Overview
ISO/IEC 29159-1:2010 defines the Biometric Fusion Information Format (FIF) - a machine‑readable data format for describing the statistical properties of biometric comparison scores used as inputs to score‑level fusion. The standard documents how to represent distributions and summary statistics of genuine and impostor scores so fusion modules can be initialized and operated interoperably across systems. It is explicit that ISO/IEC 29159-1:2010 does not standardize score‑normalization algorithms or the fusion algorithms themselves.
Key technical topics and requirements
- Fusion information format (FIF): a container and record model for exchanging score‑distribution data between comparison subsystems and fusion modules.
- Record types: the standard defines multiple record types (and subtypes) to represent different distributional forms and statistical data useful for fusion.
- Distribution representations: supports empirical representations (e.g., cumulative distribution functions - CDFs) and compact approximations (informative annex material shows spline representations and other approximations).
- Common elements and metadata: standardized fusion header block, parameter kinds/origins, number of comparisons, distribution presence flags, and pre‑normalization indicators to document how input scores were processed.
- Binary and numeric conventions: byte ordering, numeric value encoding and textual representations are specified to ensure machine‑level interoperability.
- Conformance rules: requirements for records to be conformant, enabling consistent parsing and use by fusion components.
- Limitations: FIF describes univariate score distributions per subsystem - it does not provide a format for joint multivariate distributions (i.e., explicit modelling of cross‑score dependencies is outside this part).
Practical applications and who uses it
- Biometric system integrators use FIF to initialize fusion engines when combining multiple modalities (fingerprint, face, etc.), multiple algorithms, sensors or instances.
- Algorithm and sensor vendors package statistical score data as FIF records to enable third‑party fusion modules to interoperate while keeping proprietary matchers as black boxes.
- Evaluation laboratories and researchers exchange empirical score distributions (CDFs, histograms, approximations) to benchmark fusion strategies and perform reproducible experiments.
- Developers of fusion modules read FIF payloads to implement normalization‑aware or classifier‑based fusion without needing access to underlying matcher internals.
Related standards
ISO/IEC 29159-1:2010 is part of the ISO/IEC JTC 1/SC 37 biometric standards ecosystem that includes biometric data interchange formats and interface standards for modality‑specific data (fingerprint, face, iris, etc.). Use FIF alongside modality data formats to achieve open‑systems interoperability and modular biometric system design.
Keywords: ISO/IEC 29159-1:2010, Biometric Fusion Information Format, FIF, score‑level fusion, comparison‑score distributions, CDF, pre‑normalization, interoperability.