Robot Hand Movement Detection based on Top-Down and Bottom-Up Scanpath Prediction
Toyomi Fujita, Claudio Michele Privitera · Procedia Engineering · 2012
This paper presents a method for detection of robot hand movement toward action recognition of partner robot. In a detection process, scanpath acts important role because human attends to regions-of-interests (ROIs) for recognition. We therefore inves–tigated properties of scanpaths when a subject looks at a scene of robot hand movement in a psychophysical experiment. We applied image processing algorithms based on active top-down feature patterns and bottom-up spatialkernels to generate algorith–mic regions-of-interests for scanpath prediction. The predictability is evaluated using a positional similarity index between two sets of regions-of-interests. Experimental results showed that presented method is applicable to the detection of a robot action for hand movement.