Brewing control method for food-nourishing tea based on linkage of traditional Chinese medicine constitution and intelligent equipment

Abstract

The invention discloses a traditional Chinese medicine constitution and intelligent equipment linkage food-nourishing tea brewing control method, and relates to the technical field of intelligent control. According to the traditional Chinese medicine constitution and intelligent equipment linked food-nourishing tea brewing control method, face and tongue coating images of a user are obtained, preliminary feature extraction is performed, key constitution features of the face and tongue coating are extracted through multi-level nonlinear mapping, and the constitution types in the traditional Chinese medicine theory are combined through feature fusion and classification analysis, so that the health-care tea brewing quality is improved. And constructing a traditional Chinese medicine constitution recognition network model, and performing constitution recognition and classification on the user to obtain the constitution type of the user. According to the physique type of the user, tea brewing parameters of the physique tea making machine are adjusted through a fuzzy control algorithm. Therefore, a personalized tea drinking scheme is provided for the user. The intelligent degree and the individuation degree of tea brewing are improved, and the health requirement of a user can be better met.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to a method for controlling the brewing of nourishing tea by linking a traditional Chinese medicine constitution with an intelligent device. Background Art

[0002] With the improvement of people's health awareness, personalized health management and dietary regulation have become important issues in modern life. Traditional Chinese medicine theory emphasizes the concept of "tailoring treatment to individual needs". Especially in terms of physical conditioning, the theory of Chinese medicine constitution proposes different dietary therapy plans by analyzing individual physical differences. However, traditional Chinese medicine constitution diagnosis often relies on experienced Chinese medicine practitioners to make judgments through observation, smell, questioning, and palpation. This method is not only highly subjective, but also lacks accuracy and automation, and it is difficult to meet the needs of large-scale popularization and modern health management. In order to make the Chinese medicine constitution conditioning plan more scientific and efficient, it has become a trend to combine intelligent devices for personalized adjustment, especially the preparation and health management of food and tea.

[0003] Some smart devices have been applied to tea brewing, but the design of most devices is still limited to the control of simple parameters such as time and temperature, and does not provide personalized adjustment according to the user's physical type. Smart tea machines only make simple parameter settings based on the type of tea selected by the user, without considering the specific differences in the user's physical constitution. Current smart tea equipment is difficult to accurately analyze and match the user's physical characteristics. The physical analysis process still relies on traditional subjective judgment or standardized questionnaires, and cannot perform real-time automated diagnosis based on external features such as facial and tongue coating, which greatly affects the intelligence and accuracy of the equipment in actual applications. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method for controlling the brewing of nourishing tea by linking traditional Chinese medicine constitution with intelligent devices, which solves the problems of the above-mentioned background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for controlling the brewing of nourishing tea by linking TCM constitution with intelligent devices, comprising the following steps: S1. Acquire the user's face and tongue coating images, perform preliminary image feature extraction on the user's face and tongue coating images through a convolutional neural network, and perform multi-level nonlinear mapping on the preliminary image features to extract the key physical features of the face and tongue coating; S2. Based on the key physical features of the face and tongue coating, perform feature fusion and classification analysis on the user, and establish a TCM constitution recognition network model in combination with the constitution types in TCM theory; S3. Based on the TCM constitution recognition network model, perform physical feature identification and classification on the user to obtain the user's constitution type; S4. According to the user's constitution type, adjust the tea brewing parameters of the physical tea machine through a fuzzy control algorithm.

[0006] Furthermore, the specific process of performing preliminary image feature extraction on the user's facial and tongue coating images through a convolutional neural network is as follows: inputting the facial and tongue coating images into the input layer of the convolutional neural network; filtering the input images through multiple convolutional layers to identify basic features in the image, including edges and corners; pooling the convolution results to reduce the size of the feature map through maximum pooling and reduce computational complexity; and passing the pooled image features to subsequent layers to extract preliminary image features.

[0007] Furthermore, the preliminary image features are subjected to multi-level nonlinear mapping to extract the key physical features of the face and tongue coating. The specific process is as follows: the preliminary image features are input into the nonlinear activation function layer, and the ReLU activation function is used for nonlinear transformation; the features after nonlinear activation are passed to the deep fully connected layer, and the attention mechanism is introduced. The key areas and significant features in the image are combined, and the key physical features of the face and tongue coating are extracted through further feature fusion, including facial skin color features, tongue coating color and texture features.

[0008] Furthermore, based on the key physical features of the face and tongue coating, the specific process of feature fusion and classification analysis of users is as follows: weighted fusion of facial skin color features and tongue coating color and texture features is performed, and comprehensive physical features are formed through feature splicing in the convolutional neural network; the fused feature data is trained and classified through the support vector machine to identify the physical type that conforms to the theory of traditional Chinese medicine.

[0009] Furthermore, combined with the constitution types in TCM theory, the specific process of establishing a TCM constitution recognition network model is as follows: taking the classification results of the support vector machine as the basis, the classification accuracy is further optimized through the multi-layer perceptron in the neural network; training the network through the back propagation algorithm, adjusting the weights of each layer, and performing model verification and performance evaluation through cross-validation to establish a TCM constitution recognition network model.

