In order to provide improved search and recommendation capabilities for everincreasing amounts of content, cost-effective technologies for automatic extraction of metadata from raw media objects are needed. A typical use case can be searching video segments within a database where a specific speaker talks about a specific topic (keyword).

One of the core challenges in this domain is the need to shift from using individual standalone extractors for annotation to a multimodal and context-aware analysis workflow, which can provide substantially improved search functionalities and results. To realize this potential, Fraunhofer IDMT developed solutions for:

Cost-effective integration of heterogeneous extractors

Flexible extractor workflow orchestration

An extensible, common metadata model

Availability of an expressive, media-aware query language

Recommender systems using both annotations and collaborative filtering 

Alongside with the MICO-project, Fraunhofer IDMT will be presenting its upgraded A/V-Toolbox for the automatic analysis of audio-visual content. With that, processes for broadcasting, distributing, and archiving digital content can be optimized. This toolbox comprises various components for quality control, music and speech detection, video segment matching as well as video annotation tools for actor recognition, temporal video segmentation, and motion analysis. Furthermore, the audio forensic component allows identifying devices and microphones used for recordings, and the quick validation of digital content in terms of whether the content has been edited or encoded at some point in time. 

Visit us at the Fraunhofer booth (80) in hall 8 and get an impression of the MICO-platform and feel free to test the newest A/V-Toolbox!