Voice technology “could help detect autism”, BBC News has reported. The BBC website says that a new US study found that the early speech of 86% of infants with autism differed from that of unaffected children. In the study...
Voice technology “could help detect autism”, BBC News has reported. The BBC website said that a new US study found that the early speech of 86% of infants with autism differed from that of unaffected children.
In the study researchers recorded the speech of three groups of children aged 10-48 months: 106 ‘typically-developing’ young children, 49 children with language delay and 77 children diagnosed with autism. Their fully automated recording devices were able to determine differences in speech between the groups and accurately predict which children were from each group. The technique also follows the child in their natural home setting, providing the opportunity for efficient and effective speech assessment in a familiar environment.
This research is still in the early stages, and further study will determine how this system could work alongside other developmental assessment methods. So far, the system has not been investigated as a method for diagnosing new cases of language or developmental delay. Before it is introduced into practice, the uses and feasibility of this novel approach will need to be explored.
The study was carried out by researchers from the Universities of Memphis, Chicago and Kansas and was funded by the Plough Foundation at the University of Memphis. It was published in the peer-reviewed scientific journal Proceedings of the National Academy of Sciences USA.
This was an observational study that attempted to further the techniques used in researching speech and language development. The aim was to investigate an automated method for assessing young children’s speech development on a large scale by carrying out extended recordings in the homes of infants and young children. The main goal of the research was to isolate each child’s vocalisations from other voices and background noise on candid recordings and automatically identify significant features that could be useful predictors of the child’s developmental level.
To gather audio samples, the researchers provided parents with a battery-powered recorder that was then attached to their child’s clothing, recording the child in their natural environment all day. The children recorded were drawn from three different groups: those whose parents self-reported them to be typically-developing, those reported to have language delay and those reported to have autism.
Language delay was confirmed by checking for documentation in medical records or by assessment with a speech and language clinician, and autism was confirmed by checking medical records of the diagnosis. The final sample recorded featured a total of 232 children:
The researchers carried out a total of 1,486 all-day recordings across the groups over the three years of the study, which provided a total of 23,716 hours of audio and captured a total of 3.1 million child utterances.
The recording devices were able reliably to differentiate between the child’s vocalisations and other sounds, allowing the researchers to carry out an in-depth analysis of the 12 parameters of speech known to have a role in speech development. These parameters included how the child was able to articulate each syllable, speech rhythm, pitch, their vocal characteristics and duration of speech.
The researchers looked at the relationship between a child’s overall vocalisations and the number of the 12 parameters that were as expected according to their age.
The researchers found that the automated analysis was able to predict development.
The study also found that in the typically-developing group certain vocal tendencies diminished with age, while this was not seen in the other groups. They also noted that children with autism tended to have quite unpredictable patterns of development, suggesting that they had different vocalisation from both typically-developing children and those with language delay.
Overall, the test correctly identified 90% of children who were in the ‘typically-developing’ group, 80% of those with autism and 62% of those with language delay.
The researchers considered this research to be a ‘proof of concept’, a type of developmental project designed to test how well a conceptual method translates into real-world use. They demonstrated that their method of automated assessment was able to track children’s development on acoustic parameters known to play key roles in speech, and was also able to differentiate the vocalisations of children with autism or language delay from those of typically-developing children.
They conclude that their study of ‘automated analysis’ has the potential to advance research in speech and language development.
This was valuable research that has carried out extensive all-day recordings of children and found that the automated analyses of their vocalisations could distinguish between children with normal development, language delay and autism.
The advantage of this method is that it is completely automated, requiring no human intervention. As it follows the child in their home, it provides the opportunity for efficient and effective speech assessment in a familiar environment.
This research is still in the developmental stages. Further study will be needed to see how this recording system could supplement developmental assessment of children by health professionals and the standard screening and diagnostic procedures used.
So far, the system has only been used to detect previously-diagnosed conditions, and has not yet been tested as a means of identifying undiagnosed linguistic or developmental delay. This means the accuracy of the test needs further testing. Additionally, there are likely to be many other considerations to be addressed before this could be brought into practice, including the costs and feasibility of distributing recorders on a large scale and then having trained personnel available to interpret the data from these in-depth recordings.
As the researchers say, the ability to study linguistic development in natural home environments could provide a completely objective way of detecting speech-related disorders in early childhood. Such an advance would be a highly valuable medical tool for speech and language therapists.