Multimodal real-time conversation analysis using a novel process engine

This contribution introduces a software framework enabling researchers to develop real-time pattern recognition and sensor fusion applications in an abstraction level above that of common programming languages in order to reduce and minimize programming errors and technical obstacles. Furthermore, a proof of concept using two separate instances of the process engine on different computers with audiovisual data processing is described. The scenario shows the capability of the engine to process data in realtime and synchronously on multiple machines, which are necessary features in large scale projects.