Fujitsu Laboratories Ltd developed a technology to extract a 2,048-bit "feature code" from a palm vein image for biometric authentication.
With the technology, multiple feature codes can be generated. Therefore, if registered biological information is leaked, another feature code can be registered. Also, it is possible to use different feature codes for different services.
Biological information depends on body part and cannot be changed. Therefore, there is a need for a technology that generates different feature codes from one body part in case of information leakage. This time, Fujitsu Laboratories developed a method to (1) extract the features of patterns such as the number and directions of veins from a vein image as vector quantities, (2) select elements useful for authentication and (3) convert them into 2,048-bit feature codes.
The position and shape of a hand near the sensor are corrected based on the outline of the hand so that feature codes can be stably generated. A different feature code can be generated by changing the conditions of the conversion.
For authentication, a newly-generated 2,048-bit feature code is compared with the registered code. When the number of unmatched parts is less than a certain number, the person is judged as identical.
In the past, Fujitsu Laboratories compared vein patterns, taking a long time for authentication. The time it takes to compare feature codes with the new technology is about 1μs, which is about 1/1,000 the time that the old technology requires.
When palm veins are used for authentication using the new technology, the false-positives rate is about 1/100,000, which is less accurate than those of existing authentication technologies using palm veins. Therefore, Fujitsu Laboratories plans to improve the technologies for correcting the position and shape of a hand, aiming to commercialize the new technology within fiscal 2015.