The Autism Glass Project uses machine learning and artificial intelligence to recognize other people’s faces and give real-time feedback on their expressions.
This post has been corrected.
“OK, Glass: What are other people feeling?”
This is the thrust behind a new tool that helps kids on the autism spectrum understand other people’s emotions. The Autism Glass Project, as it’s called, uses Google Glass, machine learning, and artificial intelligence to recognize other people’s faces and give real-time feedback on their expressions, a challenge for many people on the autism spectrum.
Behind the project is the Wall Lab at the Stanford School of Medicine. They’ve already run a 40-person pilot in the lab, and earlier this week (Oct. 19) the team launched a 100-person trial version of the modified smart glasses for children and families to test at home.
The glasses use a video camera to compare what the user sees to expressions from large datasets of pictures of faces, then tells the wearer what they’re looking at: happiness, sadness, calmness, fear, anger, disgust, contempt, or surprise. The tool can also recognize new and recurring faces, and adjust to differences in race as well as in lighting.
Dennis Wall, an associate professor in pediatrics at Stanford, who leads the Wall Lab, told Quartz that the goal of the device is “to empower people with autism or families who are struggling with autism with digital devices and solutions they can take home with them and leverage on their own, in a way that’s intuitive and fits within their natural environment.” He hopes that the device will someday be a compliment to, and in some cases a replacement for, traditional behavioral therapies.
After a child uses the glasses, data about his or her interactions with other people can also be collected and sent to an Android app, which parents and caregivers may review.
The engineering behind the glasses is very close to being complete, but how it works with the app still needs some tweaking.
“We’re really optimistic that this phase 2 will give us a sufficient dataset that will be massive and unique in a variety of different ways to prove the concept,” said Wall. “From there we’ll start to move it to a scalable ‘commercial’ phase—what a commercial phase looks like isn’t clear at all; we haven’t discussed it.”
Though Glass was taken off the market in January, Google has donated 35 pairs to the project.
Correction (Oct. 21 9:47 am): A previous version of this post stated that Dennis Wall is an assistant professor in pediatrics at Stanford; in fact, he is an associate professor.