[0010] Furthermore, the specific process of identifying and classifying the user's physical characteristics is as follows: according to the trained TCM physical constitution identification network model, the key physical characteristics of the face and tongue coating input by the user are passed into the network model; the features are processed by a multi-layer perceptron, the physical constitution classification calculation is performed, and the classification probability of the physical constitution type is output, and matched with the TCM physical constitution type to obtain the physical constitution matching confidence index.

[0011] Furthermore, the specific process of obtaining the user's physique type is as follows: compare the physique matching confidence index with the classification threshold; if the physique matching confidence index is greater than or equal to the classification threshold, determine that the user's physique type is the corresponding TCM constitution type, and output the matched constitution type as the user's constitution type; if the physique classification confidence index is less than the classification threshold, re-extract the key constitution features.

[0012] Furthermore, the specific process of adjusting the tea brewing parameters of the physical constitution tea machine through the fuzzy control algorithm is as follows: according to the user's physical constitution type, a fuzzy control rule library is established, and the fuzzy control rules are designed, the user's physical constitution type is set as the input variable, and the tea brewing parameters are set as the output variables; the input variables are fuzzified through the fuzzy control rule library to generate fuzzy values ​​of the tea brewing parameters; the fuzzy reasoning results are defuzzified to generate the tea brewing parameters of the user's physical constitution, including water temperature, brewing time, tea amount, and tea type.

[0013] The present invention has the following beneficial effects: (1) This method of controlling the brewing of tea based on TCM constitution and intelligent devices provides a more accurate and scientific basis for constitution identification by acquiring the user's facial and tongue coating images and extracting the key physical features of the face and tongue coating using a convolutional neural network. It can efficiently obtain key feature information related to TCM constitution. Through multi-level nonlinear mapping, the key physical features of the face and tongue coating are extracted, providing accurate data support for subsequent constitution classification analysis, thereby ensuring the scientificity and accuracy of constitution identification.

[0014] (2) This method of controlling the brewing of nourishing tea drinks by linking TCM constitution with intelligent devices, by establishing a constitution recognition network model based on TCM theory, classifies the user's constitution, accurately obtains their constitution type, and adjusts the tea brewing parameters in combination with a fuzzy control algorithm. According to the user's constitution type, the water temperature, brewing time, and amount of tea leaves are dynamically adjusted to achieve personalized brewing control, optimize the health conditioning effect of tea drinks, and improve the health management effect and user experience of tea drinks. It ensures that every user can get the tea drink that best suits their constitution, while also optimizing the user experience of the tea machine and improving the user's health conditioning effect, thereby achieving a deep integration of TCM theory and modern intelligent devices.

[0015] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of a method for controlling the brewing of nourishing tea by linking TCM constitution with intelligent equipment according to the present invention. DETAILED DESCRIPTION

[0017] The embodiment of the present application solves the problem of inaccurate matching of personalized tea brewing with the user's physique through a method for controlling the brewing of food-nourishing tea drinks by linking the traditional Chinese medicine constitution with intelligent devices. The method uses a convolutional neural network to extract features from the user's facial and tongue coating images, and analyzes them in combination with the traditional Chinese medicine constitution classification model to accurately identify the user's constitution type. Based on the recognition results, a fuzzy control algorithm is used to dynamically adjust the brewing parameters of the tea machine, so that the tea brewing method can adapt to the needs of different physiques, realize intelligent and personalized tea conditioning solutions, and improve user experience and health management effects.

[0018] The overall idea of ​​the solution in the embodiments of this application is as follows: Obtain the user's facial and tongue coating images, perform preliminary image feature extraction on the user's facial and tongue coating images through a convolutional neural network, and perform multi-level nonlinear mapping on the preliminary image features to extract key physical features of the face and tongue coating.

[0019] Based on the key physical features of the face and tongue coating, the users are subjected to feature fusion and classification analysis, and a TCM physical constitution recognition network model is established in combination with the physical constitution types in TCM theory.

[0020] According to the TCM constitution recognition network model, the user's constitution characteristics are identified and classified to obtain the user's constitution type.

[0021] According to the user's physical constitution, the tea brewing parameters of the physical constitution tea drinking machine are adjusted through fuzzy control algorithm.

[0022] See also Figure 1 The embodiment of the present invention provides a technical solution: a method for controlling the brewing of nourishing tea by linking TCM constitution with intelligent devices, comprising the following steps: S1. Acquire the user's face and tongue coating images, perform preliminary image feature extraction on the user's face and tongue coating images through a convolutional neural network, and perform multi-level nonlinear mapping on the preliminary image features to extract the key physical features of the face and tongue coating; S2. Perform feature fusion and classification analysis on the user based on the key physical features of the face and tongue coating, and establish a TCM constitution recognition network model in combination with the physical types in TCM theory; S3. Perform physical feature recognition and classification on the user based on the TCM constitution recognition network model to obtain the user's physical type; S4. Adjust the tea brewing parameters of the physical tea machine based on the user's physical type through a fuzzy control algorithm.

