Our voice is perhaps one of the mostuniversal means of communication. And even if we speak different languages, you can always use universal translation software to understand the other person. But what if a person still does not know how to speak, but still tries to tell you something? Yes, we are talking about baby crying. After all, this is not just an incoherent set of sounds. This is an attempt to say something. At least, it was this concept that was followed by researchers from the United States, developing a new translator for children crying.
Immediately it is worth noting that numerousthe questionable quality of the application that flooded Google Play and the Apple Store, we now do not take into account. Behind the creation of a new program is a serious scientific research.
How did the idea of creating a translator of the baby cry?
In fact, everything is quite simple and the idea is not new at all, but one of the developers, a professor from the University of Northern Illinois, Lichuan Liu, in the process of creation, relied on the experience of pediatricians.
"Experienced nurses or pediatricians can determinewhy baby is crying. Experts can do this based on their own experience. We talked with many doctors and they confirmed that theoretically in the children's screaming and crying can be identified some patterns that can be used when creating algorithms for the translator. ”- says Professor Liu.
Therefore, Liu and the team decided to identify featureschildren's cries that signal the presence of pain or discomfort These characteristics include the difference in the intensity and frequency of crying or crying. The team then developed an algorithm based on automatic speech recognition to detect and identify these functions.
This "screaming language recognition algorithm" wastrained on the recordings of children's cries taken from the neonatal intensive care unit. He uses a unique system that restores the signal based on incomplete data (that is, even if the system hears screaming scraps, it can “think it over” how it would look completely and make a decision). This is necessary to identify sounds that sound in a noisy environment. The algorithm can identify children's crying against the background of, say, the speech of an adult or even the loud sounds of a working television. In addition, he is able to work in conditions where there are several kids in the room, singling out from the cacophony of sounds those who can warn of danger.
This is interesting: Fish and bees have learned to communicate with the help of robot translators.
How accurately does the translator work
Then it is time to experiment. A team of scientists gathered experienced health workers who were given to listen to 48 different recordings of baby crying. The same records were analyzed by the car. And if in 70% of cases people “translated” the crying correctly, the algorithm did not lag far behind them, “agreeing” with the doctors almost completely.
Now Professor Liu is hoping to fix it.cooperation with companies for the production of special devices for the translation of baby crying. In addition, scientists continue to improve the technology for greater accuracy and expand the number of recognizable sounds. Experts also plan to add additional features, such as the ability to identify and classify facial expressions of young children. This can help to provide more detailed data on the emotional and physiological state of a small child.
Want more high-tech news? Go to our page in Yandex. Dzen. There you are waiting for unique materials that are not found on the site.