Revista Romana de Interactiune Om-Calculator

209-224

Authors: Alexandru Bardas, Mihai Dascalu, Stefan Trausan-Matu

The classification of conversations in collaborative environments is needed for a better understanding of discussed subjects. Ontologies represent an efficient and representative method of conceptualising a domain. Starting with a predefined set of ontologies for multiple domains, a precise conversation classification between multiple members is wanted. To facilitate this process, the language in which the messages are written must be detected and there should be as little grammatical errors as possible. In this manner, the keywords used by the conversation’s participants can be found among the concepts from the ontologies, while the dominant and central concepts determine the domains on which the conversations are classified. For doing this analysis, we built a web collaborative platform that enables realtime communication between users, the visualisation of keywords within ontologies using radial graphs and the transcript export in an accessible format for further analysis. The platform is accessible on any type of device, being capable of displaying the minimal needed information for a mobile device and automatically rescaling on a bigger screen. In this paper we provide details regarding the application’s implementation, the advantages of using it and its classification system, as well as the benefits of integrating it in different educational contexts.

239-258

Authors: Bradi Rares, Dorian Gorgan

The task of hand gesture recognition is actively researched and challenging because of complex backgrounds, different lighting environments and unpredictable hand movements. This paper presents an analysis along with an implementation of a system that can recognize the hand center, the fingers and the hand wrist. Hand center recognition is done by finding the minimum inscribed circle in the hand contour. Fingers are detected using the k-curvature algorithm. The hands positions along a trajectory are analyzed and based on those a general direction is established. The features are tracked in time using Kalman filter in order to achieve robustness in use and have stability in the features locations. The hand gestures are made out of a desired direction and a certain positioning of the fingers. An analysis is performed on three levels with the purpose of: quantifying the robustness of the system to lighting and background, determining the error rate and the processing rate.