[0023] In this embodiment, S2. Obtaining user's facial and tongue coating images: Facial and tongue coating images: refers to photos of the user's face and tongue coating taken by high-definition cameras and other devices. The face and tongue coating are considered by traditional Chinese medicine to be the external manifestations of the body's internal health and constitution, and can reveal the health status of qi, blood, internal organs, etc. Convolutional neural network: is a deep learning model commonly used in image processing tasks. It extracts features from images through multiple convolutional layers and can automatically identify important patterns in images, such as edges, textures, colors, etc. CNN can automatically learn and extract key features of facial skin color and tongue coating from images without manual feature annotation. Image feature extraction: refers to extracting key information from the input image that helps to identify or classify. For example, facial skin color, tongue coating color, texture, etc., are all closely related to the individual's constitution. S2. Feature fusion and classification analysis: Feature fusion: refers to combining feature information from different sources (such as facial features, tongue coating features, etc.) to obtain more comprehensive and accurate physical feature information. This fusion can improve the accuracy of physical classification, because a single feature may have limitations, and the combination of multiple features can better represent the user's physical type. Classification analysis: It is to use machine learning algorithms to assign data to different categories. In this step, based on the classification method of constitution in traditional Chinese medicine theory, the extracted features are input into the model to classify the constitution type, such as qi deficiency, blood deficiency, damp heat, etc. TCM constitution recognition network model: It is a recognition model built based on machine learning and deep learning technology combined with TCM constitution theory. The model can identify the user's constitution type based on the characteristics of the face and tongue coating, including nine different constitution types (such as yin deficiency, yang deficiency, damp heat, etc.). S3. Constitution feature recognition and classification: Constitution recognition and classification: It is to judge and classify which TCM constitution type the user belongs to based on the extracted facial and tongue coating features through the trained TCM constitution recognition network model. This step outputs the constitution category through data input to provide a reference for the next step of tea brewing. Classification probability and constitution matching: During the classification process, the system outputs the matching probability of each constitution type. Based on this probability, the system can confirm the most suitable constitution type and use it as the user's final constitution type. S4. Adjusting tea brewing parameters: Fuzzy control algorithm: is a control algorithm based on fuzzy sets and fuzzy logic, which can process uncertain or ambiguous information. In the present invention, the fuzzy control algorithm is used to automatically adjust the tea brewing parameters, such as water temperature, brewing time, type of tea, etc., according to the user's physical type. The fuzzy control algorithm can provide flexible and personalized tea brewing methods according to the needs of different physical conditions. Tea brewing parameters: including factors such as water temperature, brewing time, and amount of tea. These parameters directly affect the taste, effect, and health value of tea. According to the needs of different physical conditions, these parameters need to be appropriately adjusted to ensure that tea can achieve the best health-preserving effect.Fuzzy reasoning and defuzzification: Fuzzy reasoning refers to generating reasoning results based on the fuzzy control rule base and the user's physical type and fuzzy input. Defuzzification converts the results of fuzzy reasoning into specific control values ​​(such as water temperature, brewing time, etc.) to achieve specific control of the tea brewing process.

[0024] Specifically, the specific process of extracting preliminary image features of the user's facial and tongue coating images through a convolutional neural network is as follows: the facial and tongue coating images are input into the input layer of the convolutional neural network; the input images are filtered through multiple convolutional layers to identify basic features in the image, including edges and corners; the convolution results are pooled to reduce the size of the feature map through maximum pooling and reduce computational complexity; the pooled image features are passed to subsequent layers to extract preliminary image features.

[0025] In this embodiment, input layer: face and tongue image input: first, the user's face and tongue image are passed as input data to the convolutional neural network. Images are usually two-dimensional (width and height), but for ease of processing, images are usually converted into multi-channel (such as RGB three color channels) data format. The image enters the network through the input layer to provide basic data for subsequent calculations and feature extraction of the network. Convolution layer: filtering operation: The main function of the convolution layer is to filter the input image through "convolution operation". This process involves using a small filter (or convolution kernel) to slide the input image and calculate the local weighted sum of the convolution kernel and each area of ​​the image. The weighted sum of the convolution kernel at different positions generates a set of feature maps. The filter can identify the basic structural features of the image, such as edges, corners, and textures. Learning of convolution kernels: Through training data, the convolutional neural network can learn a set of effective convolution kernels that can extract different features in the image. For face and tongue images, the convolution layer of the convolutional neural network can identify important features related to physique, such as facial contours and tongue color changes. Pooling Layer: Maximum pooling operation: Pooling layer is often used to reduce the spatial size of convolution results to reduce computational complexity and improve computational efficiency. Maximum pooling is a common method used in pooling layer, which performs dimensionality reduction by selecting the maximum value in a small area. For example, a region of the input image may contain multiple pixel values, and maximum pooling selects the maximum value of these pixel values, reducing the resolution and amount of data of the image. Dimensionality reduction and computational complexity: Pooling can not only reduce the spatial dimension of the data, but also make the network more robust (insensitive to position changes), avoid overfitting, and speed up the training and reasoning of the network. Pass to subsequent layers: Pass the pooled image features to subsequent layers: After the pooling operation is completed, the obtained feature map (the result after dimensionality reduction) will be passed to the subsequent layers of the network (usually more convolutional layers or fully connected layers) for further processing and feature extraction. The image features after pooling have been simplified and retain the most important information in the image. The subsequent layers will further extract more advanced features on this basis, such as the comprehensive features of the face and tongue coating.

