Biometric Identity Solutions
by Helén Jansson

Commercial acceptance of biometric technology has waited patiently on a number of developments.

Early hopes for very low-cost biometrics revolved around what are essentially software-based biometrics: face or voice recognition. Any biometric process requires a sensor - a piece of hardware to capture the biometric feature - so the inherent advantage of face or voice was expected to be the ubiquity of web cameras or the PC microphone. Hopes for this have stumbled because these sensors, introduced widely from 1999 are still not of good enough quality for acceptable biometric performance.

In this time the longer established family of fingerprint recognition technologies has also made its bid for the 'low cost - high volume' market, most noticeably in the development of dedicated silicon chip sensors. These capture electronically the characteristics of the fingerprint, without the expense of an optical interface and with the efficiency advances of chip-programmable features.

Sensor cost reduction may be of little concern for markets such as physical access control where the acceptable cost of a biometric reader is on average $800, but translated to the desktop, where the target price is one tenth that of a door security system, then even a fistful of dollars packs a heavy punch. With silicon solutions in mind then the simple answer has been to cut down the size of the sensor. This can be done in one of two ways, by reduction along one or along both axes. Reduction on both gives a small, rectilinear sensor and the finger is placed stationery on this to image a far smaller area than with full finger sensors. Reduction of the sensor along just one axis gives a linear sensor array, or swipe sensor, and the fingertip is swept across the surface, imaging the full fingerprint.

The very small size of the rectilinear sensor raises a number of difficult issues for the pairing and performance of a biometric matching algorithm: fundamentally, if algorithms are at full stretch to match a wide population of fingers using full-sized sensors then they are going to struggle with much smaller sensors where the available data is radically reduced. The jury is likely to be out on this issue for some time until adequate data and customer testing has been conducted. On the other hand, the use and characteristics of the swipe sensor are already much better understood.

Swipe technology is already into its second generation of development. Inherent to the swipe sensor is the rapid transmission of the fingerprint pattern data, basically dictated by the user's speed. To resolve this Fingerprint Cards is unique among developers of core biometric technology in having a co-processor ASIC designed to control the sensor and provide the hardware support for the feature extraction algorithms. The result is an easy to use device that maximises the strengths of embedded chip design to take the burden of processing away from the host PC. The same design philosophy applies with the inclusion of a USB interface to this co-processor, pushing down hard the cost of interfacing the twin chip biometric system to a host PC or network. The final piece of the cost puzzle is compliance with the BioAPI standard - already in customer demand - to enable rapid and reliable deployment for applications.

And the final price? With demand set to increase and prices to drop then Fingerprint Cards predicts that it will be delivering a complete, twin-chip fingerprint system for around $10 in the near future.