Non-intrusive biometric recognition
Biometrics is a technology that verifies the user’s identity by measuring individual biological traits, or biological behaviour. It is one of today’s hot topics, as authorities grapple with heightened security risks across the globe.
There are several methods of biometric authentication used worldwide including fingerprint, iris, voice, face, hand, vein and signature.
While an iris biometrics system may be the most accurate, it is the most intrusive. An iris apparatus is generally bulky and is high-priced to implement. A typical voice recognition system is affordable, but not always reliable, since the voice is subject to change with illness, hoarseness, or other common throat problems. The face can be recognised differently depending on expression and age. Signature recognition is easily reproduced and is therefore not reliable.
In contrast, the fingerprint has been used for personal identification for more than 100 years. It has a couple of strong advantages. Fingerprints are unique to each individual and they are unchanging throughout life. Nowadays, fingerprint recognition is considered by many as the best choice for most applications from network security systems to compact devices, based on its accuracy, speed, reliability, non-intrusive interfaces and relative cost-efficiency.
Capacitive fingerprint Sensing
Japanese company Fujitsu (distributed by Crest Technologies )has developed a fingerprint sensor which is capacitive based (the product is distributed by Crest Technologies). The company claims that the unit’s stable operation, low power consumption and compactness make it more attractive than conventional optical sensors.
The device is based on an array of tiny metal electrodes. Each metal electrode acts as one capacitor plate and the contacting finger acts as the second. The value of each capacitor in the array is determined by the contour of the finger ridges and valleys, resulting in an “image” capture of the fingerprint pattern. (See Figure 1.)
In less than a second, the sensor captures several images of the fingerprint and selects the best. Once the fingerprint has been scanned, it is verified against the stored template of the user’s prints.
Fujitsu’s fingerprint verification algorithm is called the “Connected Minutiae Relation Method” (see Figure 2). It compares either the relations of minutiae (tiny details from the fingerprint) orientation, or the position that each ridge ending or each ridge bifurcation takes in a fingerprint pattern.
The company says the method has high performance, with a false rejection ratio of less than one percent and a false acceptance ratio of less than 0.0002 percent. Only fingerprint minutiae data is stored in the fingerprint identification system and no fingerprint images are recorded. Security comes from the fact that it is “impossible to restore the original fingerprint image from minutiae data”. The technique verifies a user’s identity with fingerprint data with an average file size of 300 bytes per print.
The company suggests that the fingerprint identification technology is easy to apply to client/server systems so that administration and verification of fingerprint data can be centred in the server.
Taking the work out of print matching
To complement its range of fingerprint sensors, Fujitsu has developed a software developer’s kit to take the work out of fingerprint matching. The company claims the verification software optimises for the best “false rejection” performance while satisfying market demand for “false acceptance” error rates.
The MBF200 is manufactured in standard CMOS technology. The 256 by 300 sensor array yields a 500-dpi image. An integrated 8-bit analogue-to-digital converter is used to digitise the output
The sensor supports three different industry standard bus interfaces. The USB core circuitry is integrated into the chip and is capable of supporting an image rate of 13 frames per second.
The 8-bit microprocessor bus can support a maximum image rate of 30 frames per second. The MBF200 SPI sustains an image rate of 10 frames per second and requires only 6 wires for connection to a microprocessor that supports the SPI standard.
11-May-2004