[0026] Specifically, the specific process of performing multi-level nonlinear mapping on the preliminary image features and extracting the key physical features of the face and tongue coating is as follows: the preliminary image features are input into the nonlinear activation function layer, and the ReLU activation function is used for nonlinear transformation; the features after nonlinear activation are passed to the deep fully connected layer, and the attention mechanism is introduced. The key areas and significant features in the image are combined, and the key physical features of the face and tongue coating are extracted through further feature fusion, including facial skin color features, tongue coating color and texture features.

[0027] In this embodiment, the preliminary image features are input into the nonlinear activation function layer. Preliminary image features: First, the facial and tongue coating images are processed by a convolutional neural network (CNN) to extract preliminary image features, which are usually some low-level features, such as edges, corners, and textures. Nonlinear activation function layer: These preliminary features are input into the nonlinear activation function layer. Nonlinear activation functions can help neural networks introduce nonlinear transformations so that the network can learn more complex data relationships. ReLU activation function: ReLU is a common activation function with the formula ReLU(x) = max(0,x), that is, if the input is a negative number, the output is 0; if the input is a positive number, the output is the value. ReLU has the advantages of simplicity, effectiveness, and small amount of calculation. It can help the network avoid the gradient vanishing problem during training and accelerate the convergence process. The role of nonlinear transformation: After using the ReLU activation function, the network can learn and express the nonlinear relationship of the input features. This means that the neural network can not only extract low-level features such as edges or textures, but also learn higher-level complex patterns, such as key physical characteristics of the face and tongue coating. Deep fully connected layer: After being transformed by nonlinear activation function, the features are passed to the deep fully connected layer of the network. The role of the fully connected layer is to further combine and process the nonlinearly activated features through the connection of multiple neurons, so as to extract more abstract and high-dimensional feature representations. Feature fusion: The fully connected layer can also fuse features from different sources (such as facial features and tongue coating features) through a multi-level structure. In this way, the network can integrate feature information from multiple sources and provide richer feature inputs for subsequent classification tasks. Attention mechanism: In order to enhance the focus on important features, the attention mechanism is introduced. The core idea of ​​the attention mechanism is to enable the network to focus on key areas or significant features by assigning different weights to input features. This mechanism is similar to the concentration process of human attention, which can help the network give priority to important parts (for example, facial skin color features or tongue coating color and texture features) when processing complex images. In image processing, the attention mechanism usually weights important areas in the image (such as facial contours, tongue coating color changes), so that the network can learn more accurate physical characteristics. Facial skin color features: The skin color of the face reflects the state of qi and blood in the body, such as health, qi deficiency, blood deficiency, etc. Through the processing of nonlinear activation functions and deep fully connected layers, the network can extract detailed information related to skin color, such as changes in facial redness or paleness, which in turn reflects the user's physical characteristics. Tongue coating color and texture features: The color and texture of the tongue coating are also very important for the diagnosis of TCM constitution. Changes in the color of the tongue coating, such as white, yellow, and red, can reveal pathological conditions such as cold, heat, dampness, and heat toxins in the body. Texture changes may be related to the body's qi and blood and the function of the internal organs.Through multi-level nonlinear mapping and feature fusion, the network can extract these detailed features on the tongue coating, providing key information for the identification of constitution types.

[0028] Specifically, based on the key physical features of the face and tongue coating, the specific process of feature fusion and classification analysis of users is as follows: weighted fusion of facial skin color features and tongue coating color and texture features, forming comprehensive physical features through feature splicing in the convolutional neural network; training and classifying the fused feature data through support vector machine to identify the physical type that conforms to the theory of traditional Chinese medicine.

