Massage device control method based on intelligent ai recognition of back acupoints

Abstract

The application provides a massage equipment control method based on intelligent AI recognition of back acupoints, which comprises the following steps: collecting a back image of a user, performing feature extraction on the back image to obtain back muscle texture and skeletal contour features in the image, highlighting the region features of the existing acupoints through an attention mechanism to obtain an acupoint candidate feature vector; inputting the acupoint candidate feature vector into a trained acupoint positioning model to obtain an acupoint positioning result; generating an optimization strategy based on the acupoint positioning result; obtaining a massage demand input by the user, matching a corresponding initial massage strategy based on the massage demand by using a decision tree algorithm; optimizing the initial massage strategy based on the optimization strategy to obtain a personalized massage strategy; and generating a control instruction of the massage equipment based on the personalized massage strategy to control the massage equipment to complete back acupoint massage. In the application, the defect that current massage strategies cannot be quickly generated based on the differences of users is overcome.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a massage device control method based on intelligent AI recognition of back acupoints. BACKGROUND

[0002] In traditional Chinese medicine theory, back acupoints are closely related to human meridians and visceral functions. Precise massage of specific acupoints can achieve the effects of regulating qi and blood, relieving pain, and enhancing immunity. With the rapid development of the health industry, automated and intelligent massage devices have gradually become popular. However, there are still significant technical bottlenecks in back acupoint massage technology, making it difficult to meet users' demand for precise and personalized massage.

[0003] Existing massage technology mainly relies on fixed programs or user manual selection of preset modes, lacking adaptability to individual differences.

[0004] On the one hand, most massage devices use a fixed acupoint map + standardized massage program scheme, without considering physiological differences in back morphology and acupoint location of different users. For example, there are significant differences in muscle thickness and skeletal structure between adults and the elderly. The same set of acupoint positioning standards may cause massage position deviation, affecting the massage effect.

[0005] On the other hand, traditional massage strategies are mostly based on experience and cannot meet the complex and diverse massage needs of users. For example, when users have complex needs such as relieving neck and shoulder fatigue and improving sleep quality, existing technology cannot quickly generate targeted massage strategies, often resulting in single massage methods, unreasonable force and duration, etc. SUMMARY

[0006] The main purpose of the present application is to provide a massage device control method based on intelligent AI recognition of back acupoints, aiming to overcome the defect of being unable to quickly generate targeted massage strategies based on user differences.

[0007] To achieve the above purpose, the present application provides a massage device control method based on intelligent AI recognition of back acupoints, comprising the following steps:

[0008] Collecting the back image of the user, extracting features from the back image, obtaining the back muscle texture and skeletal contour features in the image, highlighting the regional features of the existing acupoints through attention mechanism, and obtaining the acupoint candidate feature vector;

[0009] Inputting the acupoint candidate feature vector into the trained acupoint positioning model to obtain the acupoint positioning result; based on the acupoint positioning result, generating an optimization strategy;

[0010] Obtaining a massage requirement input by a user, matching an initial massage strategy corresponding to the massage requirement based on the massage requirement by using a decision tree algorithm, optimizing the initial massage strategy based on the optimization strategy to obtain a personalized massage strategy;

[0011] Generating a control instruction of a massage device based on the personalized massage strategy to control the massage device to complete back acupoint massage.

[0012] Further, the personalized massage strategy includes parameters of massage methods, massage intensity, and massage duration.

[0013] Further, based on the acupoint positioning result, an optimization strategy is generated, including:

[0014] Performing acupoint correlation analysis on the acupoint positioning result to construct an acupoint neighborhood relationship graph to calculate the spatial distance, meridian correlation, and functional synergy between acupoints, and obtain an acupoint correlation matrix;

[0015] Based on the acupoint correlation matrix, performing acupoint clustering analysis by using a graph neural network algorithm to identify functionally related acupoint groups, extracting core acupoints and edge acupoints of each acupoint group, and obtaining an acupoint clustering result;

[0016] Performing massage priority evaluation on the acupoint clustering result to obtain a priority ranking result;

[0017] Based on the priority ranking result, an optimization strategy including acupoint group massage order, time allocation proportion, and intensity adjustment coefficient is generated; wherein the intensity adjustment coefficient is dynamically adjusted according to the position of the acupoint in the cluster, and the intensity adjustment coefficient of the core acupoint is higher than that of the edge acupoint.

[0018] Further, based on the massage requirement, an initial massage strategy corresponding to the massage requirement is matched by using a decision tree algorithm, including:

[0019] Performing semantic feature extraction on the text corresponding to the massage requirement, identifying the intensity level of the requirement in combination with sentiment analysis technology, and constructing a multi-dimensional demand feature vector;

[0020] Obtaining a traditional Chinese medicine massage knowledge graph; wherein the entities corresponding to acupoints, meridians, diseases, massage methods, and intensity in the traditional Chinese medicine massage knowledge graph are connected through association relationships to form a semantic network;

[0021] Aligning the multi-dimensional demand feature vector with the traditional Chinese medicine massage knowledge graph semantically, calculating the semantic similarity between the massage requirement and each acupoint, and screening out candidate acupoints with a similarity greater than a threshold value;

[0022] Obtaining a pre-constructed decision tree forest, the decision tree forest including multi-dimensional decision trees;

[0023] inputting the candidate acupoints into the decision tree forest for parallel reasoning to obtain initial parameter suggestions of each dimension;

[0024] conflict resolution is performed on the initial parameter suggestions of each dimension to construct a parameter compatibility matrix; based on the parameter compatibility matrix, a synergistic effect score of different parameter combinations is calculated, and a parameter combination with the highest synergistic effect score is selected as an initial massage strategy.

