Automatic Conversion of Communication Media

This thesis answers the question whether it is possible to build a system for personal communication support, which is able to make automatic decisions within limits defined by the user, regarding the selection of communication terminals as well as the conversion of the communication medium. Therefore, the system relieves the user from predefining and foreseeing every possible situation in communication. Instead, the user can focus on his major goals in communication and general preferences in reachability management and let the system do the detailed work. Thereby, such a system can be considered as an 'electronic secretary'. The work is embedded in the concepts of Personal Communication Support. An appropriate prototype system has been built. It is able to evaluate the assumed outcome of such media conversions and to automatically propose the most suitable way of conversion and delivery, based on the reachability of the user as well as the technical quality and the semantics of the communication content. To reach this goal, a new model of media handling and conversion is proposed, and the impacts of media conversion on the semantic of information are discussed from the theoretical point of view. A new generic model of telecommunication services is proposed, a new concept of evaluating the quality of complete chains of concatenated conversion modules is applied. Methods from the area of automatic decision making are adapted to the specific problem, and algorithms for selecting and configuring chains of concatenated software and hardware modules and techniques for controlling and stream binding such modules are developed. Supporting complex kinds of media conversion, the system provides a generic interface for integrating third party products such as Text-to-Speech, Speech recognition, Optical Character Recognition. This concept of easy integration of conversion and filtering tools allows further enhancements. Beyond considering the technical quality of the message, content semantics (intelligent content handling), ranging from keyword recognition to multimedia content screening, can be used for evaluation. Closely following design principles of the Reference Model of Open Distributed Processing, and based on a vendor-open middleware using CORBA, the system proves interoperability between heterogeneous components, and is widely scalable, allowing tailoring for various purposes and environments.