correctly classified 43 per cent of the time. That wasn't the best success ratio, but it was far better than human recognition, the researchers said.
The computer was most accurate in identifying the "fight" and "stranger" contexts, and was least effective at matching the "play" bark.
The results, which appear in this week's Animal Cognition journal, suggest that dogs have acoustically different barks depending on their emotional state, the researchers said.
The researchers also performed a second test, in which the computer identified individual dogs by their bark. The software correctly identified the dogs 52 per cent of the time, again much better than the human result, suggesting there are individual differences in barks even though humans are not able to recognize them.