The Research Of Tongue Feature Extraction Based On Optimization Dictionary Algorithm

With the modern living standards continue to improve,the health care as a basic livelihood growing demand.While Chinese medicine as one of the essence of Chinese traditional culture,Western medicine has irreplaceable advantages in some areas.Therefore,how to make contact contemporary science and technology of modern medicine diagnosis has important significance.Tongue Diagnosis in Medicine by inspection as an important part,tongue objective is not only to become a hot topic of domestic scholars,and attracted the attention of foreign mathematical medical profession.Tongue diagnosis is the most major judgments based tongue color,tongue different colors may exhibit physical differences and the degree of human illness.Some scholars now has to do with the HSI color space components division of tongue color,but very close to the color characteristics of border demarcation is not clear.To address this issue,the article mainly do the work as following:First of all,in order to more accurately quantify the color of the tongue,and thought Based on sparse representation of the particle filter optimization algorithm presents a dictionary.The algorithm sparse linear representation established a theoretical model.When taking pictures with the segmentation method of overlapping blocks to establish a collection of dictionaries,so that the selected block to a greater extent dictionary contains feature information.Secondly,the paper did not like the existing methods as direct color characteristic information processing qualitative analysis,but based on the idea of Monte Carlo particle filter to particles as characteristic information,the establishment of a KD tree as particle collection by similarity to determine particle tongue color classification.Under a program designed to improve the conditions of this study selected three common tongue conducted experiments to verify the feasibility and effectiveness of the algorithm.Finally,due to the special nature of gestational diabetes,the subject to select it as a case analysis.Gestational diabetes is a disease early,it's tongue red and purple tongue tongue-based,clinically is not easy to discern.Process optimization model based on dictionary paper established gestational diabetes tongue image experiment,because it is hard to determine characteristics of tongue color in KD tree appeared to match the workload and time-consuming long time to find the particle problem.Therefore,this article on the solution of the model made improvements,the introduction of accelerated proximal gradient method,the target particle pre-filter,removing particles smaller probability,the probability of larger particles retained.Fast solution reduces the complexity of the model calculations,but also improve the accuracy of feature extraction.The subject of the study of tongue color is only a small part of TCM four diagnostic methods of inspection,with the pattern recognition and computer image processing technology continues to mature,the dictionary algorithm used in this paper also need to continuously optimized,so as to further enhance the accuracy and efficiency of feature extraction.This will also be our future research learning an important goal.