Douglas McArthur
Director, Biometric Sensors Product Group
Fujitsu Microelectronics America
Sunnyvale, Calif.
www.fujitsumicro.com


Fingerprint sensors are getting smaller and more reliable and provide security for automotive entry, cell phones, and PDAs.

 

Fingerprint sensors from Fujitsu capture a fingerprint image. Software then digitizes the image data, creating a template unique to the fingerprint. The system tests for matches between the sample and a candidate template. False accept and reject rates are small, well under 1%.

 

Capacitive-based sensor ICs require little power and current to deliver a rapid, reliable match. This cross section shows the three layers in the Fujitsu MBF300 sweep sensor and the differing capacitance associated with fingerprint ridges and valleys.

More and more products now incorporate fingerprint sensors. They include computer keyboards and mice, along with the keypads of access-control systems. With approximately 1048 possible fingerprint patterns, duplicate patterns are virtually impossible.

Fingerprint sensor ICs are showing up in PCs, PDAs, and cell phones. They're replacing identification techniques requiring passwords and personal ID numbers (PINs) that are often lost, forgotten, or stolen. They're also generally more efficient and dependable than older methods used by law-enforcement agencies, such as the basic ink-fingerprinting method, because they store only data about specific points of the fingerprint rather than the full image.

Three different types of fingerprint-sensing devices are now in production or planned. These include sweep, single-touch, and in the future, trusted sensors. Sweep sensors are ideal for compact, computer, and communications products that need rapid, easy access by users. Single-touch sensors are finding use in consumer products such as vehicles and home-security systems.

Each sensor type uses a four-stage process that begins with the acquisition stage, in which a PC, phone, or other device captures a fingerprint sample. Data points are extracted from the sample and converted into a template using a preestablished mathematical algorithm.

The verification process extracts at least seven characteristic matching points of the more than three dozen that exist on a person's finger. These include bifurcations and endings that create a minutia of relations, which are defined as the distances between and among key characteristic points. This information is stored as a reference template for future authentication. To access the system, a user sweeps a finger across the sensor area. The live scan template is compared with the reference template, a process taking about 2 sec.

The system then decides whether the live-scan template includes enough biometric data matching the reference template and makes a match or no-match decision. If it fails, the user is cleared for a new sweep. If the wrong user tries to access the device, the equipment will stay locked. For instance, repeated attempts to authenticate a cell phone will cause it to power off.

The sweep sensor detects capacitance when a fingerprint ridge is above the sensing-capacitor pixel electrodes. The pixels in the sensor are 45 µm square and are on a 50-µm grid and have a resolution of about 500 dpi. The sensor's base technology is a standard single-polysilicon, triple-metal, 0.5-µm CMOS process.

A layer of titanium nitride forms the capacitor pixel plates. It is covered with a layer of silicon nitride 7,000 Angstroms thick. The hard metal electrodes and abrasive-resistant coating means the sensor can withstand years of repeated use.An individual fingerprint consists of tightly spaced ridges and valleys. The presence or absence of a ridge affects how fast a standard reference discharge current bleeds off nearby capacitive plates. Each pixel is precharged to a reference current then discharged with the rate of change of the voltage on the capacitor plate proportional to the capacitance. The capacitive cell is a charge-storage-device, or CSD, network. The distance of the skin acts like a shunt resistor across the capacitive sensing plates, which then discharge the voltage on the CSD network.

Capacitance under a ridge is greater, so the discharge time is longer. A pixel underneath a fingerprint valley discharges more quickly. Sample-and-hold circuitry detects the difference in discharge rate and converts it to an 8-bit output. Software captures the raw fingerprint image, digitizes the image data, extracts the resulting minutia template, and then tests to see if the extracted-minutia template matches a reference template.

Capacitive sensors have some advantages over other schemes such as resistive, thermal, and pressure sensors. For starters, they have a high signal-to-noise ratio. Also, fast response times of under 2 sec makes the sensor user-friendly. The standard CMOS process used to produce the hardware requires less than 70 mW of power and negligible standby current.

Fingerprint sensor ICs promise to simplify several typical mobile applications. One example is voice mail. Users can dial the voice-mail number, then sweep a finger across the sensor for authentication. Fingerprint authentication replaces the need to key in mailbox or personal-identification numbers.

A typical single-touch sensor has an array measuring 15 X 15 mm, which rapidly captures the largest fingerprint images. This type of sensor is easy to use and provides rapid transmission of a full fingerprint at 500 dpi. The sensor is a 256 X 300 column-and-row configuration of tiny metal electrodes. Each column is linked to a pair of sample-and-hold circuits. The fingerprint image is recorded in sequence, row by row. Each metal electrode acts as a single capacitor plate, and the contacting finger acts as the second plate. A passivation layer on the surface forms the dielectric between the two plates. Placing a finger on the sensor creates a variety of capacitive values across the array, which are determined by ridges and valleys of the fingerprint, and linked into a full image for authentication.

Single-touch sensors are now being designed and perfected for federal, state, and regional government agencies and police departments for use in identification. They use the same basic self-authentication process as sweep sensors in computers and phones. Single-touch sensors are also being used in automotive applications for one-touch keyless-entry systems as well as homeland security.

In automotive applications, a fingerprint module might mount on a vehicle dashboard or steering column. The driver would touch the sensor then wait a second while the fingerprint template is matched with stored templates. Then all preprogrammed settings for the driver kick in, including mirrors, car seats, radio stations, climate-controlled preferences, and even cruise or speed controls.

As another example, consider a police officer equipped with a PDA having a built-in fingerprint sensor. At a crime stop, a fingerprint sample might take the place of a photo ID or vehicle-registration papers. Beaming fingerprint data back to a central database can quickly verify a suspect's identity and bring up any prior arrests or convictions.

As single-touch and sweep sensors move into computer and consumer applications, industry leaders are preparing for the next move forward: entrusted sensors. Authentication takes place on the sensor itself, so only authorization credentials need to pass to the host CPU. Data exchanged with the sensor is encrypted. Work is being done on the first versions of entrusted sensors with the first commercial versions expected by the end of the year.