基于模糊支持向量机的手写字符串识别

Handwritten Numerical String Recognition Based on Fuzzy Support Vector Machine

由于手写数字串书写的随意和分割技术的局限性,手写数字串的识别成为了手写字符识别领域的一个难点。本文针对手写数字串的分割和识别采用了一种有效的模型,该模型针对手写数字串的分割的不准确设置模糊支持向量机作为检测器进行再次识别确认。实验证明,设置模糊支持向量机作为检测器后,大大地减少了因为分割离群点造成识别错误率增加的问题;其次,模糊隶属度函数的正确选择才能发挥模糊支持向量机的优越性。

Due to the limitation of writing arbitrariness and segmentation technology, the segmentation and recognition of handwritten numerical string becomes a problem of handwritten character recognition. This paper adopts an effective model for handwritten numerical string recognition which could deal with the inaccuracy of segmentation problems by setting verifier based fuzzy support vector machine. The experiment results show that this method with verifier based fuzzy support vector machine can greatly reduce recognition error rate owing to the inaccurate segmentation. In addition, the sound implementation of fuzzy membership function will exploit the advantage of performance of fuzzy support vector machine to the full.