Method for discriminating traditional Chinese medicine constitution and giving corresponding life suggestions based on neural network model by fusing eye features
By extracting multiple eye characteristics and combining data on the secondary physical constitution of traditional Chinese medicine, iterative training is performed using the multi-classifier neural network model, the problem of inconsistent eye characteristics and the secondary physical constitution judgment standards of traditional Chinese medicine is solved, significantly improving the accuracy of physical constitution identification, and personalized health management suggestions are provided.
Patent Information
- Application Number
- CN202510079482.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-18
- Publication Date
- 2025-05-27
AI Technical Summary
In the prior art, the judging criteria for eye characteristics and the secondary physical fitness of traditional Chinese medicine are not unified, resulting in insufficient accuracy of physical fitness identification and insufficient accuracy, breadth and refinement of data sources.
By extracting multiple eye features, such as eye sight, white eye color, dark circle degree and eyelid morphology, combined with data from the secondary physical constitution of traditional Chinese medicine, iterative training is used to improve the accuracy of physical constitution identification.
It significantly improves the accuracy of physical fitness identification, provides life suggestions that are more in line with the actual needs of users, enhances the personalization of health management, and promotes the integration of traditional Chinese medicine wisdom and modern technology.
Smart Images
Figure FT_1 
Figure FT_2 
Figure SMS_1
Abstract
Description
Technical Field:
[0001] The present invention discloses a method for fusing and analyzing the current eye features of a service object by extracting eye features, predicting traditional Chinese medicine constitutions, and giving corresponding life suggestions. The present invention relates to the field of health service management. Using the data corresponding to the eye expression, sclera color, degree of dark circles under the eyes, eyelid shape and traditional Chinese medicine secondary constitutions integrated according to traditional Chinese medicine constitution theory and relevant textbooks as a training data set, iterative training is carried out on a multi-classifier neural network model (CNN), which further improves the accuracy of the model in constitution fusion, so that the suggestions provided to users are more in line with the actual needs of users, and the reliability of the entire fusion method is improved. Background Art:
[0002] Because there is no unified judgment standard between eye features and traditional Chinese medicine secondary constitutions, the accuracy of constitution identification is insufficient, specifically reflected in: the accuracy, breadth and refinement of data sources are insufficient.
[0003] Related Technologies: Machine learning framework, multi-classifier neural network model (CNN), Panda All are mature technical methods.
[0004] Reference Materials: Classification and Judgment of Traditional Chinese Medicine Constitutions, Huangdi Neijing, Diagnosis of Traditional Chinese Medicine, Traditional Chinese Medicine Constitution Theory, Basic Theory of Traditional Chinese Medicine.
[0005] The key lies in improving the accuracy of the fusion model through training with the training set, so as to improve the accuracy of constitution identification.
[0006] In summary, the existing fusion models have insufficient accuracy in constitution identification due to the insufficient accuracy, breadth and refinement of data sources. Summary of the Invention:
[0007] Currently, using a single eye feature to judge the type of human body constitution has certain limitations and deficiencies. The eye itself contains multiple aspects of information, and it is difficult for a single feature to comprehensively capture these details. The information provided by a single eye feature is relatively less, which may not be sufficient to comprehensively reflect a person's health status and constitution type. Combining multiple eye features can provide more comprehensive information, which helps to more accurately judge the constitution type and physical state.
[0008] This method comprehensively analyzes multiple single eye features instead of relying solely on a single feature. The correspondence between eye features and traditional Chinese medicine constitutions is more refined. Moreover, the trained model can significantly improve the accuracy of constitution identification. Effectively identifying constitutions helps to formulate health preservation and treatment plans that better meet individual needs. This method not only strengthens the personalization of health management but also provides users with a constitution analysis report, health risk assessment, and suggestions for improving clothing, food, housing, and transportation based on the analysis results. Furthermore, it provides more scientific and effective health management support for individuals and promotes the integration of traditional Chinese medicine wisdom and modern technology.
[0009] This system includes:
[0010] Facial video receiving module: This module is designed to receive the facial video transmitted from the front end.
[0011] Eye feature extraction module: This module is designed to extract eye features from the video. The obtained facial video is subjected to feature extraction through an eye feature extraction system. The eye feature extraction method has been submitted for patent application.
[0012] Eye feature calling module: This module is designed to call eye features through a function for use by the next module.
[0013] Eye feature fusion module: This module is designed to fuse the retrieved eye features through a multi-classifier neural network model to obtain the final constitution type. Using the dataset of the correspondence between eye expression, sclera color, dark circle degree, eyelid shape and traditional Chinese medicine secondary constitutions integrated from "Traditional Chinese Medicine Constitution Science", "Huangdi Neijing", and relevant textbooks, input the high-dimensional feature vectors extracted by CNN, and the output is different constitution categories. The cross-entropy loss function and Adam optimizer are used for backpropagation training. After iterative training until the conditions are met, the performance of the model (accuracy, recall rate, F1 score, etc.) is evaluated using the test set. After the model training is completed, the retrieved eye features are input into the model to obtain the probability of each constitution, and then the three most likely constitutions are output according to the principle of the largest probability. (The dataset is shown in Table 2.) Data table of the correspondence between eye expression, sclera color, dark circle degree, eyelid shape and traditional Chinese medicine secondary constitutions
[0014] Module for suggesting food, clothing, housing and transportation corresponding to TCM constitution: This module uses Panda to read TCM secondary constitution and suggestion comparison table 1 according to the TCM constitution with the highest output probability, and provides corresponding suggestions for food, clothing, housing and transportation.
[0015] Result return module: This module aims to transmit the acquired physical type, food, clothing, housing, transportation, and department recommendations to the front end.
[0016] See System Architecture for details. Figure 1 .
[0017] System flow diagram Figure 2 .
[0018] Comparison table 1 of TCM secondary constitution and recommendations. Comparison table of TCM secondary constitution and recommendations Table 1 Table 2 Specific implementation method:
[0019] Step one: Integrate the entire system into the company’s health management system.
[0020] (The website and WeChat applet are online, and the APP has been listed on the Honor, Xiaomi, OPPO, and vivo app stores; the listing of Hongmeng and iOS systems is under review. Website: https: / / yinrk.com / ; APP name: Yinrunkang; WeChat applet name: Yinrunkang Shenzhen Technology Co., Ltd.)
[0021] Step 2: Receive the video transmitted by the front end.
[0022] Step 3: Extract eye features.
[0023] Step 4: Call the extracted eye features.
[0024] Step 5: Input the received eye features into the trained neural network model, and output the three most likely physical types according to the maximum probability principle.
[0025] Step 6: Read the TCM Level 2 Constitution and Recommendation Comparison Table 1 to get recommendations on food, clothing, housing and transportation.
[0026] Step 7: Combine the results of the recommendation data, physical characteristics, and physical performance data and return them to the front end. Description of the drawings:
[0027] System Architecture Figure 1 .
[0028] System flow diagram Figure 2 .
Claims
1. Based on the multi-classifier neural network model (CNN), the eye features are integrated to determine the final TCM constitution and give corresponding life suggestions.
2. Using the data on eye expression, white eye color, dark circles, eyelid morphology and TCM secondary constitutions integrated from "Chinese Constitution", "Huangdi Neijing" and other related textbooks, a data set was constructed for iterative training of the multi-classifier neural network model (CNN). (i.e. Table 2 in the manual)