[0029] In this embodiment, facial skin color features are weightedly fused with tongue coating color and texture features. Facial skin color features: facial skin color features can reflect the state of human qi and blood. For example, a healthy person has a ruddy complexion, while a person with qi deficiency or blood deficiency may have a pale complexion. Facial skin color features are used in traditional Chinese medicine to diagnose whether qi and blood are full. Tongue coating color and texture features: the color and texture of the tongue coating play an important role in traditional Chinese medicine diagnosis. For example, a white tongue coating may indicate that the body has a heavier cold, while a red tongue coating may indicate that the body has a stronger heat. The texture of the tongue coating helps to further analyze the health status of the internal organs. Weighted fusion: After the facial skin color features and the color and texture features of the tongue coating are extracted, in order to optimize the contribution of these features in classification, they are respectively given certain weights (i.e., weighted fusion). For example, the reflection of facial skin color on the state of qi and blood may be more important than the color and texture of the tongue coating, so the facial skin color features can be given a higher weight, and vice versa. Weighted fusion helps to highlight the features that have a greater impact on physical identification. Feature splicing: The weighted fusion of facial skin color features and tongue coating color and texture features are combined into a comprehensive physical feature through the feature splicing method in the convolutional neural network. Through splicing, feature data from two different sources can be combined to generate a comprehensive physical feature vector containing multi-dimensional information. In this way, the model can consider multiple signals of the face and tongue coating at the same time, thereby providing more comprehensive feature data for physical classification. Support vector machine (SVM): Support vector machine is a supervised learning algorithm that is often used for classification tasks. It separates data of different categories by finding an optimal hyperplane and has strong generalization ability. Through training, SVM can establish a classification model based on the relationship between known physical types and features. Training and classification: The SVM model is trained using training data, which includes the user's facial skin color and tongue coating features, as well as their corresponding physical types (such as "Qi deficiency", "Yin deficiency", etc.). The trained SVM model will be able to predict new user data and classify the user's physical type. Through SVM classifier analysis, the corresponding physical type of the user is finally output. According to TCM theory, physique types are generally divided into the following nine types: Peaceful physique: manifested by good health, good complexion, and vigorous spirit, which is the most ideal physique type in TCM. People with a peaceful physique are not prone to illness, with abundant qi and blood, and balanced organ functions. Qi deficiency physique: manifested by symptoms such as fatigue, shortness of breath, laziness, and easy cold. People with a qi deficiency physique have insufficient qi (vitality) in their bodies, resulting in a decrease in the body's ability to resist disease. Blood deficiency physique: manifested by symptoms such as pale complexion, dizziness, palpitations, and insomnia. People with a blood deficiency physique have insufficient blood, resulting in a lack of nourishment in the body, weakness, dryness, etc. Yin deficiency physique: manifested by symptoms such as dry mouth, insomnia, and hot hands and feet. People with a yin deficiency physique have insufficient yin fluid (such as body fluids, blood, etc.) in their bodies, and are prone to symptoms such as internal heat and thirst. Yang deficiency physique: manifested by symptoms such as fear of cold, cold hands and feet, and pale complexion.People with Yang deficiency constitution have insufficient Yang in their bodies, which leads to poor blood circulation, manifested as a feeling of coldness and weak digestive function. Phlegm-damp constitution: manifested by obesity, easy sweating, greasy tongue coating and other symptoms. People with phlegm-damp constitution have too much moisture in their bodies, which leads to poor blood circulation, easy to form phlegm and dampness, affecting health. Damp-heat constitution: manifested by bitter mouth, dry mouth, yellow greasy tongue coating and other symptoms. People with damp-heat constitution have too much moisture and heat in their bodies, which can easily cause discomfort in the body, such as indigestion, skin diseases, etc. Wind-cold constitution: manifested by fear of cold, wind-cold symptoms on the body surface, easy to catch cold and other symptoms. People with wind-cold constitution are invaded by wind and cold in the body, which can easily lead to symptoms such as joint pain and muscle stiffness. Special constitution: also called special constitution, including allergic constitution, susceptible constitution, etc. People with this constitution have strong reactions to the external environment or certain substances (such as pollen, food), and are prone to allergies or other special reactions.

[0030] Specifically, combining the constitution types in TCM theory, the specific process of establishing a TCM constitution recognition network model is as follows: taking the classification results of the support vector machine as the basis, further optimizing the classification accuracy through the multi-layer perceptron in the neural network; training the network through the back propagation algorithm, adjusting the weights of each layer, performing model verification and performance evaluation through cross-validation, and establishing a TCM constitution recognition network model.