[0025] Further, the multi-dimensional decision tree includes a method decision tree, a force decision tree, and a time length decision tree.

[0026] The method decision tree generates decision rules based on the meridian properties and demand types of the candidate acupoints;

[0027] The force decision tree fuses user constitution data and acupoint sensitivity coefficients and optimizes decision boundaries using a random forest algorithm;

[0028] The time length decision tree determines the optimal massage time through a gradient boosting algorithm according to the demand intensity level and acupoint tolerance model.

[0029] Further, after selecting the parameter combination with the highest synergistic effect score as the initial massage strategy, the following steps are included:

[0030] Based on the preset traditional Chinese medicine knowledge base, reverse reasoning is performed on the initial massage strategy to check whether there are meridian compatibility taboos or method conflicts; if there are, a backtracking correction mechanism is started to adjust the conflicting parameters until the verification is passed.

[0031] Further, based on the optimization strategy, the initial massage strategy is optimized to obtain a personalized massage strategy, including:

[0032] The acupoint massage order in the initial massage strategy is analyzed and compared with the acupoint group massage priority in the optimization strategy, and a matching degree score of the acupoint massage order and the acupoint group massage priority is calculated; if the matching degree score is lower than a set threshold, the acupoint massage order is adjusted to obtain an adjusted acupoint massage order;

[0033] According to the time allocation proportion of each acupoint group in the optimization strategy, combined with real-time physiological data of the user, based on an adaptive weighted fusion algorithm, the massage time of each acupoint in the initial massage strategy is dynamically adjusted;

[0034] Based on the force adjustment coefficient in the optimization strategy, the force parameter in the initial massage strategy is calibrated to obtain a calibrated massage force;

[0035] The adjusted acupoint massage order, the adjusted massage time of each acupoint, and the calibrated massage force are integrated to form an optimized massage parameter combination;

[0036] The optimized massage parameter combination is compared and verified again with the initial massage strategy, if there is a parameter conflict, a constraint satisfaction algorithm is used to resolve the conflict based on the traditional Chinese massage theory and equipment performance constraints, and finally a personalized massage strategy is obtained.

[0037] The application also provides a massage device based on intelligent AI recognition of back acupoints, comprising:

[0038] An extraction module is configured to collect a back image of a user, perform feature extraction on the back image, obtain back muscle texture and skeletal contour features in the image, highlight the area features of the existing acupoints through an attention mechanism, and obtain an acupoint candidate feature vector.

[0039] A generation module is configured to input the acupoint candidate feature vector into a trained acupoint positioning model to obtain an acupoint positioning result, and generate an optimization strategy based on the acupoint positioning result.

[0040] An optimization module is configured to obtain a massage requirement input by a user, match a corresponding initial massage strategy based on the massage requirement by using a decision tree algorithm, optimize the initial massage strategy based on the optimization strategy, and obtain a personalized massage strategy.

[0041] A control module is configured to generate a control instruction of a massage device based on the personalized massage strategy to control the massage device to complete back acupoint massage.

[0042] The application also provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method according to any one of the preceding embodiments when executing the computer program.

[0043] The application also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the method according to any one of the preceding embodiments.

[0044] This invention provides a control method for a massage device based on intelligent AI-based recognition of back acupoints, comprising: acquiring a user's back image; extracting features from the back image to obtain muscle texture and skeletal contour features; highlighting the regional features containing acupoints through an attention mechanism to obtain acupoint candidate feature vectors; inputting the acupoint candidate feature vectors into a trained acupoint localization model to obtain acupoint localization results; generating an optimization strategy based on the acupoint localization results; acquiring the user's input massage needs; matching a corresponding initial massage strategy based on the massage needs using a decision tree algorithm; optimizing the initial massage strategy based on the optimization strategy to obtain a personalized massage strategy; and generating control commands for the massage device based on the personalized massage strategy to control the massage device to perform back acupoint massage. In this invention, by highlighting the regional features containing acupoints in the user's back image to obtain acupoint candidate feature vectors and generating targeted optimization strategies; combining the user's input massage needs to generate a corresponding initial massage strategy; and then optimizing the initial massage strategy based on the optimization strategy to obtain a personalized massage strategy, this overcomes the current limitation of not being able to quickly generate targeted massage strategies based on user differences. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the steps of a massage device control method based on intelligent AI to identify acupoints on the back in one embodiment of the present invention;

[0046] Figure 2 This is a structural block diagram of a massage device based on intelligent AI to identify acupoints on the back, according to one embodiment of the present invention.

[0047] Figure 3 This is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0048] The implementation, functional features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0050] Reference Figure 1 One embodiment of the present invention provides a control method for a massage device based on intelligent AI recognition of acupoints on the back, including the following steps:

[0051] Step S1, collect the back image of the user, perform feature extraction on the back image, obtain the back muscle texture, skeleton contour features in the image, highlight the region features of the existing acupoints through the attention mechanism, and obtain an acupoint candidate feature vector;

[0052] Step S2, input the acupoint candidate feature vector into a trained acupoint positioning model to obtain an acupoint positioning result; and generate an optimization strategy based on the acupoint positioning result;

[0053] Step S3, obtain a massage demand input by the user, match a corresponding initial massage strategy by using a decision tree algorithm based on the massage demand, and optimize the initial massage strategy based on the optimization strategy to obtain a personalized massage strategy;

[0054] Step S4, generate a control instruction of a massage device based on the personalized massage strategy to control the massage device to complete back acupoint massage.

