SBIR-STTR Award

We propose to improve biometric authentication performance by incorporating multimodal data quality analysis into a commercially practical embodiment of an extremely high-performance multimodal biometric verification and/or identification solution. We will utilize our proven multispectral imaging platform to create a whole-hand sensor that incorporates four modalities: five fingerprints, a palmprint, chromatic texture, and handshape. Our proposal includes a unifying method for creating a quality metric independent of modality and a method for combining quality and match information into a single value for each modality. Because the single value has the same meaning across modalities, the values can be combined using any number of methods such as the sum or maximum value. Collecting multiple biometrics with a single insertion reduces the complexity of the hardware and software integration in addition to creating a simpler system for user interaction. A single method for creating quality metrics using data-driven techniques creates a unified method for assessing quality across biometric modalities. Fusion of the multiple biometrics with their respective quality metrics creates a more robust matching metric.

Benefits:
This research will result in a multimodal whole-hand biometric system with exceptional authentication performance. Commercialization will result in enhanced security in numerous access control situations such as borders, ports and airports, bio-labs, nuclear facilities, chemical plants, refineries, and anywhere access control is critical to overall system security.

Keywords:
biometric, multimodal, quality metric, fingerprint, hand geometry, access control, multispectral, chromatic texture

This proposal is for a program that will develop a fully-functional multibiometric sensor based on multispectral imaging of the hand. Four distinct biometric signals will be collected for each hand insertion: fingerprints (4 + partial thumbprint), palmprints, skin textural characteristics, and hand shape. The multispectral aspect of the sensor provides a very robust and reliable means to collect each of the biometric sources of signal. Algorithms for biometric feature extraction and matching will be developed for each of the individual biometric sources. Score-level multibiometric fusion will be used to combine these four different biometric information sources to provide an overall matching score. One method of multibiometric fusion that will be tested is based on the quality of the individual biometric images. Therefore a component of this program will be focused on establishing suitable quality measures for each of the four biometrics contained in the whole-hand data. Based upon the excellent results generated by a proof-of-concept device generated in the first phase of this effort, it is anticipated that the multibiometric whole-hand sensor will achieve extremely high levels of biometric performance. A relatively large multi-person study will be conducted to estimate the performance characteristics of the full system.

Keywords:
Biometric, Multispectral Imaging, Multibiometric, Whole Hand, Skin Texture, Access Control, Forensic, Spoof Detection