Open Competition 1 - Information Technology

Robust, Multi-Modal Biometric Algorithm Technology

Sponsor: Quantum Signal, LLC

• Project Performance Period: 11/1/2002 - 10/31/2005
• Total project (est.): $1,930,918.00
• Requested ATP funds: $1,838,970.00

Biometrics--the measurement of physical or behavioral characteristics to verify human identity--is being considered as a security tool in both the public and private sectors. However, many commercial biometric systems cannot accommodate the variations and noise common to real-world situations, and such variation can cause accuracy to drop to 50 percent or less for some systems. Variations that pose difficulties include those inherent in an individual (such as the presence or absence of a beard or eyeglasses) and those due to aspects of data collection (such as microphone quality). Quantum Signal plans a three-year project to develop and test signal processing techniques that will achieve 90 percent accurate biometric authentication through face or voice recognition. Conventional biometric systems use mathematical algorithms (rules) to interpret data by characterizing facial characteristics or other signatures, and extracting and classifying a subset of features that best define differences among individuals. By contrast, the proposed technology will incorporate much of the signal as a unit--relying on unvarying, inherent characteristics of the data--and therefore will be less influenced by unimportant variables. The project will combine and build on two research algorithms invented previously by the principal investigator based on work done at the University of Michigan and Woods Hole Oceanographic Institution, where the algorithms originally were used to identify specific dolphins from their vocalizations in the noisy, difficult environment of underwater sound. This work will be used to characterize transient, multi-component signals such as those in speech, and the other to compare and classify images. In addition, the company will use a "subspace analysis" technique to separate relevant variations from background clutter. The ATP funding is needed because the company is small and the project is too risky to attract private investments. If successfully developed and commercialized, the new technology could be used to improve safety and security in vehicles through occupant sensing, improve airport security through better screening of passengers, and enable faster automated verification in telecommunications applications. The core technology will be useful not only in face and voice recognition but also in other current and future biometric modalities, such as fingerprints, iris recognition, gait analysis, or even complete MRIs of the human body.

For project information:

Dr. Mitchell M. Rohde, (734) 994-0028
rohde@quantumsignal.com

ATP Project Manager

Jack Boudreaux, (301) 975-3560
jack.boudreaux@nist.gov