[0055] In this embodiment, as described in step S1, the collected back image is first preprocessed to eliminate light difference and noise interference. Subsequently, a deep separable convolution network (such as MobileNetV3) is used to extract a multi-scale feature map, and shallow texture details and deep semantic information are obtained. To enhance the expression of acupoint region features, a CBAM attention mechanism is introduced. The mechanism adaptively enhances the feature response of the acupoint candidate region through two sub-modules of channel attention and spatial attention, and suppresses irrelevant background information. Specifically, the channel attention module generates channel descriptors through global average pooling and maximum pooling, and generates channel weights through a multi-layer perception. The spatial attention module generates spatial weights through convolution operation based on the channel weighted feature map, and finally realizes adaptive enhancement of the feature map. The acupoint candidate feature vector output in this step encodes spatial position information and semantic association information in the form of a three-dimensional tensor, providing a basis for subsequent accurate positioning.

[0056] As described in step S2 above, the acupoint candidate feature vector is input into the spatio-temporal attention LSTM network for acupoint coordinate regression. This network structure contains a spatio-temporal attention mechanism, which can capture the dynamic change characteristics of acupoints in continuous frame images, effectively dealing with the displacement of acupoint position caused by factors such as respiratory movement and body position adjustment. After the model outputs the acupoint coordinate matrix with confidence, a graph neural network (GNN) is used to construct an acupoint correlation graph. Each acupoint is a graph node, and the weight of the edge between nodes is predefined based on traditional Chinese medicine meridian theory and anatomical knowledge, and is dynamically adjusted in combination with real-time positioning results. Through graph convolution operation, the importance score and cooperative relationship of each acupoint are calculated, and an optimization strategy containing massage sequence priority, force distribution coefficient, and time proportion is generated. This strategy not only considers the meridian correlation between acupoints, but also combines the accuracy of real-time positioning, providing an optimization basis for personalized massage.

[0057] As described in step S3 above, the user's massage requirements are semantically analyzed, the semantic vector of the requirement text is extracted, and the requirement intensity is quantified using sentiment analysis technology. Subsequently, the semantic vector is input into the decision tree forest for multi-dimensional reasoning: the manipulation decision tree selects the optimal massage manipulation (such as rubbing, pressing, and pushing) based on the requirement type and acupoint characteristics; the intensity decision tree determines the initial intensity in combination with the user's body mass index (BMI) and muscle hardness data; the duration decision tree allocates massage time for each acupoint according to the demand urgency. To further optimize the initial strategy, a multi-objective optimization algorithm is used, with the maximum efficacy and maximum comfort as the objective functions, and the device physical limitations and human tolerance as the constraint conditions, to perform Pareto optimization on the massage parameters. By introducing a parameter synergy matrix, the interactive effects of different parameter combinations are evaluated, and finally a set of Pareto optimal solutions is generated. The solution with the highest satisfaction is selected as the personalized massage strategy to ensure the best balance among multiple objectives.

[0058] As described in step S4 above, based on the personalized massage strategy, a model predictive control (MPC) algorithm is used to generate the motion trajectory of the massage device. The MPC algorithm predicts the system response in the future period of time by establishing a device dynamics model, and optimizes the control input in a rolling manner to ensure that the massage head accurately reaches the target acupoint. During execution, the pressure sensor collects real-time feedback data from the user's back, and the Kalman filter is used to fuse the predicted value and the measured value to dynamically adjust the massage intensity. In particular, a dual-mode control architecture is designed: for acupoint stimulation, PID precise position control is used; for meridian line dredging, impedance control algorithm is used to adjust the force adaptively according to the contact stiffness. If the user's muscle tension is detected to be abnormally high or the pressure feedback exceeds the safety threshold, a protection mechanism is automatically triggered to adjust the massage parameters and generate warning information. This closed-loop control system ensures that the massage process is accurate, effective, safe, and comfortable, realizing the intelligent transformation from strategy to execution.

[0059] In an embodiment, the personalized massage strategy includes parameters of massage methods, massage intensity, and massage duration.

[0060] In the present embodiment, the user massage demand semantic vector is parsed by the BERT model, and a decision tree forest is combined to intelligently match the most suitable operation mode from a method library including rubbing, pressing, pushing, and tapping, according to the demand type (such as relieving fatigue or improving sleep) and the acupoint characteristics (such as meridian properties and therapeutic effects). For example, for the demand of relieving headache at Fengchi acupoint, the point pressing method is preferentially selected to accurately stimulate the acupoint.

[0061] The intensity decision tree comprehensively considers user constitution data (BMI, muscle hardness), acupoint sensitivity coefficients, and real-time pressure feedback, and dynamically determines the intensity by using fuzzy control theory. For example, for a young user with tight muscles, the intensity is automatically increased when massaging the acupoints of the bladder meridian; when detecting that the muscle tension of the user is abnormal, the intensity is reduced in real time through the Kalman filtering algorithm, so as to ensure that the treatment effect is achieved and discomfort is avoided.

[0062] According to the demand urgency, the acupoint tolerance model, and the fatigue accumulation curve, a gradient boosting algorithm is used to allocate the massage time of each acupoint. For example, for acute neck and shoulder pain, the massage time of Tianzong acupoint is extended; combined with real-time heart rate variability (HRV) data of the user, the overall massage time is dynamically adjusted to ensure the treatment effect while improving comfort. The three work together to form the core parameter system of the personalized massage strategy, realizing precise and personalized massage experience.

