“It’s a neural network, so you need to show the algorithm examples of what you’re trying to detect,” said Fabrice Matulic, senior researcher at Preferred Networks Inc. and a former postdoctoral researcher at Waterloo. “Some people will make gestures a little bit differently, and hands vary in size, so you have to collect a lot of data from different people with different lighting conditions.”
The team recorded a database of hand gestures with dozens of research volunteers. They also had the volunteers do tests and surveys to help the team understand how to make the program as functional and versatile as possible.
“We’re always setting out to make things people can easily use,” said Daniel Vogel, an associate professor of computer science at Waterloo. “People look at something like Typealike, or other new tech in the field of human-computer interaction, and they say it just makes sense. That’s what we want. We want to make technology that’s intuitive and straightforward, but sometimes to do that takes a lot of complex research and sophisticated software.”
The researchers say there are further applications for the Typealike program in virtual reality where it could eliminate the need for hand-held controllers.
The study, Typealike: Near-keyboard hand postures for expanded laptop interaction, authored by Chhibber, Matulic, Vogel, and team-member Hemant Bhaskar Surale, was recently published in the journal for the proceedings of ACM Human Computer Interaction.
Explore further
Simple, accurate, and efficient: Improving the way computers recognize hand gestures