Data science meets the real experience
Meanwhile, on winter 2016, in a town in La Mancha, whose name I do not wish to recall…
Ana was pregnant of Yago. As every mum expecting a baby she was so happy but she also had so many conflicting emotions and feelings. Should I be a good mum? Should I be able to understand my baby needs?
I’m sure many of you reading this article are parents yourselves. So you understand well, being a parent is a wonderful experience but it brings with it some tough implications not really specified in the babies instruction guide 😉
Even if you feel extremely glad, so many doubts emerge, particularly during the late pregnancy and the very first weeks and months. Main reason: the baby is so fragile and you feel so inexperienced. Hopefully, in most of the cases the natural instinct starts doing its job quickly.
However, recent statistic says that about 80% of new parents feel both anxious, and completely unprepared for the raising of their new child. What are we so scared about on this early stage? Well, mainly two reasons:
- Our responsibility as parents taking care of our baby. Are we going to be good parents? Do will be able to understand our baby’s needs? Does he/she need to eat? Or maybe, is he/she having enough milk. Does he/she needs to burp? Or maybe,is he/she suffering from baby colic?
- And that second thing that we worry about our baby’s health. From the first fetal ultrasound we are concerned about his/her health (is his/her heart beating? Has he/she 5 fingers? Is every vital organ working properly?), and once he/she is born (ok, he/she can move, see, ear…). We also worry about something that can happen suddenly, and we didn’t see it coming: sudden death, for example. How many times do we touch our baby’s chest while he/she sleeps, just to be sure if he/she’s still breathing? Who didn’t do that? We pretty much all did that.
Indeed, all this it is absolutely normal, we have also good reasons to be scared. Scientific researches state that 75% of child deaths happen in the first months of the life of the baby. And it is estimated that two thirds of newborn deaths could be prevented by early diagnosis and prevention.
Part of this stress or suffering is also due to the fact that newborns can not talk. Small babies have only ONE way to communicate. They can just cry. However, it is demonstrated that a CRY brings a lot of uncoded information about our baby state.
So, what are the type of information baby cries can tell us? Mainly two types:
- First, we can know what the baby needs. Why he/she is crying? Is he/she hungry, does he/she need to burp, or maybe does he/she have stomach problem? Every baby in the world has some basic needs, and each of these needs has a very specific, distinct cry pattern (audio signal), that can be learnt and understood. Problem is, that particularly at the beginning, it is very difficult for the human ear to understand that language.
- Second, baby cry can be used to identify newborn pathologies. Like asphyxia, or autism, And probably many more as well, such as hyperthyroidism or impaired hearing among others. There are already lots of researches started in this field, that demonstrated that cry analysis is a very powerful and reliable method to have an early diagnosis of certain pathologies. For example, the autism-risk babies’ cries had a rougher, less clear quality, indicating tenseness in the vocal cords. The autism-risk babies’ cries were also higher and varied more wildly in pitch.
One important additional point that needs to be mentioned is that newborns language is universal across all cultures, countries, races and genders at the very beginning. Baby simply cry is pretty much the same way all over the world: in Switzerland, in Spain, in China – wherever. Minor differences in pitch and intonation could vary as the baby grows up without being a problem for our audio signal analysis purpose.
Keeping this in mind, 2 years ago, still pregnant with my first son, I had an interesting and necessary challenge as Data Scientist and future mum: translating my own baby’s cries. I decided to start recording his cries one he would be born in order to develop an algorithm able to recognize my baby’s needs. Just for fun, as a hobby during my maternity leave (as you can imagine I was very bored 😉)
After achieving encouraging results with Yago, the ‘evil’ experiment continues 2 years later with my second newborn Marco, 5 months old now.
After some months of data collection (not only from my son, but also from other babies donations) I was able to “teach” my deep learning algorithm how to learn to recognize patterns on the crying sound signals.
And that’s exactly what we want to do at Zoundream, reason why Ana is part of the team after our meeting in Madrid in July 2019.
The next coming months we will continue working hard in order to use artificial neural networks to translate baby cries in order to make parents and babies life easier.
References: “¿Qué quiere un bebé cuando llora? La inteligencia artificial aprende a traducir los llantos”, el Pais, 2019/01/19 https://elpais.com/tecnologia/2019/01/19/actualidad/1547926456_910406.html