Sketching in air: A single stroke classification framework

We describe a trainable, hand drawn, single stroke 3D Sketch-based classification system, using a motion detecting depth sense camera. Our system captures data from a user, who is free to sketch any desired shape in a 3D environment. The overall system is based on a set of previously defined and well developed classifiers, which are, the Rubine Classifier, $1 recognizer and the Image based classifier. The novelty of this paper comes from 1) the classification of sketches drawn in a 3D environment; 2) extending the pixel based image representation to a voxel-based scheme; and 3) combining the results from individual classifiers using a sensitivity matrix. To evaluate the performance of the system, user studies were performed. To validate the significance of results obtained from the user studies, we performed at-test. Our system outperforms the individual classifiers and is able to achieve an average overall accuracy of 93+%.