[0063] In an embodiment, based on the acupoint positioning result, an optimization strategy is generated, including:

[0064] The acupoint positioning result is subjected to acupoint correlation analysis, and an acupoint neighborhood relationship graph is constructed to calculate the spatial distance, meridian correlation, and functional synergy between acupoints, so as to obtain an acupoint correlation matrix;

[0065] Based on the acupoint correlation matrix, a graph neural network algorithm is used for acupoint clustering analysis to identify functionally related acupoint groups, extract the core acupoints and edge acupoints of each acupoint group, and obtain an acupoint clustering result;

[0066] The acupoint clustering result is subjected to massage priority evaluation to obtain a priority ranking result;

[0067] Based on the priority ranking result, an optimization strategy including acupoint group massage order, time allocation proportion, and intensity adjustment coefficient is generated; wherein the intensity adjustment coefficient is dynamically adjusted according to the position of the acupoint in the cluster, and the intensity adjustment coefficient of the core acupoint is higher than that of the edge acupoint.

[0068] In this embodiment, first, multi-dimensional association analysis is performed on the acupoint positioning results to construct an acupoint neighborhood relationship graph. The acupoint neighborhood relationship graph takes acupoints as nodes, and the edge weights between nodes are determined by spatial distance, meridian correlation, and functional synergy. The spatial distance is calculated by the Euclidean distance, reflecting the proximity of acupoints in physical space; the meridian correlation is based on the theory of Chinese meridians, and acupoints on the same meridian are given higher weights; the functional synergy is determined by co-occurrence frequency analysis, for example, acupoints commonly used together for treating a particular disease have higher weights. The three-dimensional relationship graph is mapped to a two-dimensional acupoint association matrix through tensor decomposition technology, and the matrix elements represent the comprehensive association strength between acupoints. This matrix not only retains the spatial distribution information of acupoints, but also encodes the meridian and functional relationships in traditional Chinese medicine theory, providing a structured data basis for subsequent clustering analysis.

[0069] Further, graph neural network (GNN) is used to perform clustering analysis on the acupoint association matrix. GNN aggregates the neighborhood information of each acupoint through multi-layer graph convolution operations, and learns the latent semantic representation of acupoints. The embedding vectors output by GNN are clustered using a spectral clustering algorithm to identify functionally related acupoint groups. The core acupoints within each cluster are determined by calculating node centrality (such as PageRank value), which usually have higher connectivity and intermediary centrality and play a key role in the meridian network. The edge acupoints are determined by the K-nearest neighbor algorithm, which have similar characteristics to core acupoints but lower association strength. The clustering results are represented in a hierarchical structure, each acupoint group contains explicit core-edge relationships, providing classification basis for priority evaluation.

[0070] Further, based on the clustering results, massage priority is evaluated from three dimensions of clinical demand, physiological effect, and safety threshold. The clinical demand dimension scores the matching degree of user health data (such as pain location, muscle tension) and acupoint treatment function; the physiological effect dimension simulates the influence of acupoint stimulation on blood circulation and nerve conduction through biomechanical models; the safety threshold dimension considers the sensitivity of acupoints and user physical limitations (such as force limitation of specific acupoints for hypertension patients). A multi-criteria decision-making model is constructed using the analytic hierarchy process (AHP) to calculate the comprehensive priority score of each acupoint group. The final priority sequence is generated by the Pareto optimal sorting algorithm to ensure that high-priority acupoint groups can meet the main needs and comply with physiological safety constraints.

[0071] Finally, based on the priority ranking results, a three-dimensional optimization strategy containing massage order, time allocation and intensity adjustment is generated. The massage order follows the principle of core first and edge later, and main points first and auxiliary points later, and high priority point groups are preferentially stimulated to quickly activate the meridian system. The time allocation ratio is dynamically adjusted according to the importance score of the point group, and an exponential decay function is used to ensure that key point groups receive sufficient stimulation time. The intensity adjustment coefficient is determined based on the position of the point in the cluster, and the adjustment coefficient of the core point is set to 1.5-2 times that of the edge point. The coefficient is converted to the actual massage intensity through a nonlinear mapping function. The strategy also includes a dynamic adjustment mechanism that can automatically correct parameters based on real-time physiological feedback (such as heart rate and skin electrical response) during the massage process, forming a closed-loop optimization system.

[0072] In an embodiment, based on the massage demand, a decision tree algorithm is used to match the corresponding initial massage strategy, including:

[0073] The text corresponding to the massage demand is subjected to semantic feature extraction, and the intensity level of the demand is identified by combining sentiment analysis technology to construct a multi-dimensional demand feature vector;

[0074] A traditional Chinese massage knowledge graph is obtained; wherein in the traditional Chinese massage knowledge graph, entities corresponding to points, meridians, diseases, techniques and intensity are connected through association relationships to form a semantic network;

[0075] The multi-dimensional demand feature vector and the traditional Chinese massage knowledge graph are semantically aligned to calculate the semantic similarity of the massage demand and each point, and candidate points with a similarity greater than a threshold value are selected;

[0076] A pre-constructed decision tree forest is obtained, and the decision tree forest includes multi-dimensional decision trees;

[0077] The candidate points are input into the decision tree forest for parallel reasoning to obtain initial parameter suggestions for each dimension;

[0078] The initial parameter suggestions for each dimension are subjected to conflict resolution to construct a parameter compatibility matrix; based on the parameter compatibility matrix, the synergy effect score of different parameter combinations is calculated, and the parameter combination with the highest synergy effect score is selected as the initial massage strategy.

