The University of Glasgow has unveiled experimental research into an eye-based phone control system. The results will be presented at the Human Factors in Computing Systems conference in Yokohama at the end of April.
Experiments have shown that eye tracking accuracy can be greatly improved by making on-screen interactive elements taller and narrower, and by using horizontal eye movements to activate specific areas.
Additionally, accuracy was improved by avoiding placing elements at the top and bottom of the screen. The researchers were also able to determine the optimal viewing angle that maximizes the efficiency of eye-to-eye interaction.
Eye tracking is arguably a technology that could be considered futuristic, as Apple recently added eye tracking to iOS as an accessibility option.
"Apple's eye tracking feature helps people with disabilities interact with devices without touching them," explained study supervisor Mohammed Khamis. "However, it currently works in very limited settings. We have been researching for years how to make eye tracking available in any environment, whether it's a train ride or a rainy night walk. This technology could also be useful in medicine, for example, when doctors need to operate devices while wearing sterile gloves."
Laboratory tests included 24 volunteers between the ages of 22 and 44, including those wearing glasses and contact lenses. They were able to control an iPhone X using eye movements in various interactive scenarios.
The following results were achieved:
- Optimal target size: The highest accuracy (up to 70%) is achieved when interactive elements occupy 4 degrees of the user's field of view.
- Accuracy decreases as the device gets closer: The closer the phone is to your face (from 49 cm to 25 cm), the worse the tracking becomes.
- Horizontal movement is more accurate than vertical movement: The camera records left and right eye movements better than up and down movements.
"Eye tracking is a promising technology for hands-free smartphone control, but its accuracy currently suffers when moving," said PhD student Omar Namankani from Glasgow's School of Computer Science. "Our research provides clear recommendations for developers to help make this feature more practical and widespread."