In this paper, we experiment the capabilities of three preprocessing models to eliminate the human-machine conversation signal noise produced by the movement of a pen when drawing a diagram sketch such as an electrical circuit diagram. Aiming at some sketch capturing problems, such as the different resolution of sampling device, different speed of the sketching tool, the different of writing style, the three models found on two sorts of normalization, that is equal distance normalization and unit-based normalization. On a dataset of 300 hand-drawn sketches and two different recognition algorithms, the proposed pre-processing models allow to locate most of the input track, especially, the unit-based models is able to be independent of different environments of sampling devices.
展开▼
