摘要
将SVM应用于手写数字的识别,即将采集、预处理后的样本图像的各个像素点直接作为SVM的输入进行训练,通过交叉验证得到最佳参数,取得了较高的识别率.通过与BP神经网络的实验对比表明了在小样本、高维度的应用环境中,SVM具有训练简单、识别率高的特点.
An application of SVM algorithm in handwritten numeral recognition is proposed in this paper.The pixels of preprocessed images were directly input to SVM,after cross-validation,a set of optimal parameters was obtained.This method gains a high recognition rate.This method is compared with the method based on BP neural-network by experiments.the results show that,in the case of small samples and high dimension,the method of SVM has the characteristics of simple training and high recognition rate.
出处
《泉州师范学院学报》
2010年第4期18-21,共4页
Journal of Quanzhou Normal University
