基于监督局部线性嵌入的中药材分类鉴别研究*

电子鼻所采集的中药材气味信息往往具有高维性和非线性。针对气味信息的这种特性,提出一种基于监督局部线性嵌入(SLLE)和线性判别分析(LDA)的气味数据分析方法。首先利用SLLE对所采集的高维非线性气味信息进行降维,目的是提取出气味数据内在的低维流行特征,并增大类别间的辨别信息。然后,在低维空间中,利用LDA进行特征分类判别。通过实验,分别将该方法与单独使用SLLE方法及PCA LDA方法进行对比分析,结果表明,该方法可以很好地对五种不同种类的中药材及三种不同产地的何首乌进行分类鉴别,其个体识别率和整体识别率均可达到100%,为使用电子鼻对中药材进行分类鉴别提供了一种行之有效的方法。

Research on discrimination of Chinese herbal medicines based on supervised locally linear embedding

The odor data of Chinese herbal medicines(CHMs) collected by electronic nose is high-dimensional and nonlinear. For these feature data , propose a new method based on SLLE and LDA. First, the SLLE algorithm was employed to reduce the dimension of the feature data. The purpose is to discover the hidden structure from the raw data automatically and increase the information between classes. Then, the LDA algorithm was employed to implement a linear classifier in the reduced space. Through the experiment,compare this method with SLLE and PCA LDA.The results show that, the method can well distinguish five different kinds of CHMs and three different production batches of Radix Polygoni Multiflori with 100% recognition rate of all tested samples, which provides a new way for the classification of CHMs based on E-Nose.