[0031] In this implementation scheme, support vector machine (SVM): First, support vector machine is used to preliminarily classify the physical features in the facial and tongue coating images. Support vector machine is a supervised learning model used for classification tasks, which separates sample data of different categories by finding the best segmentation hyperplane. SVM has strong classification ability and is particularly suitable for classification problems of high-dimensional data. SVM results as a basis: After preliminary classification by SVM, the classification results are used as input of the network model to provide a basis for further optimization and accuracy improvement. This process provides a preliminary division of physical types and provides a good starting point for subsequent neural network optimization and accurate classification. Multilayer perceptron (MLP): Multilayer perceptron is a feedforward neural network composed of multiple layers of neurons, including at least one hidden layer. Multilayer perceptron processes input data through nonlinear activation functions (such as ReLU) and uses fully connected layers for information transfer. It can learn complex patterns and relationships from data and improve the accuracy of classification. Optimize classification accuracy: Take the SVM classification results as input and use multilayer perceptron to further optimize the classification results. Through the training of multi-layer perceptron, the complex relationship between facial and tongue coating features and physical type can be better explored, thereby improving the accuracy of classification. Back propagation algorithm: Back propagation is one of the core algorithms for neural network training. It calculates the gradient of the loss function, back propagates the error, and updates the weights of each layer in the network according to the gradient. Through back propagation, the neural network can adjust the weights in each training, thereby continuously optimizing the classification results. Training process: During the network training process, the input facial and tongue coating features enter the network through forward propagation, are processed by the multi-layer perceptron, and finally the classification results are obtained. Then, by calculating the loss function (such as cross entropy loss), the network weights are updated through the back propagation algorithm, so that the network gradually approaches the optimal classification result. Cross validation: Cross validation is a commonly used method to evaluate model performance, especially for preventing overfitting. In the cross validation process, the data set is divided into multiple subsets, and the model is trained on different training sets and evaluated on different validation sets to obtain the generalization ability of the model. Performance evaluation: Through cross validation, the model can be validated on multiple different data sets to ensure that the network model has good robustness and accuracy. The evaluation indicators include accuracy, recall, F1 score, etc. These indicators help understand the model's classification ability in different constitution types. TCM constitution recognition network model: Through SVM, neural network optimization (MLP), back propagation training and cross-validation, a high-precision TCM constitution recognition network model is finally constructed. The model can accurately identify and classify different constitution types based on facial and tongue coating feature data, and provide a basis for subsequent personalized health management.

[0032] Specifically, the specific process of identifying and classifying the user's physical characteristics is as follows: according to the trained TCM physical constitution recognition network model, the key physical characteristics of the face and tongue coating input by the user are passed into the network model; the features are processed by a multi-layer perceptron, the physical constitution classification calculation is performed, and the classification probability of the physical constitution type is output, and matched with the TCM physical constitution type to obtain the physical constitution matching confidence index.

[0033] In this implementation scheme, the formula for calculating the physical fitness classification is: is the input facial and tongue key physical feature vector, and For the The weight matrix and bias term of the layer neural network, is the activation function (such as ReLU or Sigmoid). Then in each layer of the multi-layer perceptron, the calculation process of the neural network can be expressed as: ; Parameter explanation: :No. The linear combination result of the layer represents the weighted calculation of the input features through the weight matrix and the bias term. :No. The weight matrix of the layer determines the importance of the input features in the network. :No. The activation output of the layer is used as the input of the current layer. :No. The bias term of the layer helps adjust the range of the network output. :No. The activation function of the layer (such as ReLU or Sigmoid) is used to introduce nonlinear relationships and help the model learn complex features. :No. The activation output of the layer represents the response of the neurons in this layer to the input features. Output classification probability calculation ; Parameter explanation: : The activation output of the last layer represents the result after feature conversion and processing of all layers. Softmax: The activation function converts the output of the network into a probability distribution so that the sum of the probabilities of all categories is 1. Each output represents the probability that the user belongs to that type of physique. Physical type matching and confidence calculation ; Parameter explanation: :No. The classification probability of a type of physique indicates the probability that the user belongs to this type of physique. : Matching function, measuring the user's constitution and the constitution type in traditional Chinese medicine theory The value is 1 for a complete match and less than 1 for a partial match. Final physical match confidence index calculation ; Parameter explanation: :No. The matching confidence index of the physical constitution reflects the confidence of the physical constitution classification. :The total number of constitution types. The final physique matching confidence index is obtained by weighted aggregation of the matching confidences of all physique types.

[0034] Specifically, the specific process of obtaining the user's physique type is as follows: compare the physique matching confidence index with the classification threshold. If the physique matching confidence index is greater than or equal to the classification threshold, the user's physique type is determined to be the corresponding TCM constitution type, and the matched constitution type is output as the user's constitution type; if the physique classification confidence index is less than the classification threshold, the key constitution feature extraction is performed again.

[0035] In this embodiment, comparison of the physique matching confidence index and the classification threshold: First, the system will compare the matching confidence index calculated by the physique classification with the preset classification threshold. The classification threshold is a numerical value, usually a value between 0 and 1, which indicates the system's confidence requirement for the classification result. The classification threshold can be adjusted according to system requirements to ensure that the output physique type has sufficient credibility. Matching confidence index: As mentioned above, the physique matching confidence index is a quantitative value that indicates the system's confidence in identifying the user's physique type. This value is usually calculated based on the user's facial and tongue coating features through methods such as multi-layer neural networks and support vector machines. Classification threshold: It is a fixed value set by the system, indicating that when it is above a certain confidence level, the system believes that the recognition result is accurate enough and can output the result. If the matching confidence is lower than this threshold, the system believes that the result is not certain enough and may need further confirmation. When the matching confidence index is greater than or equal to the classification threshold, the system will believe that the recognition result is reliable enough, output the physique type, and use it as the user's physique type. At this time, the system will provide the matched constitution type (for example, TCM constitution types such as Qi deficiency and Yin deficiency) to the user for subsequent adjustment of tea brewing parameters. For example, if the system determines that a user's constitution type is "Yang deficiency constitution" and the matching confidence index is higher than the set classification threshold, the constitution type is output as the user's constitution type. When the matching confidence index is less than the classification threshold, the system will consider the current constitution classification result to be uncertain or inaccurate, and the system will re-extract the constitution features and perform the constitution classification analysis again. At this time, it may be necessary to obtain more facial and tongue coating images, or re-evaluate under other conditions to improve the confidence of constitution matching. For example, if the system believes that the key constitution features in the facial or tongue coating images are not obvious enough or are affected by environmental factors (such as light, angle, etc.), the matching confidence may be lower than the set classification threshold. The system will return to the feature extraction step and re-extract the key constitution features of the facial and tongue coating images to improve the confidence of constitution matching.

