The project aims at developing approaches for building intergrated representations of multimedia data by modelling certain basic properties of the Human Information Processing system (HIP). In MIAMI, we want to study full multimodality aspects in accessing, representing and producing multimedia information by visual, acoustical and tactile/gestural systems. The main aspect studied is how to build intergrated representations of different modalities as they occur at many levels of processing. This integration is useful, both in the disambiguation of the human input to the system and in the disambiguation of the system output.
APPROACH AND METHODS
MIAMI aims to follow the human information processing by providing and analyzing data on selected aspects of human multimodal data processing, and by devising an architecture for revealing and exploiting its integrative capabilities. To avoid bottlenecks, a systematic, bottom-up approach combining accumulated knowledge, psychophysical experiments and deductive thinking is applied to developing new algorithms, their experimental verification and practical applications. The application side is stressed by developing two typical scenarios for multimedia information acquisition, representation, and use. The first scenario is related to symbolic information retrieval and develops an "Information City" metaphor. The second scenario is related to the interaction and manipulation in information space as exemplified by teleoperation. The scenarios cover fundamental aspects of multimedia data handling and are close to practical applications.
Multimedia is the technology of the future and this project is of primary importance for next generation information processing systems. These systems will be evolving from the present "computers" to integrated information assistants, managers and communicators. This project is focused on a central issue of the integrated handling of information acquired by different modalities and presented in a multimedia context. The ideas, models, and algorithms produced in this project are most likely to exert a profound influence on a whole range of future information technology products.