Artificial intelligence has learned to distinguish the "language" of infants

Artificial Intelligence Developers from New Jersey College has introduced a neural network that can distinguish baby cries from each other and classify them.



The study was published in the journal Journal of Automatica Sinica.

Artificial intelligence, built on speech recognition technology, works by the method of compressed sounding - with the help of which the neural network can receive and restore the signal value using data from its previous context.

As part of the experiment, engineers used 26 infants aged from three days to six months in the intensive care unit of newborns. With the help of doctors, the algorithm remembered five main cries, which mean the desire for a baby to eat, sleep, change diapers, the need for attention and discomfort.

According to the leading author of Luchuan Liu's article, in the sounds that are issued by children, as in the case of adults, there is a wealth of health information. For all five categories of sounds, scientists have identified specific frequencies and time intervals.

Within a few days, artificial intelligence recorded 422 cries of hunger, 182 shouts of discomfort, 151 screams of attention, 137 cries of change of diapers, and 79 cries of a dream. In the future, engineers are going to conduct more extensive research.