By Anastasia Ntracha

04 September 2018 - 12:03

Knitted baby shoes
'We identified a need for communication between parents and babies, for the healthy development of the infant.' Photo ©

Janko Ferlič used under licence and adapted from the original.

 

Anastasia Ntracha is part of the team that developed the iCry2Talk app, which tells parents and carers what babies are saying when they cry. 

Why do babies cry?

Babies cry to tell us if they are hungry, sleepy, in pain, uncomfortable, and if they need a burp, diaper change or hug.

Crying is a baby's first effort to communicate, and an alarm for the parents or carers, to respond to promptly.

When parents can determine the reason for a baby's cry and respond to their needs promptly, they can also build a stronger bond with their infant and lay the foundations for a relationship of security and trust. A quick response also contributes to the infant's healthy physiological and psychosocial development. 

How does your app work?

The parent records the baby's cry for seven seconds, using the iCry2Talk app on a smartphone. The algorithms are trained on a database of cries, and instantly identify the type of cry in the recording. The app displays the result with a text, voice, image and sign language message.

What were the first steps toward developing your app?

The first step is to identify a need. We identified a need for communication between parents and babies, for the healthy development of the infant.

Then we researched ways that scientists have analysed babies' cries. We needed to gather samples of cries, so we researched what information we can legally, physically and technically gather, and how to process cry samples for classification. We also familiarised ourselves with technologies for signal processing, machine learning and deep learning techniques, and Android and iOS app development.

What did you do next? 

We developed a plan of the steps toward the final product.

We tested our algorithms on baby cry data to check our accuracy. For example, how accurately was our app telling the difference between a hungry cry and a cry of pain?

We realised that we didn't have enough cry samples to train our algorithms. Artificial intelligence can offer better results and predictions when it is trained with a greater amount of data. It is similar to humans in that way; when we gather more knowledge on a topic we can become experts in the field.

So, we created a donation campaign to collect our own cry samples.

We developed a simple-to-use cry donation app, and shared it with parents who have babies up to 12 months old and wanted to support our project. We found the parents by contacting neonatologists, pediatricians and gynecologists. We shared our idea with them, and asked them to help us find volunteers.

Doctors and hospitals also verified cry samples (for example, cry samples recorded during vaccinations are pain cries) and gave us their medical expertise on the acoustic analysis of the cry.

How did you gain funding for the app?

We did all the developing on our own and used student credit to access technologies needed for implementation. The Signal Processing and Biomedical Technology Unit (SPBTU) of the faculty of Electrical Engineering and Computer Sciences, at our university, the Aristotle University of Thessaloniki, supported us with marketing costs. That included promotional material and some hardware for experimentation, like the Alexa digital assistant. We expect to raise future funds by participating in innovation and startup competitions, or by pitching our idea to investors.

Anastasia won FameLab Greece in 2015. Join us for Hall of FameLab at the Natural History Museum in London or watch the live-stream on 28 September 2018. 

The iCry2Talk app won second place in Microsoft's Imagine Cup competition. 

If parents or carers with a baby up to 12 months old want to participate in the iCry2Talk cry donation campaign, they can contact the team on Facebook

Hall of FameLab is part of EU Researchers' Night 2018.

You might also be interested in