摘要
给出一种嵌入重要性权值的局部线性嵌入(LLE)算法,并将其应用到中医舌象识别中。根据中医医理知识,从舌像中提取出来的特征向量,其各分量对舌象类别具有不同的重要性。基于这一认识,首先根据训练数据的分布计算各舌象类别的重要性权值;然后将该权值嵌入到传统的LLE算法中,计算各舌象类别的低维流形;最后采用最近邻分类器识别舌象类别。实验表明所提算法具有很好的识别性能。
An importance weighted LLE algorithm was proposed, and was applied to the tongue recognition in Traditional Chinese Medicine （TCM）. According to the theory of TCM, the components of a feature vector extracted from a tongue image have different impact on the classification of the tongue. Based on this knowledge, the proposed method firstly computed the importance weights for all of the tongue manifestation categories according to the distribution of the training data. Then, the weights were embedded into the formula of LLE, and to compute the low dimensional manifold for all categories of tongue manifestation. Finally, a nearest-neighbor classifier was employed to judge the category of a queried tongue image. The experiments show that the proposed method has good recognition performance.
基金
国家重点基础研究发展计划(973计划)项目(2006CB504801)
关键词
图像识别
中医舌象识别
局部线性嵌入
重要性权值
image recognition
recognition of TCM tongue manifestation
locally linear embedding
importance weight