[0079] In this embodiment, massage demand texts input by users are first received, which are expressed in natural language, such as "hope to relieve severe shoulder and neck pain", "want to have a light back relaxation massage", etc. In order to convert these unstructured texts into computer-processable information, the system uses a pre-trained language model (such as BERT) to extract semantic features from the texts, and parses key semantic information from the texts, such as the body parts involved and the effects expected to be achieved. At the same time, sentiment analysis technology is used to analyze the words, tone, and other elements in the text to identify the intensity level of user demand, such as classifying the demand into different levels such as light, moderate, and severe. Finally, the extracted semantic features and demand intensity levels are integrated to construct a multi-dimensional demand feature vector, which comprehensively describes the user's massage demand and provides basic data for subsequent processing.

[0080] A pre-constructed traditional Chinese massage knowledge graph is obtained, which is a structured semantic network. In this network, acupoints, meridians, diseases, massage techniques, and force levels are regarded as independent entities. These entities are connected to each other through various association relationships, such as the "treatment" relationship between acupoints and diseases, indicating that a specific acupoint has a therapeutic effect on certain diseases; the "application" relationship between acupoints and massage techniques, indicating that different massage techniques are suitable for different acupoints; and the "belonging" relationship between acupoints and meridians, reflecting the position of acupoints in the meridian system. In this way, the professional knowledge in the field of traditional Chinese massage is presented in a graphical and structured form, facilitating knowledge retrieval and reasoning.

[0081] Furthermore, the multi-dimensional demand feature vector constructed previously is subjected to semantic alignment operation with the traditional Chinese massage knowledge graph. This process aims to match the user's massage demand with the acupoints and other entities in the knowledge graph, and find the acupoints related to the demand. Specifically, the similarity between the multi-dimensional demand feature vector and the semantic vector corresponding to each acupoint in the knowledge graph is calculated to measure the association degree between the massage demand and each acupoint. The similarity calculation can use algorithms such as cosine similarity. After the calculation is completed, a threshold is set, and the acupoints with a similarity greater than the threshold are selected as candidate acupoints that meet the user's massage demand. These candidate acupoints are an important basis for generating massage strategies in the subsequent process, and they can meet the user's massage goals to a certain extent.

[0082] A pre-constructed decision tree forest is obtained, which is composed of multiple decision trees of different dimensions. Each decision tree focuses on a specific dimension of the massage strategy, for example, a massage technique decision tree is used to determine which massage technique should be used for different acupoints and needs; a massage intensity decision tree is used to determine the appropriate massage intensity; and a massage duration decision tree is used to determine the massage time for each acupoint. These decision trees are constructed based on a large amount of traditional Chinese medicine massage knowledge, clinical experience and experimental data, and through learning and analysis of the data, corresponding decision rules and logic are formed, which can provide accurate guidance for the generation of massage strategies.

[0083] Then, the selected candidate acupoints are respectively input into each decision tree in the decision tree forest for parallel reasoning. Each decision tree reasons and judges the parameters of the corresponding dimension according to its own decision rules and the input candidate acupoint information. For example, the massage technique decision tree reasons out the appropriate massage technique for the acupoint according to the characteristics of the acupoint, user needs and associated information in the knowledge graph; the massage intensity decision tree gives appropriate massage intensity suggestions in combination with the sensitivity of the acupoint, the user's physical condition and other factors; and the massage duration decision tree determines the massage time according to the efficacy of the acupoint and the intensity of the need. Through this parallel reasoning method, initial parameter suggestions for massage techniques, intensity, duration and other dimensions are quickly obtained, which provide specific parameter basis for constructing the initial massage strategy.

[0084] Since the initial parameter suggestions of each dimension are respectively reasoned by different decision trees, there may be conflicts or incompatibilities between the parameters, for example, a certain massage technique combined with a specific massage intensity and duration may cause discomfort to the user or fail to achieve the expected effect. Therefore, conflict resolution is needed for these initial parameter suggestions. The specific method is to construct a parameter compatibility matrix, which is used to describe the compatibility and mutual influence relationship between different parameter combinations. Then, according to the preset evaluation standard and algorithm, the parameter compatibility matrix is used to calculate the synergistic effect score of each parameter combination, which reflects the comprehensive performance of the parameter combination in meeting the user's massage needs, achieving the massage effect and ensuring the user's comfort, etc. Finally, the parameter combination with the highest synergistic effect score is selected from all parameter combinations, which is determined as the initial massage strategy. This initial massage strategy takes into account various factors and can theoretically maximize the satisfaction of the user's massage needs, providing guidance for subsequent massage operations.

[0085] In an embodiment, the multi-dimensional decision tree includes a technique decision tree, an intensity decision tree, and a duration decision tree;

[0086] The technique decision tree generates decision rules based on the meridian properties and demand types of the candidate acupoints;

[0087] The intensity decision tree fuses user constitution data and acupoint sensitivity coefficients, and adopts a random forest algorithm to optimize the decision boundary.

[0088] The duration decision tree determines the optimal massage time through a gradient boosting algorithm according to the demand intensity level and acupoint tolerance model.