[0036] Specifically, the specific process of adjusting the tea brewing parameters of the physical constitution tea machine through the fuzzy control algorithm is as follows: according to the user's physical constitution type, a fuzzy control rule library is established, and the fuzzy control rules are designed, and the user's physical constitution type is set as the input variable, and the tea brewing parameters are set as the output variables; the input variables are fuzzified through the fuzzy control rule library to generate fuzzy values ​​of the tea brewing parameters; the fuzzy reasoning results are defuzzified to generate the tea brewing parameters of the user's physical constitution, including water temperature, brewing time, tea amount, and tea type.

[0037] In this implementation scheme, a fuzzy control rule base is established. First, a fuzzy control rule base needs to be established according to different user constitution types and corresponding tea brewing requirements. This rule base contains the fuzzy relationship between constitution type and tea brewing parameters. According to different types of TCM constitutions (such as qi deficiency, yin deficiency, yang deficiency, etc.), each constitution is suitable for different tea brewing parameters. The rule base will list such rules: if the user's constitution is "yin deficiency", the water temperature should be lower, the brewing time should be shorter, and the amount of tea should be moderate; if the user's constitution is "yang deficiency", the water temperature should be higher, the brewing time should be longer, and the amount of tea should be moderate. Design fuzzy control rules. In the rule base, the user's constitution type is used as an input variable, and the parameters of tea brewing (water temperature, brewing time, tea amount, tea type) are used as output variables. Specific fuzzy control rules can be set based on experience and TCM theory. For example, the constitution type "Qi deficiency" corresponds to milder brewing parameters; the constitution type "Yang deficiency" corresponds to higher temperature brewing parameters. The key to fuzzy processing of input variables lies in the "fuzzification" and "defuzzification" steps. First, the input variable (i.e. the user's constitution type) is fuzzified. Specifically, the discrete constitution types (such as "Qi deficiency" and "Yang deficiency") are mapped to fuzzy values ​​(usually a value between 0 and 1). For example, for the constitution type "Qi deficiency", a fuzzy value of 0.8 can be obtained; for the constitution type "Yin deficiency", a fuzzy value of 0.6 can be obtained; for the constitution type "Yang deficiency", a fuzzy value of 0.9 can be obtained. Fuzzy reasoning generates tea brewing parameters. Through the fuzzy control rule base and the fuzzified constitution type, the system will perform fuzzy reasoning to generate fuzzy values ​​of tea brewing parameters. The core of fuzzy reasoning is to match the input fuzzy constitution type with the brewing parameters in the rule base through reasoning rules to obtain the fuzzy value of each output variable. Specifically, the following reasoning process can be used: if the constitution is "Qi deficiency" and its fuzzy value is 0.8, then the fuzzy value of water temperature in the "medium" range (for example, 65-70°C), the amount of tea is "medium", and the brewing time is "short" is generated; if the constitution is "Yang deficiency" and its fuzzy value is 0.9, then the fuzzy value of water temperature in the "high" range (for example, 85-90°C), the amount of tea is "more", and the brewing time is "long" is generated. The result of fuzzy reasoning is a fuzzy value, and we need a clear output (specific water temperature, brewing time, and amount of tea). At this time, the system converts the fuzzy value into a specific tea brewing parameter through defuzzification. Defuzzification methods include: Weighted average method: weighted average calculation of all possible output values ​​to obtain specific parameter values. Maximum membership method: select the result with the largest membership of the fuzzy set as the final output parameter. Example Assume that a user is identified as a "Qi deficiency" constitution, and the fuzzy value of this constitution type is 0.8.According to the designed fuzzy control rules, the system will generate the following brewing parameters: Water temperature: Based on the reasoning in the rule base, the system may determine that the water temperature should be in the "medium" range (e.g. 70°C). Brewing time: The brewing time is short (e.g. 3 minutes). Amount of tea: The amount of tea is moderate (e.g. 5 grams). Type of tea: The type of tea that may be suitable for Qi deficiency constitution, such as "yellow tea" or "flower tea".

[0038] In summary, this application has at least the following effects: A method for controlling the brewing of tea based on traditional Chinese medicine constitution and intelligent devices. By combining facial and tongue coating image feature extraction, constitution type recognition and fuzzy control algorithm, it can intelligently adjust tea brewing parameters according to the user's traditional Chinese medicine constitution type and provide personalized and accurate tea customization solutions. By combining convolutional neural networks with support vector machines, this method can accurately extract key constitution features from facial and tongue coating images, ensure that the user's constitution classification is highly matched with traditional Chinese medicine theory, and improve the accuracy and practicality of constitution recognition. It not only improves the intelligence and accuracy of tea brewing, but also realizes personalized health management, optimizes user experience, and ensures the effective integration of traditional Chinese medicine theory and modern intelligent technology.

