Feature Extraction for handwritten Chinese character recognition using X-Y graphs decomposition and Haar wavelet

In this paper, a new approach of feature extraction method for handwritten Chinese character recognition called X-Y graphs decomposition is presented. Central to the proposed method is the idea of capturing the geometrical and topological information from the trajectory of the handwritten character using two unique decomposed graphs: X-graph and Y-graph. For feature size reduction, Haar wavelet is applied on the graphs, in which this is a new attempt of wavelet transform. Features extracted using X-Y graphs decomposition with Haar wavelet not only cover both the global and local features of the characters, but also are invariant of different writing styles. As a result, the discrimination power of the recognition system can be strengthened, especially for recognizing similar characters, deformed characters and characters with connected strokes. Experimental results have proved the efficiency of our proposed method and it is superior to other representative traditional feature extraction schemes with high recognition rate of 95.5%, despite of small dimensionality between 64 (inclusive) and 128 (exclusive) and less processing time.