Develop a software that understands the barking of dogs

has the ability to "understand", classify and identify the barking of a dog. According to its creators, this development is unique in that it does its job even in different contexts,such as "play", "fight" or "walk. " The results of this study suggest that dogs emitdifferent acoustic sounds based on their emotional state. This system can alsorecognize the individual barking of a dog in particular, which is unable to humans. Thesoftware opens up new perspectives and paths in the study of animal behavior. By Raúl Morales.

A new computer program developed by Professor Csaba Molnar, from Eotvos Lorand University in Budapest, has become the most accurate tool for studying acousticcommunications amongst animals, reported Reuters.

A new computer program developed by Professor Csaba Molnar, from Eotvos Lorand University in Budapest, has become the most accurate tool for studyingacoustic communications amongst animals, reported Reuters.

One of the most exciting fields of computing and computing is communicationbetween man and machine, explore different possibilities for a computer will understand when we talk. Have already been taken important steps in this research.Surprisingly, however, that computer scientists have set their sights on another type of communication: between dog and machine.

The results of the investigation, which were presented in the journal Animal Cognition,suggests that dogs have different acoustic sounds in terms of their emotional state.

According to Molnár and his colleagues, his research has shown that the piece ofsoftware they have developed is able to classify dog barks according to different situations and even identify the barking of a dog in particular, something that a human being is almost unable to do.

Behind this project is the interest of Molnár experience the ability of a computer algorithm to identify and differentiate the acoustic sounds of dogs barking andclassify them according to different contexts and individual animals.

For this, the software analyzed more than 6,000 barks in six different situations:"stranger", "fight", "walk", "ball", "game" and "alone. " The barks were recordedbefore being transferred to a computer, where they were digitized. The individualbarks were coded, classified and evaluated.
Degree of success

In the first experiment, which sought a classification according to the situations, thesoftware was able to recognize the barks in 43% of cases. The best recognition rateswere in the situation of "struggle " and "bizarre. " By contrast, rates were lowest for the context of "play. "

These data suggest that different motivational states of dogs in contexts submissive, friendly or aggressive would be to acoustically different barks.

In the second experiment was to see to what extent the algorithm was able to know what dog is doing what bark. In this case, correctly classified the barks in 52% of cases.

This is a very high percentage if we think that man is totally unable to discriminatebetween the barking of dogs. Clearly, then, that there are nuances in the bark that ourear can not appreciate.

According to the authors, this work may be helpful for owners and dog trainers. "If we can find the characteristic sound of a dog that represents some emotion, we mayhave valuable information about animal welfare."