Research In Domain-oriented Sketch Recognition Method

Human-computer interaction (HCI) is the technology that study the human, the computer and the impact between them.It is one of the important component in the computer system. Pen-based interaction is the major research area in HCI. Result from the the natural character of Pen-based interaction, people could process documents, exchange ideas, record opinions with it in daily life. Because drawing sketches is the most common drawing task, the on-line sketch recognition is important research area of pen-based interaction.In this paper we focus on sketch recognition of two field:the flowchart field and the cognitive assessment field. For flowchart recognition, our approach is based on dynamic programming and can be divided into three stages: candidate symbol generation, symbol classification and candidate symbol combination. We choose WEKA as our classification tool and Random Forest as our classification algorithm.For the cognitive assessment field, we combine the traditional medical congnitive assessment tool-clock drawing test (CDT) with the on-line sketch recognition. The recognition includes two procedures:1) CDT stroke recognition; 2)CDT stroke scoring.At the meantime, we want to explore the potentials of pen’s 3D orientation information in cognitive assessment.At last, we carry out experiments for each field. For flowchart recognition, the experiment is tested on FCinkML database, which includes 177 flowcharts as test samples. In the result, the symbol recognition precision is 74.40% and the stroke recognition precision is 80.25%. For CDT recognition, We test the accuracy of our scoring system using 40 samples. And the accuracy for clock face, digits and hands are respectively 0.8,0.73 and 0.8...