[0039] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0040] The present invention is described with reference to flowcharts and / or block diagrams of systems, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0041] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0042] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0043] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0044] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for controlling the brewing of food-nourishing tea by linking traditional Chinese medicine constitution with intelligent equipment, characterized in that: The following steps are involved: S1. Obtain the user's facial and tongue coating images, perform preliminary image feature extraction on the user's facial and tongue coating images through a convolutional neural network, and perform multi-level nonlinear mapping on the preliminary image features to extract key physical features of the face and tongue coating; S2. Based on the key physical features of the face and tongue coating, perform feature fusion and classification analysis on the user, and establish a TCM physical identification network model based on the physical types in TCM theory; S3. Identify and classify the user's physical characteristics according to the TCM physical constitution identification network model to obtain the user's physical constitution type; S4. According to the user's physical condition, the tea brewing parameters of the physical condition tea drinking machine are adjusted through the fuzzy control algorithm.

2. According to claim 1, a method for controlling the brewing of food and tea by linking TCM constitution with intelligent devices, characterized in that: The specific process of performing preliminary image feature extraction on the user's face and tongue coating images through a convolutional neural network is as follows: Feed the face and tongue images into the input layer of the convolutional neural network; The input image is filtered through multiple convolutional layers to identify basic features in the image, including edges and corners; Pooling is performed on the convolution results to reduce the size of the feature map and the computational complexity through maximum pooling. The pooled image features are passed to the subsequent layers to extract preliminary image features.

3. According to claim 2, a method for controlling the brewing of food and tea by linking TCM constitution with intelligent devices, characterized in that: The specific process of performing multi-level nonlinear mapping on the preliminary image features and extracting the key physical features of the face and tongue coating is as follows: The preliminary image features are input into the nonlinear activation function layer and the ReLU activation function is used for nonlinear transformation; The features after nonlinear activation are passed to the deep fully connected layer, and the attention mechanism is introduced. The key areas and significant features in the image are combined, and through further feature fusion, the key physical features of the face and tongue coating are extracted, including facial skin color features, tongue coating color and texture features.

4. According to claim 3, a method for controlling the brewing of food and tea by linking TCM constitution with intelligent devices, characterized in that: Based on the key physical characteristics of the face and tongue coating, the specific process of feature fusion and classification analysis of users is as follows: The facial skin color features and tongue coating color and texture features are weighted and fused, and the comprehensive physical characteristics are formed through feature splicing in the convolutional neural network; The fused feature data is trained and classified through support vector machine to identify the constitution type that conforms to the theory of traditional Chinese medicine.

5. According to claim 4, a method for controlling the brewing of food and tea by linking TCM constitution with intelligent devices, characterized in that: Combining the constitution types in TCM theory, the specific process of establishing a TCM constitution identification network model is as follows: Based on the classification results of support vector machine, the classification accuracy is further optimized through the multi-layer perceptron in the neural network. The network is trained through the back-propagation algorithm, the weights of each layer are adjusted, the model is verified and the performance is evaluated through cross-validation, and a network model for TCM constitution recognition is established.

6. The method for controlling the brewing of food and tea by linking TCM constitution with intelligent equipment according to claim 5, characterized in that: The specific process of identifying and classifying users' physical characteristics is as follows: According to the trained TCM constitution recognition network model, the key constitution features of the face and tongue coating input by the user are transmitted to the network model; The features are processed by a multi-layer perceptron to perform constitution classification calculations, output the classification probability of the constitution type, and match it with the TCM constitution type to obtain the constitution matching confidence index.

7. The method for controlling the brewing of food and tea by linking TCM constitution with intelligent devices according to claim 6, characterized in that: The specific process of obtaining the user's physical type is as follows: Compare the constitution matching confidence index with the classification threshold. If the constitution matching confidence index is greater than or equal to the classification threshold, determine that the user's constitution type is the corresponding TCM constitution type, and output the matched constitution type as the user's constitution type. If the confidence index of the physical fitness classification is less than the classification threshold, the key physical fitness features are extracted again.

8. The method for controlling the brewing of food and tea by linking TCM constitution with intelligent equipment according to claim 7, characterized in that: The specific process of adjusting the tea brewing parameters of the physical tea machine through the fuzzy control algorithm is as follows: According to the user's physical type, a fuzzy control rule base is established, and fuzzy control rules are designed, with the user's physical type set as the input variable and the tea brewing parameters set as the output variable; The input variables are fuzzified by using the fuzzy control rule base to generate fuzzy values ​​of tea brewing parameters; The fuzzy reasoning results are defuzzified to generate tea brewing parameters for the user's constitution, including water temperature, brewing time, tea amount, and tea type.

Images