摘要
目的:探讨基于面部特征测量的中医五行人体质类型判别研究,为体质分类研究提供方法学参考。方法:以《灵枢·阴阳二十五人》相关论述为依据,采用图像识别技术对327名在读本科生面部轮廓二维图像等进行聚类分析,计算均值和标准差,并分析五行人之间的差异性,判断五行人体质类别。继而建立五行人体质类型辨识的Bayes判别模型,在此过程中分组变量是根据五行人的体质分类,采用逐步判别分析的方法筛选自变量,进而计算出Bayes判别系数函数结果,利用自身回代性检验与交叉检验来验证模型的预测效果。结果:采用木、火、土、金、水作为判别分析的分组变量,模型拟合良好,自身验证的准确率为93.9%,交叉验证的准确率为91.4%。结论:建立的Bayes判别模型辨识五行人体质准确率很高,具有较高的应用价值;将人脸图像识别技术引入五行人体质辨识系统研究中,可为中医体质分类研究提供新的方法和技术。
Objective: To discuss the research of the identification of five elements of traditional Chinese medicine based on facial feature measurement, and to provide a methodological reference for the research on physique classification. Methods:Based on the relevant expositions of Lingshu·Twenty-five People of Yin and Yang, we use the image recognition technology to collect two-dimensional facial images of 327 undergraduates to calculate the mean and standard deviation and analyze the five elements for cluster analysis. The differences between the five elements were used to determine the constitution of it. Then we established the Bayes discriminant model for identifying the five elements of human body type. In this process, the grouping variables are classified according to the physical fitness of the five elements. The independent variable is selected by the method of stepwise discriminant analysis, and the Bayes discriminant coefficient function result is calculated. Back-testing and crosstesting to verify the predictive effect of the model. Results: Using wood, fire, soil, gold, and water as the grouping variables of the discriminant analysis, the model fits well, the accuracy of self-verification is 93.9%, and the accuracy of cross-validation is 91.4%.Conclusion: The established Bayes discriminant model has a high correct rate of identifying the five-element body constitution and has high application value;the introduction of face image recognition technology to the research of the five-element body constitution identification system can provide a new method and technology.
作者
任慧霞
马晓梅
王锦玉
司帆
王常海
李灿东
REN Hui-xia;MA Xiao-mei;WANG Jin-yu;SI Fan;WANG Chang-hai;LI Can-dong(School of Basic Medicine,Henan University of Traditional Chinese Medicine(Zhongjing College),Zhengzhou 450046,China;The Third Affiliated Hospital of Henan University of Traditional Chinese Medicine,Zhengzhou 450000,China;Research Base of Traditional Chinese Medicine Syndrome,Fujian University of Traditional Chinese Medicine,Fuzhou 350122,China)
出处
《中华中医药杂志》
CAS
CSCD
北大核心
2022年第3期1635-1638,共4页
China Journal of Traditional Chinese Medicine and Pharmacy
基金
2020年河南中医药大学重点学科建设项目(No.15102044-2020-2)
中国中西医结合学会妇产科专委会专项研究基金(No.CR901FC01)
基于人工智能(AI)深度学习技术的“体质-面体型-舌脉象”中医辅助诊断设备研究与应用(No.STG-ZYX09-202181)。
关键词
中医体质
五行人
人脸识别
判别分析
面部特征
TCM physique
Five elements
Face recognition
Discriminant analysis
Facial features
