Analysis Of Facial Features And Its Application In Disease Diagnosis

With the development of computer technology,accelerating the standardization and modernization of TCM diagnosis has become an important task of TCM research.Face-to-face diagnosis is one of the important components of the four diagnosis of TCM.Its objectification and standardization research have important significance for the standardized research of TCM diagnosis.In this paper,we have carried out in-depth research on facial features extraction,facial features fusion,multi-classification of diseases,and design of Chinese medicine facial examination system.First of all,the features of our face are rich.The previous research work on facial features mostly focused on facial color features,extracted and optimized facial color features,and verified the effectiveness of facial color features for disease diagnosis.However,in addition to the color features,the face also has very rich texture features,traditional Chinese medical theory shows that the se facial texture features are also closely related to the health of the human body.Therefore,this paper uses the texture feature extraction algorithm based on Gabor operator to measure the facial texture features by the response of facial images on Gabo r filters in different directions and different scales.Experiments show that facial texture features can effectively achieve disease classification.At the same time,through the classification experiment of different skin blocks,it is proved that some disease information will be projected on the skin in a specific position of the face,which further verifies the correctness and rationality of the theory of TCM.After that,in order to further improve the classification accuracy of the disease,we combine the acquired facial texture features with the facial color features,using the fusion algorithm which called Discriminative Shared Gaussian Latent Variable Model(DS-GPLVM),and the new fusion algorithm we proposed in this paper which called Improved Joint Similarity and Specific Learning(IJSSL)is used to classify the disease.The experimental results show that the above two fusion algorithms can achieve higher classification accuracy.Therefore,we further integrate the fusion features obtained by DS-GPLVM and IJSSL,and use the integrated learning method of Stacking to input the new features of the two models into the target classifier,and the output of the target classifier as the final result.The comparative experimental results of disease classification show that the proposed integrated learning-based approach can achieve better classification results in the two-category experiment.Finally,by dissolving the multi-category into multiple sub-category strategies,we have complete multi-classification learning of the face database,and finally can distinguish five common diseases with a high correct rate.After that,we design and optimize the Chinese medicine facial examination system on this basis,it can analyze the face images collected online or offline in time,and give corresponding analysis reports,which has a good application prospect.