Blickgesteuerte Interaktion bei der visuellen Analyse komplexer Szenen zur Entlastung des Arbeitsgedächtnisses

The performance of image-based scene analysis at computer workstations is expected to be improved by usage of eyetracking instead of conventional pointing techniques. Two interaction techniques controlled by eye movements: pure fixation (RFix) and fixation with confirmation (FixMB) and the conventional technique using a computermouse (Maus) were modelled by the Human Processor Modelling Language (HPML). Subsequently, based on these models an indicator of workload was calculated and related to performance. The predicted ranking relating to performance was RFix before FixMB before Maus, which is reverse to the ranking relating to the workload of the interaction techniques. The theoretical prediction was confirmed by a subsequent empirical validation. With the usage of the less straining interaction technique via eye tracking, it is possible to achieve a higher performance in scene analysis than by using the computermouse, although the probands were not accustomed to the eye-tracker and therefore were less well trained in usage of the eye-tracker than in the usage of the mouse. The number of memory errors during a classification task was much higher using the mouse than using interaction techniques controlled by eye movements