Baby crying translator, created by a Spanish expert

Fans The Simpsons they will remember episode of the third season —released in the distant year 1992—in which Homer’s brother creates a baby cry translator, an invention with which he manages to recover his fortune. Off the screen and deep into the 21st century, advances in artificial intelligence They allow technological genius of this kind to materialize in the real world.

When babies cry, many parents wonder, “What’s wrong with my baby? Will he be hungry? Could it be a dream? Will your stomach bother you?” Instead of just appealing to maternal instinct, the data scientist Ana Laguna Pradas He set to work and designed a system that provides answers. “The idea came about seven years ago when I was pregnant with my first child. At that moment, the curiosity of a scientist combined with my status as a new mother,” she says in an interview with Hypertext A 36-year-old specialist who lives in Madrid.

His first translator of crying babies appeared long before the current boom generative artificial intelligence, which has ChatGPT as a paradigm, image generators and systems that efficiently convert speech to text. Now, for the first time, their ingenuity is embodied in a commercially available product. It is not surprising that the merger takes place in the video surveillance camera that many parents place in the children’s room.

“There’s a lot of science behind this system,” says the creator of the baby cry translator

baby cry translator
“Since there were no databases, I started recording my son and testing the algorithms. deep learning“says the creator of the baby cry translator. (Photo: Courtesy)

You can hear the faint voice of Laguna Pradas on the other end of the phone. “Sorry for speaking softly. “My little one is sleeping on top of me,” the mother of three, the youngest is only six months old, tells us. As the conversation continues, he comments that when he started his project some time ago, deep learning techniques were just beginning to be applied to unstructured data like audio. With his idea, he was looking for bases with a baby crying. “Since they didn’t exist, I started recording my son and testing the algorithms. deep learning“, Remember.

Progress in this area has been intense since…

Definitely. What we do wouldn’t happen without these backups. In fact, attempts have been made in the past to develop similar cry interpretation applications, albeit with much less advanced technology. These systems did not offer a satisfactory experience. Since then, there have been major changes and improvements. Mainly for processing unstructured data: sound, image and text. Now, quite revolutionary, a whole world of pre-trained generative models has emerged. It is undoubtedly a great revolution.

These types of systems require intensive training. How did they do it in the case of the baby cry translator? How then do they connect the recordings with the emotions of children who do not yet communicate verbally?

“The data is verified by experts, doctors and parents,” explains Laguna Pradas. (Credit: DALL-E via Bing Chat)

The big difference here was big data. A large amount of data is required. These neural networks are still artificial brains that learn from experience. In our case, we collect samples with families all over the world for years and years. There is a very tedious process in providing model data. It is not only verified by experts, but also by doctors and parents. In addition, there are very strict rules to ensure that what the algorithm sees is a cry associated with a specific emotion or need. So it’s quite labor intensive because it requires a lot of data, a lot of pre-processing and big algorithms.

We recently published Scientific articles, which support this. Basically, what we do is not just about acoustics. We have made several clinical findings in which we correlate acoustics (crying) with the baby’s brain activity using electroencephalogram and NIRS (near infrared spectroscopy) technology. In addition, facial expressions and body movements are evaluated. This shows that there is a lot of science behind this system.

How does the monitor that includes the baby cry translator they just launched work?

We entered the market with the baby monitor Maxi Cosi See Pro, it is still a monitoring camera, the typical one that we have in the room when the children are sleeping. Moreover, when they cry, not only do we see them, but they also warn us and help us, help us, support us. The key is in that analyzes the cry and converts it into a reason. It indicates those that are most innate during the first months of life: hunger; dream; flatulence; discomfort; changes in posture; who wants you to pick him up or what we call “irritated”.

baby cry translatorbaby cry translator
Maxi Cosi See Pro, the device in which the baby cry translator developed by Ana Laguna Pradas debuts. (Credit: Zoundream/Dorel Juvenile)

How will this information reach parents?

Through the application, they will receive notifications about what is happening with the child, the reason for crying and how to behave. In addition, there is a small guide that will tell you the most common reasons why he cries. The type of sound is also described so that parents gain strength and increasingly understand what the different cries are. We show that it is normal, that there is also nothing to worry about. In addition, all this information is recorded and statistics are displayed, so parents can monitor their baby very closely and everything is recorded objectively.

Have you used a baby cry translator in your own home?

Yes, it was great because I just got to test first-hand the product that will be launched. So it was very good to see what was right and what was wrong. I think This type of technology is supportive, it empowers you, lets you know that you believe the baby is crying for a specific reason, and you control it in an almost gamification way, i.e. a game. So on the one hand, there’s that moment when he’s crying, you think it’s happening to him and he confirms it to you, it’s really cool. Like I said, it’s a bit of a game.

Right now my little one had an ear infection, I knew something was wrong with him because I looked at the crying stats and saw that he cried a lot more because of pain or irritation. In addition, the amount of crying was generally higher than average. When the pediatrician asked me how she was doing this morning, I told him she was crying a lot more and here it is being objectified. In that sense I find it very useful and I am without a doubt the number one user.

Next steps

Ana Laguna Pradas is a co-founder and data scientist at Zoundream. (Photo: Courtesy)

“We are not discovering America,” admits Laguna Pradas, referring to the translator of the children’s cries. His comment points to the preexistence of ideas about the neurological aspects of sobbing. In this context, she points out that both pediatricians and parents make interpretations that she and her team try to objectify.

“The purpose is to show the potential of crying as a biological marker, which not only supports the baby’s well-being to understand what is happening to him from a communication point of view, his needs and emotions. Also help doctors in the early diagnosis of certain pathologies or neurodevelopmental disorders such as autism. Our goal is to provide additional tools for physicians. And that we can act in time, through therapy and early intervention, to completely change the lives of these children, these families and society in general.

The conversation ends when little Irene, a six-year-old girl, makes some whimpers. Before saying a polite farewell, his mother, Laguna Pradas, tells us about the next steps in the initiative she leads. It hopes to open a new round of funding in 2024 to “continue to scale the product” and conduct further clinical trials. The ambitions are not small: “We want to be a world reference in the analysis of children’s cries,” he concludes.

Receive from us every morning reporter. A guide to understanding what matters in relation to technology, science and digital culture.

Processing…

Ready! You have already subscribed to this subscription

An error occurred, please refresh the page and try again

Also in Hipertextual: