A dynamic Time Warping Algorithm for Recognition of Multi-Stroke On-Line Handwriten Characters

Abstract

Dynamic time warping (DTW) is a famous distance to compare two mono-stroke symbols online. It obeys boundary and continuity constraints. The extension of DTW to multi-stroke symbols raises specific problems. A naive solution is to convert the multi-stroke symbol into a single one via a direct concatenation in the handwriting order. However, people may write a symbol with different stroke directions and orders. Applying a brute force method by searching all the possible directions and orders leads to prohibitive calculation times. To reduce the searching complexity, DTW-A* algorithm, which keeps the continuity constraint during each partial matching and reduces the searching complexity by using the A* algorithm, is proposed. Experimental results on a flowchart dataset mainly containing multi-stroke symbols indicate that DTW-A* algorithm helps to achieve the highest recognition rate and stability in cross-validation, as compared with such two other algorithms as DTW and modified Hausdorff distance.