At first blush, using lip movement to authenticate a device might just seem like another frivolous method to accomplish something we already do on our devices every day easily enough, but there is actually a very strong case to be made for the idea, as the project’s leader Cheung Yiu-ming explained to the South China Morning Post.
“You can use English, Cantonese, or Putonghua,” the computer science engineer explained. “You can even mimic a bird chirping.”
Many forms of authentication rely on language, or numbers, or a complex software interface. What makes lip reading different is that anyone, anywhere can say a word or make a motion with their mouth. What makes it even more ingenious is that no two people will speak or move in exactly the same way.
“An imposter reading the same phrase would still be rejected by the system,” Cheung said, adding that, like other biometrics technologies, the user will have to demonstrate the triggering action multiple times so that the system can build a number of accepted responses on which to base a tolerance.
Right now, Cheung says his team of three has achieved a level of 90 percent accuracy and one of the major roadblocks to wide-scale implementation would be guaranteeing functionality in a variety of different lighting conditions. While the system should be harder to fool with a video recording than the way in which facial recognition prompts can be breached with a photograph, Cheung did not address that potential exploit.
The university has reportedly patented the technology and is exploring options for commercialization. Cheung says he hopes it could be ready for public use with another year of development.