[0089] In this embodiment, the construction of the manipulation decision tree is based on the theory of meridians in traditional Chinese medicine and clinical experience. First, the meridian properties of the candidate acupoints are analyzed to determine their corresponding meridian systems, such as the Foot-Taiyang Bladder Meridian and the Governor Vessel. Different meridians correspond to different blood and qi running and visceral associations. At the same time, the type of massage demand input by the user is considered, such as relieving pain, promoting blood circulation, and regulating visceral function. Through the learning of a large number of historical cases and the application relationship between acupoints and manipulation in traditional Chinese medicine classics, a set of decision rules is established. For example, when the candidate acupoint belongs to the Foot-Taiyang Bladder Meridian and the user's demand is to relieve muscle soreness, the decision tree will generate a decision rule that prefers the pressing and rubbing method and is supplemented by the tapping method according to the past data of this type of acupoint. These rules are stored in a tree structure, and each node represents a judgment condition, such as the type of meridian of the acupoint and the classification of demand intensity. The branches point to different manipulation options, thereby realizing the function of quickly matching appropriate massage manipulation according to the input information.

[0090] The intensity decision tree focuses on determining the precise massage intensity, and its core lies in comprehensively considering the individual differences of the user and the characteristics of the acupoint. First, the user's constitution data is obtained, including age, gender, body mass index (BMI), muscle content, and skin sensitivity, which can reflect the user's body's ability to withstand massage intensity. At the same time, the inherent sensitivity coefficient of each acupoint is combined, which is obtained through a large number of acupoint stimulation experiments and clinical feedback data statistics, representing the reaction degree of the acupoint to different intensity stimulation. In order to more accurately divide the reasonable intensity interval under different conditions, the random forest algorithm is adopted. This algorithm constructs a forest composed of multiple decision trees, randomly samples and splits the user's constitution data and acupoint sensitivity coefficients, and learns the rules and patterns in the data from different angles. Each decision tree makes decisions independently, and finally determines the optimal decision boundary through voting or averaging. For example, for an elderly user with a high acupoint sensitivity coefficient, the random forest algorithm optimizes the decision boundary of gentle intensity by analyzing numerous similar sample data, ensuring that the massage intensity can achieve therapeutic effect without causing damage to the user.

[0091] The duration decision tree aims to scientifically plan the massage duration of each acupoint. First, according to the intensity level of the user's massage needs, such as mild fatigue, moderate soreness, severe discomfort, etc., combined with the analysis of the tolerance model of different acupoints. The acupoint tolerance model is established based on traditional Chinese medicine theory and clinical practice data, which describes the physiological response and treatment effect change rule of acupoints under different duration stimulation. In order to find the optimal massage time from the complex relationship between demand and acupoint characteristics, the system uses gradient boosting algorithm. This algorithm constructs multiple weak prediction models in an iterative manner, each iteration is based on the error of the last model to adjust and focus on the data points that are not accurate in prediction, gradually improving the prediction accuracy of the model. In the duration decision tree, the gradient boosting algorithm determines the optimal massage duration of each acupoint under the premise of meeting the user's needs by continuously adjusting and optimizing the prediction model based on the input of demand intensity level and acupoint tolerance model. For example, for the case of severe shoulder and neck soreness demand and high acupoint tolerance, the gradient boosting algorithm optimizes through multiple iterations to make a decision to appropriately extend the massage time to enhance the treatment effect, while avoiding user discomfort or excessive stimulation of acupoints due to too long time.

[0092] In an embodiment, after selecting the parameter combination with the highest synergistic effect score as the initial massage strategy, the following steps are included:

[0093] Based on the preset traditional Chinese medicine knowledge base, the initial massage strategy is reverse inferred to check whether there are meridian compatibility taboos or manipulation conflicts, and if there are, the backtracking correction mechanism is started to adjust the conflict parameters until it passes the verification.

[0094] In this embodiment, after determining the parameter combination with the highest synergistic effect score as the initial massage strategy, further verification according to traditional Chinese medicine theory is needed to ensure the safety and effectiveness of the strategy.

[0095] Firstly, the preset traditional Chinese medicine knowledge base is called, which integrates classical theories of traditional Chinese medicine, clinical experience and authoritative medical literature, covering core knowledge such as meridian compatibility taboos, acupoint function characteristics, massage manipulation application range, etc. Then, the parameters such as acupoint selection, massage manipulation, force and duration in the initial massage strategy are substituted into the traditional Chinese medicine knowledge base for reverse reasoning.

[0096] During the reverse reasoning process, it is strictly checked whether there are meridian compatibility taboos in the acupoint combination. For example, according to traditional Chinese medicine theory, some acupoints may have antagonistic effects when massaged at the same time, or there may be potential risks for people with certain constitutions. The system needs to identify such conflict combinations. At the same time, it also verifies whether there are compatibility problems between massage manipulation and acupoint or meridian, such as some acupoints that are not suitable for certain manipulations due to their special location or sensitive function. If there is such a conflict in the initial strategy, it is determined as a manipulation conflict.

