Optimal design of reference models using simulated annealingcombined with an improved LVQ3

摘要

For the recognition of large-set handwritten characters,classification methods based on pattern matching have been commonlyused, and good reference models play a very important role in achievinghigh performance in these methods. Learning vector quantization (LVQ)has been studied intensively to generate good reference models in speechrecognition since 1986. However, the design of reference models based onLVQ has several drawbacks for the recognition of large-set handwrittencharacters. To cope with these, the authors propose a method for theoptimal design of reference models using simulated annealing combinedwith an improved LVQ3 for the recognition of large-set handwrittencharacters. Experimental results reveal that the proposed method issuperior to the conventional method based on averaging and otherLVQ-based methods