Multi-purpose and extensible framework for multimedia content description

(eng) Multimedia content is very rich in terms of meanings and archiving systems have to be improved to consider such richness. This thesis defines a multi-purpose, extensible and scalable framework for multimedia archiving to empower the capacity of content description. It is compatible and embeddable in a pragmatic formal process to integrate different stages of multimedia processing. Its extensibility and multi-purpose design allows the application on conceptually distinct problems. These improvements consider a set of techniques to segment and annotate different kinds of media. To validate this framework, we have applied it in two distinct case studies. The first one presents a technique for segmentation and annotation applied on the description of user interfaces, whose generated meta-data may support several approaches for user assistance, observation and user interface evaluation. The second case study aims to propose a multimodal fusion framework for high-level data fusion between two or more modalities. It takes as input media streams from different system devices, analyzes and identifies intrinsic meanings in these data. The fusion combines modalities to capture a context and enhance the user experience. Both experiments show the robustness of the framework to deal with different and complex problems.