The goal of the AudioTrust+ project is to develop next-generation audio forensics components, which can be used for manipulation detection, quality assessment, content provenance analysis and data integrity assessment. AudioTrust+ algorithms are very heterogeneous in terms of input/output formats, development language and sometimes also regarding hardware requirements, e.g. the need for a dedicated GPU. And a core element of the project is to exploit synergies among the heterogeneous algorithms developed by combining them into multi-modal workflows. Therefore, MICO platform was a natural choice to address the need for AudioTrust+ component integration, by means of:
- The MICO Metadata Model, i.e., an easily-extensible RDF-based ontology for cross-media publishing
- The MICO platform, i.e., a set of services that allows distributed analysis of multimedia content, secure storage, and RDF-based metadata publishing and querying
- The MICO extractor C++ API and Java API, i.e., cross programming language APIs for the integration of customized services for media analysis
For IBC 2016, we integrated 3 AudioTrust+ components for audio tampering detection within the new version of the MICO platform, and we noticed that integration was significantly simpler than with older MICO platform versions (not to talk about integration “from scratch”). This lead to the idea of reporting the integration process step by step, thereby highlighting the differences between integration with this platform versions versus previous versions.