Free-hand Sketch Understanding Based On Hidden Markov Model

The purpose of this thesis is to find a solution for sketch understanding, in order to eliminate its ambiguity. We take advantage of simple free-hand sketch which supports normal graphics application. By studying the problem of semantics representation and semantics understanding, we design a sketch understanding frame. And more, we take use of statistical learning method, try to frame a qualitative and quantitative theory for sketch understanding. The main contribution of this thesis includes:Free-hand sketch understanding based on Hidden Markov Model: In the light of fact that sketch understanding is familiar with speech recognition, we look on user habit as sketch information in time sequence .We introduce Hidden Markov Model (HMM), which has been successfully applied in speech recognition, into sketch understanding field.We make experimental comparisons of common descriptive capacity. By analyzing users'inputted strokes and some familiar graphics used in design software, the experimental result shows that adaptive HMM have good performance in free-hand sketch understanding. We also develop a gesture recognition system based on"Quill"system, in which we build HMM algorithm as its classifier.