[0097] Once the meridian compatibility taboo or manipulation conflict is detected, the backtracking correction mechanism is immediately started. This mechanism is guided by the conflict point, traces back to the decision tree reasoning process and parameter generation link, and adjusts the conflict parameters according to the rules and suggestions in the traditional Chinese medicine knowledge base. For example, if it is found that there is a meridian compatibility taboo in a certain acupoint combination, the candidate acupoints will be re-evaluated, and alternative acupoints with similar efficacy and no compatibility conflict will be selected according to the relevant knowledge; if there is a manipulation conflict, the more suitable massage manipulation will be selected from the manipulation library according to the acupoint characteristics and meridian direction. After adjustment, the modified parameter combination is again substituted into the traditional Chinese medicine knowledge base for verification, and the above reasoning and checking process is repeated until the strategy fully meets the requirements of traditional Chinese medicine theory and there is no meridian compatibility taboo or manipulation conflict, and finally a safe and effective initial massage strategy is formed.

[0098] In an embodiment, the initial massage strategy is optimized based on the optimization strategy to obtain a personalized massage strategy, including:

[0099] The acupoint massage order in the initial massage strategy is analyzed and compared with the acupoint group massage priority in the optimization strategy, and the matching degree score of the acupoint massage order and the acupoint group massage priority is calculated; if the matching degree score is lower than the set threshold, the acupoint massage order is adjusted to obtain an adjusted acupoint massage order;

[0100] According to the time allocation proportion of each acupoint group in the optimization strategy, combined with the real-time physiological data of the user, the massage time of each acupoint in the initial massage strategy is dynamically adjusted based on the adaptive weighted fusion algorithm;

[0101] The force parameter in the initial massage strategy is calibrated based on the force adjustment coefficient in the optimization strategy to obtain a calibrated massage force;

[0102] The adjusted acupoint massage order, the adjusted massage time of each acupoint, and the calibrated massage force are integrated to form an optimized massage parameter combination;

[0103] The optimized massage parameter combination is compared and verified with the initial massage strategy again, if there is a parameter conflict, the constraint satisfaction algorithm is used to resolve the conflict based on the traditional Chinese massage theory and the equipment performance constraint, and finally a personalized massage strategy is obtained.

[0104] In this embodiment, first, the preset acupoint massage sequence in the initial massage strategy is deeply analyzed, and it is compared with the sorting result based on the acupoint group massage priority in the optimization strategy item by item. In the comparison process, by constructing a matching degree calculation model, considering the factors such as the meridian to which the acupoint belongs, the functional characteristics, and the position difference in the priority sequence, the matching degree score of the acupoint massage sequence and the acupoint group massage priority is calculated. If the score is lower than the pre-set threshold, it indicates that the current acupoint massage sequence fails to fully reflect the priority logic emphasized by the optimization strategy, which may affect the massage effect. At this time, the simulated annealing algorithm is started, which simulates the physical annealing process, exchanges the acupoint massage sequence randomly, evaluates the matching degree score of the new sequence, and accepts the worse solution with a certain probability, gradually searches for a better acupoint massage sequence in the solution space, and finally obtains the adjusted acupoint massage sequence after multiple iterations, to ensure that the massage process conforms to the principles of meridian conduction and acupoint synergistic effect in traditional Chinese medicine.

[0105] Further, according to the time allocation ratio set for each acupoint group in the optimization strategy, the system combines the real-time physiological data collected by the user, such as heart rate, electromyographic signal, skin temperature, etc. These data can reflect the user's body response and fatigue state to the massage in real time. In order to realize accurate time adjustment, the system uses an adaptive weighted fusion algorithm, which dynamically assigns appropriate weights to each index according to the influence of different physiological indicators on the massage effect, and evaluates the user's current physical condition through weighted calculation, and then intelligently adjusts the massage time of each acupoint in the initial massage strategy. For example, if the electromyographic signal of a certain part of the user shows a high degree of fatigue, the massage time of the corresponding acupoint group is automatically increased; on the contrary, if the physiological indicators show that the part has reached the expected relaxation effect, the massage time is appropriately shortened, ensuring that the massage time can meet the treatment needs and avoid excessive stimulation.

[0106] Further, the force adjustment coefficient set for different acupoints and acupoint groups in the optimization strategy is used as the reference to calibrate the force parameter in the initial massage strategy. First, compare and analyze the force parameter output by the current massage device with the force adjustment coefficient, according to the difference between the two, combine the real-time feedback of the user's pressure perception data, and adjust the massage force in real time through the PID (Proportional-Integral-Derivative) control algorithm. The PID algorithm can dynamically calculate the appropriate adjustment amount according to the error size, error change rate, and error accumulation, to ensure that the massage force remains stable while accurately meeting the requirements of the optimization strategy. For example, if the force adjustment coefficient indicates that the force needs to be increased, and the user has not feedback discomfort, the massage force will be gradually increased according to the PID algorithm until the ideal value after calibration is reached.

[0107] After completing the adjustment of the acupoint massage sequence, the dynamic optimization of the massage time length, and the calibration of the massage intensity, the adjusted acupoint massage sequence, the updated massage time length of each acupoint, and the calibrated massage intensity are integrated. By constructing a standardized data structure, these parameters are combined in order to form a complete set of optimized massage parameter combinations. This combination optimizes the initial massage strategy from three dimensions of massage process, time allocation, and intensity control, providing more scientific and user-demand-oriented parameter guidance for subsequent massage operations.

[0108] Finally, the optimized massage parameter combination is compared and verified again with the initial massage strategy. By establishing a parameter conflict detection model, it checks whether there are problems such as mismatch between acupoint sequence and time length allocation, conflict between intensity and acupoint characteristics, or parameters exceeding the performance range of the massage device, etc. in the new parameter combination. If parameter conflicts are detected, a constraint satisfaction algorithm is immediately started. This algorithm is guided by traditional Chinese massage theory and combined with the physical performance constraints of the massage device, such as maximum massage intensity, minimum massage time, etc. Through searching and adjusting the parameter value space, it finds the optimal solution that satisfies all constraint conditions. After repeated iteration and verification, all parameter conflicts are eliminated, and a personalized massage strategy that meets the requirements of traditional Chinese medicine theory, device operation, and individual differences of users is finally generated, providing users with a safe, efficient, and comfortable massage experience.

