北京生物医学工程

___________多标记学习在中医舌象分类中的研究_________

Research on multi-label learning in the classification of tongue images in TCM

作者：张静张新峰王亚真蔡轶珩胡广芹

单位：           北京工业大学电子信息与控制工程学院(北京100124)    

关键词：中医；舌象；舌质；舌苔；多标记学习

分类号：           R318.04    

出版年·卷·期（页码）：2016·35·2（111-116）

摘要：

目的中医舌诊中，一幅舌象对应舌色、苔色和苔厚等多个类别，而且舌象的多个类别间存在一定的相关性。传统的数据挖掘技术无法利用这些相关性同时进行建模，本文拟探索用多标记学习方法解决舌象这种多标记数据的分类问题。方法首先对舌象进行苔质分离，分别提取舌质和舌苔的颜色特征，再对舌苔图像分块，提取每一块的纹理特征，随后通过多标记学习算法(multi-label learning by exploiting label dependency, LEAD)进行分类。最后将LEAD的分类结果和ML-kNN的结果进行对比，评价指标为汉明损失（Hamming loss）、平均精度（average precision）和(-评估)（-evaluation）。结果相对于SVM等传统的单标记学习算法，LEAD可以将多个类别同时赋予一幅舌图像，而且在三个指标上的分类效果均优于ML-kNN。结论多标记LEAD算法用于舌象分类能够使得对舌象的描述更全面、准确，可以辅助中医进行舌诊。

Objective In tongue inspection of traditional Chinese medicine (TCM), a tongue image is associated with multiple labels of tongue body color, tongue coat color, the coat thickness and so on, and there are certain correlations between these labels. Modeling can not be carried on with the correlation at the same time by traditional data mining technology. So, we explore with multi-label learning to solve the classification of tongue images with multiple labels. Methods First, color features are extracted after separating tongue coat and tongue body, then blocking is done on tongue coat only and texture features are extracted on each block, and multi-label learning algorithm LEAD is subsequently used for classification. Finally, the classification results of LEAD and ML-kNN are compared, and the evaluation metrics adopted are Hamming loss, average precision and -evaluation. Results A set of proper labels can be assigned to a tongue image simultaneously through this method compared with the traditional single-label learning such as SVM. What’s more, LEAD can achieve better classification results on all the three metrics than ML-kNN. Conclusions LEAD can make the description of the tongue image more comprehensive and more accurate, providing an objective reference for the TCM tongue diagnosis.

参考文献：

［1］刘春雨.基于纹理特征的舌象分类研究及应用［D］.哈尔滨:哈尔滨工业大学,2006.

Liu Chunyu. Texture-based tongue classification and its application［D］.Harbin: Harbin Institute Technology, 2006.

［2］楚宇燕.基于中医望诊的舌象特征提取与健康信息分析技术研究［D］.厦门:厦门大学,2008.

ChuYuyan. The research of extraction of the tongue image's feature & analysis technology of healthy information based on traditional Chinese medicine inspection［D］. Xiamen: Xiamen University, 2008.

［3］李志欣，卓亚琦，张灿龙，等.多标记学习研究综述［J］.计算机应用研究,2014,31(6):1601-1605.

Li Zhixin, Zhuo Yaqi, Zhang Canlong, et al. Survey on multi-label learning［J］.2014,31(6):1601-1605.

［4］Zhang Minling, Zhang Kun. Multi-label learning by exploiting label dependency［C］. Washington D: Proceedings of the 16th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD’10), 2010.

［5］韩锋.舌象颜色空间分析及颜色特征研究［D］.哈尔滨:哈尔滨工业大学,2011.

Han Feng. Tongue color space analysis and color feature study ［D］.Harbin: Harbin Institute of Technology, 2011.

［6］王永刚,杨杰,周越,等.中医舌象颜色识别的研究［J］.生物医学工程学杂志, 2005,22(6):1116-1120.

Wang Yonggang, Yang Jie, Zhou Yue, et al. Tongue image color recognition in traditional Chinese medicine［J］. Journal of Biomedical Engineering, 2005, 22(6):1116-1120.

［7］崔宁海，刘丽萍，李长智.空间主颜色描述符的图像特征提取算法［J］.沈阳理工大学学报，2011,30(4)：19-22.

Cui Ninghai, Liu Liping, Li Changzhi. The spatial dominant color descriptor for feature extraction algorithms of image［J］. Journal of Shenyang University of Technology, 2011, 30(4):19-22.

［8］高程程，惠晓威.基于灰度共生矩阵的纹理特征提取［J］.计算机系统应用，2010,19(6)：195-199.

Gao Chengcheng, Hui Xiaowei. GLCM-based texture feature extraction［J］. Computer Systems & Applications, 2010, 19(6):195-199.

［9］Boutell MR, Luo Jiebo, Shen Xipeng, et al. Learning multi-label scene classification［J］. Pattern Recognition, 2004,37(9):1757-1771.

［10］Daniel Eaton, Kevin Murphy.BDAGL. Bayesian DAG learning［EB/OL］(2013-04-09). http://www.cs.ubc.ca/~murphyk/Software/BDAGL/index.html.

［11］Zhang Minling, Zhou Zhihua. ML-kNN: a lazy learning approach to multi-label learning［J］. Pattern Recognition, 2007,40(7):2038-2048.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录