Multi-order Standard Deviation Based Distance Metrics and its Application in Handwritten Chinese Character Recognition

Distance metric is the most popular metrics in the area of pattern recognition, and it is always used as a measure of similarity between the test pattern and the reference patterns. In this paper, a new distance metric based on Manhattan distance is proposed. In the distance metric, not only the standard deviation but also the multi-order standard deviation of the reference patterns´ feature vectors is involved. This paper develops this metric and the experiments based on the distance metric are discussed. According to our experiments on HCL2004 handwritten Chinese characters database, the proposed distance metric shows its efficiency by improving the recognition accuracy of the system 4.01% compared with the system performance based on the standard deviation weighted distance metric