Appearance-based proficiency evaluation of micro-operation skill in removing individual habit

The purpose of this paper is to evaluate proficiency of manual micro-operation skill based on appearance-based information, considering effects of individual habit. First, we extract trajectory features from micro-operation using Dense Trajectories. Second, we calculate histogram from the features using Bag of Features. Then, common elements of histogram corresponding to experts are evaluated using Random Forests to remove individual habit. Finally, we calculate similarity of the histograms as the proficiency. We have experimentally verified that the proposed methodology demonstrated proficiency evaluation in removing individual habit.