Development of Next Generation Biometric Data Interchange Formats and Related Standards
Greg Cannon September 23, 2008
What are Biometric Data Interchange Format Standards
• • Why would anyone care? Proprietary Systems – Enrolled Data only as good as the support the vendor has the capability to provide – “Interoperable” only if every agent has access to the proprietary biometric technology – Superior performance usually due to IP that could be standardized and licensed Standards – Enables “True” Interoperability, especially for applications that require data sharing across application domains (intra-agency, international, etc) – Significantly mitigates the risk of the loss of enrollment data – For many applications, fundamental capability is much more important than accuracy or speed. – The cost structure of the basic level interoperable performance is driven by support costs – Innovation will improve performance Revisions are required to improve performance and conformance
•
•
Because the choir is a receptive audience
Development - Status
• INCITS M1 – in revision – INCITS 377:200x – close to publication – INCITS 378:200x – INCITS 381:200x – 3D Face Amendment – close to publication – Speaker – INCITS 423:200X – close to publication • JTC1-SC37 – 19794-1 –> 19794-6 in active revision – 19794-11, 13, 14 in early development – 29794-X in late stage development – 29109-X in active development – 29159-1 in early development
Multiple Part Disorder
Fingerprint – Minutia (19794-2)
• Minutia Quality • Zonal Cell Quality • Level 3 Conformance (29109-X) – Minutia Location and Angle Detection – Core and Delta Location and Angle Detection – Minutia Type Detection (Other Types) – Ridge Count Accuracy – Maximum Number of Minutia • False Minutia • Curvature • Ridge Density
But they seem so important at the time
Fingerprint – Pattern 19794-3
• Cell Quality • Level 3 Conformance (29109-X) – Cell Characterization Accuracy – Frequency Component Selection – Core and Delta Location and Angle Detection • Interoperability Between Various Subparts (triplets, fourier, gabor)
I am sensing a trend here
Fingerprint – Image 19794-4
• Swipe Sensor Combination Accuracy • Roll Image Quality • Level 3 Conformance (29109-X) – Real Grayscale Values • Annotation • • • • • • • Fingerprint Type Classification Multispectral Imaging 3D Fingerprint Images Major Case Prints Latent Deformation Estimation Improved Compression Liveness
Can fingerprints taste like gummy bears?
Face – Image 19794-5
•
• • • • • • • •
Difficult to Test Normative Requirements – When is the pose angle out of specification, uncertainty metric – When are shoulders not square – When is lighting not equally distributed – When is exposure not correct – When shall Face be free of shadows or hot spots – How is the accuracy of Feature Points determined – How well is the Frankfurt Horizon determined Auto White Balance Correctness Grayscale Conversion Correctness What is a closed eye? How much hair/glasses can obscure an eye? Non Frontal Matching Multispectral Imaging “Wide Angle” Deformation Estimation Improved Compression Liveness
A faceless man is just one to many
Iris – Image 19794-6
• • • • • • • • • •
“Region of Interest” Compression – How much Sclera Polar Format – How much Pupil, How much Sclera Iris Quality Level 3 Conformance (29109-X) – Real Grayscale Values Uncentered Eyes Eyelid, Eyelash, Reflection Masking Multispectral Imaging, Color Iris Images Iris at a Distance Metadata Improved Compression – Especially that deals with occlusions, specular reflections Liveness
Iris I knew where the sclera was
Signature Sign 19794-7
• Signature/Sign Quality • Real Pressure Force Values • Resolution Sufficiency Profiles • Improved Compression
We are our own signature authority
Fingerprint – Skeletal 19794-8
• Skeleton Quality • Zonal Cell Quality • Level 3 Conformance (29109-X) – Skeletonization Accuracy – Segment Starting Locations – Adjacent Line Link Accuracy – Ridge Count Accuracy – Sweat Pore Detection and Location Accuracy – Trifurcations • Improved Compression
Ridges? We don’t need no stinking ridges
Vein 19794-9
• Level 3 Conformance (29109-X) – Real Grayscale Values – Orthographic Projection Accuracy • Vein Quality • Improved Compression
Sometimes the bloodlines matter
Hand Geometry 19794-10
• Level 3 Conformance (29109-X) – Hand Detection Correctness – Silhouette Accuracy – Distortion Accuracy • Hand Geometry Quality
Do parallel fingers meet?
Signature/Sign Processed 19794-11
• A Work in Progress… • Other Features Besides Turning Points • Interoperability Between Disparate Features
Perhaps this entire presentation has been altered
Speaker (19794-13)
• A Work in Progress… • • • • • • • Input Device Characterization Text Dependence and Independence Environment Conditions Communication Channel Models Signal to Noise Ratio Compression And more…
Are you talking to me?
DNA (19794-14)
• A Work in Progress… • Representation, Characterization and Qualification of – Single Tandem Repeats STRs (including Y-STRs) – Single Nucleotide Polymorphisms SNPs – Mitochondrial DNA – Messenger RNA
That was my evil twin, Gerg
Fusion 29159-1
• A Work in Progress… • Imposter and Genuine Distribution Statistics • Imposter and Distribution Modeling, Sampling • Interoperability between various encodings • Multi-sensor • Multi-algorithm
In search of more genuine imposters
Roadmaps
• • • • • • • • • • •
Common Look and Feel and Harmonization Platform Independence Separate Transfer Syntax from Content Definition Enlist more experts Conformance Testing Avoid Match On Card Conflicts with SC17 XML Encoding Level 3 Conformance Testing Biometric Databases for Conformance, Quality Capture Device Certifications Biometric Quality on Other Modalities
• Reference Implementations?
What if life is not a highway?