Software that lets a camera phone recognise its owner鈥檚 face could provide a handy new security measure, according to a Japanese company.
The digital cameras fitted to many modern cellphones already provide a nifty way to capture and share low resolution images with others. But Omron, based in Kyoto, Japan, has developed software that it hopes could turn these built-in cameras into a useful security tool.
鈥淔unctionality in mobile phones and other mobile devices is upgrading significantly,鈥� says Masato Kawade at Omron鈥檚 Sensing Technology Laboratory. 鈥淢obile devices are carrying increasingly personal information, including address books, schedules and payment information. This technology has been designed to protect this information even when the mobile phone is lost or stolen.鈥�
Advertisement
The OKAO Vision Face Recognition Sensor software was created for existing phones with a digital camera fitted. After taking a picture of their face for reference, a user can configure their handset to lock itself or limit its functionality until another picture is taken.
Identity theft
The software works by measuring key parameters, such as the distance between the eyes, nose and mouth. Omron says the software takes up just 370 Kb of a cellphone鈥檚 memory, and takes about one second to perform the check. In testing, it correctly recognised its owner鈥檚 face in 99 out of 100 attempts, the company claims.
Mark Nixon, a computer vision expert at the University of Southampton, UK, agrees that current smart phones should be powerful enough to perform the task fairly well. He also sees a strong potential appeal for the application. 鈥淕iven current concerns over identity theft, if the phone is for secure transactions such as banking, then I think people would be very interested,鈥� he says.
But Alan Robinson, a research scientist at Sheffield Hallam University, UK, says recognising a face from different positions is likely to pose a significant problem.
Different angles
鈥淚f the face is turned through a few degrees compared with their stored face, then the measurements will change,鈥� he told 快猫短视频. 鈥淎lthough it is possible to deal with this pose problem, the better solution is to use 3D methods, which will record the surface of the face irrespective of pose.鈥�
Robinson adds that without 3D face recognition it might be difficult to prevent someone from fooling the system by holding up a picture of the authorised user. But 3D face recognition requires stereo vision, so two or more cameras would need to be used simultaneously.
The OKAO Vision Face Recognition Sensor program currently works on phones running the Symbian and Linux operating systems. It was demonstrated at the Security Show Japan 2005, which took place in Tokyo between 2 March and 4 March.
