Scientists create an AI tool to decipher baby’s cries
Any new parent will find that a common occurrence is waking up frustrated and confused to the wailing of their baby. Infants cry. They cry for many reasons, whether the cause is due to hunger, tiredness, illness, pain... And it's definitely bewildering and overwhelming at first. Yet, surprisingly, seasoned parents have shown expertise in accurately identifying a baby's needs through its crying which suggests it is entirely possible to recognize and analyze a bay's well-being solely through their cries.
However, using technology to do so has never been attempted before or even heard of. Nevertheless, researchers based at Northern Illinois University have devised a way to use artificial intelligence to prevent any unnecessary worry for parents and, in some cases, save the lives of infants by detecting illnesses.
In a study published in IEEE/CAA Journal of Automatica Sinica, the group of researchers demonstrated how they designed a new algorithm to distinguish between what they called 'normal' and 'abnormal' cries – the latter being caused by a medical issue. The new deep learning algorithm was designed based on a pre-existing automatic speech recognition system, which they initially used to detect and recognize features of infant cries. To process the large data collected, the team used 'compressed sensing', which reconstructs a signal based on sparse data. It is especially useful for noisy environments – in other words, where babies typically cry.
The researchers then created the algorithm which could identify particular waveforms in cries by looking for variations in loudness, pitch, and timbre which could be compared to a database of recorded baby cries previously identified by experienced neonatal nurses and caregivers. For example, the "neh" sound is correlated to the baby being hungry due to the fact that the sucking reflux pushes their tongue to the roof of their mouth which creates the "neh" sound. Likewise, the "eh" sound generally suggests the baby needs to burp.
Furthermore, although different babies have unique cries, the AI technology is independent of each baby, therefore enabling all parents to potentially use the tool to aid themselves while raising their child. “Like a special language, there are lots of health-related information in various cry sounds. The differences are represented by different features of the cry signals.
To recognize and leverage the information, we have to extract the features and then obtain the information in i,” said Prof Lichuan Liu. Liu and her team are planning on collaborating with more hospitals and medical research centres to collect data. The technology not only promises to be extremely beneficial in medical settings where a doctor can discern cries among sick children but also at home.