Method for discriminating traditional Chinese medicine constitution and giving corresponding life suggestions based on neural network model by fusing face image features

Abstract

The invention discloses a method for performing fusion analysis on current face image features of a service object through extracted face image features, predicting traditional Chinese medicine physique and giving corresponding life suggestions. The invention relates to the field of health service management, and relates to the technical field of health service management, and the method comprises the steps: carrying out the iterative training of a multi-classifier neural network model (CNN) through employing a marked face image as a training data set (the data set is placed in other proof files), and further improving the accuracy of the model in constitution fusion, thereby enabling the suggestions provided for a user to better meet the actual needs of the user, and improving the user experience. And the reliability of the whole fusion method is improved.

Description

Technical Field

[0001] The present invention discloses a method for fusion analysis of the current facial features of a service object through extracted facial features, predicting the TCM constitution, and providing corresponding life advice. The present invention relates to the field of health service management, and uses annotated facial images as a training data set to iteratively train a multi-classifier neural network model (CNN), further improving the accuracy of the model in constitution fusion, thereby making the advice provided to the user more in line with the actual needs of the user, and improving the reliability of the entire fusion method. Background Art

[0002] Because there is no unified standard for judging facial features and TCM constitution, the accuracy of constitution identification is insufficient, which is specifically reflected in: (1) the judgment between facial features and TCM constitution is unprofessional; (2) the standard for labeling facial training sets is not unified.

[0003] Related technologies: Multi-classifier neural network model (CNN) and Panda are both mature technical methods.

[0004] Reference materials: "Classification and Determination of Constitution in Traditional Chinese Medicine", "Huangdi Neijing", "Diagnosis in Traditional Chinese Medicine", "Constitution in Traditional Chinese Medicine", and "Basic Theory of Traditional Chinese Medicine".

[0005] The key is to improve the accuracy of the fusion model through training set training, thereby improving the accuracy of physical constitution identification.

[0006] In summary, the existing fusion model has insufficient accuracy in constitution identification due to the unprofessional judgment between facial features and TCM constitution, as well as the inconsistent standards for labeling facial training sets. Summary of the invention

[0007] At present, there are certain limitations and shortcomings in using a single facial feature to judge TCM constitution. The facial image itself contains information from multiple aspects, and a single feature is difficult to fully capture these details. A single facial feature provides relatively little information and may not be enough to fully reflect a person's health status and constitution type. The combination of multiple facial features can provide more comprehensive information and help to more accurately judge the constitution type and physical condition.

[0008] This method comprehensively analyzes multiple single facial features rather than relying solely on a single feature. The correspondence between facial features and TCM constitution is more refined, and the model trained with labeled facial images can significantly improve the accuracy of constitution identification. Effective identification of constitution helps to formulate health and treatment plans that better meet personal needs. This method not only enhances the personalization of health management, but also provides users with constitution analysis reports, health risk assessments, and suggestions for improvement in food, clothing, housing, and transportation based on the analysis results, thereby providing individuals with more scientific and effective health management support, while promoting the integration of traditional Chinese medicine wisdom and modern technology.

[0009] The system includes:

[0010] Image receiving module: This module is designed to receive facial images uploaded by the front end.

[0011] Feature extraction module: This module aims to extract facial features from images. Use labeled facial images as data sets to train a neural network model; perform feature extraction using the trained neural network model. A patent application has been submitted for the facial feature extraction method, and the facial dataset has been uploaded to other supporting materials.

[0012] Facial feature calling module: This module aims to retrieve facial features through functions for use by the next module.

[0013] Facial feature fusion module: This module aims to fuse the retrieved facial features through a multi-classifier neural network model to obtain the final constitution type. Use labeled facial images as data sets to train the classifier neural network model, input high-dimensional feature vectors extracted by CNN, and output different constitution categories. Use cross entropy loss function and Adam optimizer for back propagation training. After iterative training until the conditions are met, use the test set to evaluate the model performance (accuracy, recall rate, F1 score, etc.). After the model training is completed, input the retrieved facial features into the model, and obtain the probability of each constitution, and then output the three most likely constitutions according to the principle of maximum probability.

[0014] TCM constitution corresponding to food, clothing, housing and transportation suggestion module: This module uses Panda to read the constitution suggestion comparison table 1 according to the TCM constitution with the highest output probability, and provides corresponding food, clothing, housing and transportation suggestions.

[0015] Result return module: This module aims to pass the obtained physical type and food, clothing, housing and transportation suggestions to the front end. See System Architecture for details. Figure 1 . System flow diagram Figure 2 . Physical constitution recommendation comparison table 1. Physical fitness comparison table 1 Table 1 DETAILED DESCRIPTION

[0016] Step one: Integrate the entire system into the company’s health management system.

[0017] Step 2: Receive facial images transmitted by the front end.

[0018] Step 3: Extract facial features.

[0019] Step 4: Call the extracted facial features.

[0020] Step 5: Input the retrieved facial features into the neural network model, and output the three most likely physical types based on the maximum probability principle.

[0021] Step 6: Read the physical constitution recommendation comparison table 1 and get recommendations on food, clothing, housing and transportation.

[0022] Step 7: Combine the results of the recommendation data, physical characteristics, and physical performance data and return them to the front end. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a system architecture diagram.

[0024] Figure 2 It is a system flow chart.

Claims

1. Based on the multi-classifier neural network model (CNN), the facial features are integrated to determine the final TCM constitution and give corresponding life suggestions.

2. Face training set for iterative training of multi-classifier neural network model (CNN).

Images