Overview
ISO/IEC 15938-4:2002/Amd 2:2006 - Multimedia content description interface, Part 4: Audio - Amendment 2: High-level descriptors extends the MPEG‑7 audio description framework with XML schema definitions and semantics for high‑level audio descriptors. The amendment adds standardized descriptor types for representing scalable series of scalar and vector audio features (SeriesOfScalarType, SeriesOfVectorType) and introduces a rhythmic pattern descriptor (RhythmicBaseType). These descriptors are expressed as XML complexTypes extending MPEG‑7 base types and are intended for interoperable audio metadata, indexing and content‑based retrieval.
Key topics and technical requirements
- XML schema descriptors: Defines complexType syntax for SeriesOfScalarType and SeriesOfVectorType (using mpeg7 types such as floatVector, FloatMatrixType) and for RhythmicBaseType.
- Scalable series representation: Supports multiple resolution representations (Raw, Min, Max, Mean, Random, First, Last, Variance, etc.) to store full‑resolution or downsampled/aggregated feature series. This enables a scalable approach to storing and transmitting audio features.
- Weighting and downsampling rules: Optional Weight vectors control downsampling and aggregation; semantics specify how to ignore zero‑weight samples and how weighted statistics are computed.
- Logarithmic pre‑processing: A LogBase parameter (default = 1.0) indicates if input values must be transformed with a logarithm before series calculation; LogBase must be > 0.
- Vector statistics: SeriesOfVectorType adds multi‑dimensional statistics - Covariance, VarianceSummed, MaxSqDist (Maximum Squared Distance) - and a vectorSize attribute to describe dimensionality.
- Statistical details: Variance is defined using N (not N−1) in the denominator (easy to convert to unbiased form). Rules and formulas are provided for mean, variance, covariance and downsampling behavior.
- Rhythmic pattern model: RhythmicBaseType captures a single rhythmic pattern using ordered prime indices and velocity vectors to enable compact, importance‑sorted rhythmic representations.
Applications and users
Practical uses of this amendment include:
- Audio feature export/import in MPEG‑7 metadata pipelines for content management systems.
- Content‑based audio retrieval, music information retrieval (MIR), and similarity matching where scalable descriptors reduce storage and speed up comparison.
- Metadata generation for broadcasting, streaming platforms, digital libraries and audio archives that need interoperable, high‑level audio descriptors.
- Developers and engineers building tools for audio analysis, indexing, and MPEG‑7 compliant metadata authoring and parsing.
Who will use it:
- Standards implementers, MPEG‑7 tool vendors, audio metadata architects, MIR researchers, broadcasters and archive managers.
Related standards
- ISO/IEC 15938 (MPEG‑7) family - especially Part 4: Audio (base specification) and other Parts that define multimedia content description interface elements.