[0109] In an embodiment, after obtaining the personalized massage strategy, the following steps are included:

[0110] Obtain the user's biometric information and identity information, and perform a hash operation to generate a user unique key seed;

[0111] Obtain a multi-dimensional chaotic mapping encryption model, and decompose the parameter matrix of the personalized massage strategy into basic data blocks;

[0112] Use the user unique key seed to drive the multi-dimensional chaotic mapping encryption model to generate a chaotic sequence;

[0113] Perform feature confusion on the chaotic sequence and the basic data blocks to generate encrypted data segments;

[0114] Store the encrypted data segments in multiple nodes, and each node stores data with an access permission verification contract based on zero-knowledge proof.

[0115] In this embodiment, the biometric information of the user is first collected, including but not limited to fingerprint feature vector, iris texture data, facial recognition feature code, etc., which have high individual uniqueness and stability; at the same time, the identity information of the user is obtained, such as ID number, registered account ID, etc. Then, the biometric information and the identity information are fused and processed, and the fused information is subjected to a hash operation. The hash operation converts the input data into a fixed-length hash value, and this process is one-way, and the original data cannot be deduced from the hash value. The generated hash value is used as a unique key seed of the user, and its uniqueness is guaranteed by the individual differences of the biometric information and the specificity of the identity information, which provides a core key basis for the subsequent encryption process.

[0116] Further, a pre-constructed multi-dimensional chaotic mapping encryption model is called, which is based on chaos theory and uses the randomness, ergodicity and initial value sensitivity of chaotic systems to achieve encryption function. At the same time, the parameters contained in the personalized massage strategy, such as acupoint massage sequence, massage duration, massage intensity, etc., are structured in matrix form. In order to facilitate encryption processing, a block algorithm is used to decompose the parameter matrix into multiple basic data blocks according to a preset rule, each data block containing part of the parameter information. This decomposition method can not only ensure the efficiency of data processing, but also provide suitable data units for subsequent hybrid encryption with chaotic sequences.

[0117] The generated user unique key seed is used as the initial input parameter of the multi-dimensional chaotic mapping encryption model. The chaotic model generates a chaotic sequence with pseudo-random characteristics through multi-dimensional chaotic mapping iterative operations such as Logistic mapping and Henon mapping under the driving of the key seed. Since the chaotic system is extremely sensitive to initial conditions, even if there is a slight difference in the key seed, the generated chaotic sequence will be completely different. Therefore, the chaotic sequence generated based on the user-specific key seed has high uniqueness and unpredictability, providing strong encryption key resources for the encryption process.

[0118] The generated chaotic sequence is subjected to feature confusion operation with the basic data blocks obtained by decomposition bit by bit or element by element. Specifically, XOR operation, modulo operation, etc. can be used to integrate the pseudo-random characteristics of the chaotic sequence into the basic data blocks, changing the feature distribution of the original data. After feature confusion, the original basic data blocks are transformed into encrypted data segments, which no longer have the semantic characteristics of the original data. Even if the data segments are illegally obtained, without the corresponding key seed and encryption model, the original personalized massage strategy parameters cannot be restored, thereby achieving effective encryption protection of the data.

[0119] To further enhance the security of data storage, a distributed storage architecture is adopted to disperse encrypted data segments in multiple independent storage nodes. This dispersed storage method reduces the risk of data leakage due to single node failure or attack. When storing encrypted data segments in each node, the system attaches an access permission verification contract based on zero-knowledge proof. Zero-knowledge proof allows users to prove their access rights to the data without revealing the specific data content to the storage node. Only when the user can pass the zero-knowledge proof verification, the storage node allows it to access the corresponding data segment, thus ensuring data privacy while achieving secure and controllable data access management.

[0120] In an embodiment, after obtaining the personalized massage strategy, the following steps are included:

[0121] Hashing the user's biometric data and the registered mobile phone number generates a user unique identification hash value;

[0122] A dynamic coding mapping table is constructed, which is designed based on the integration of the five-element theory of traditional Chinese medicine and the Morse code rules. Based on the dynamic coding mapping table, the parameters in the personalized massage strategy are converted into five-element symbol combinations and Morse code sequences;

[0123] A matrix permutation coding algorithm is used, with the user unique identification hash value as the random seed, to perform matrix row and column permutation on the five-element symbol combinations and Morse code sequences to form a confusion coding matrix;

[0124] The confusion coding matrix is XORed with the user unique identification hash value to generate encrypted coding data;

[0125] A relational storage structure is established in the database, with the user unique identification hash value as the index primary key, and the encrypted coding data is stored in association.

[0126] In this embodiment, first, the user's biometric data such as fingerprint feature vector, facial contour geometric parameters, and other individual unique physiological information are collected, and the user's registered mobile phone number is obtained. After information fusion of the biometric data and the registered mobile phone number, the fused data is calculated. Hashing operation maps the input data to a fixed-length hash value through complex mathematical transformation, and this process is irreversible and has an avalanche effect, i.e. a small change in the original data will result in a significant difference in the hash value. The generated user unique identification hash value not only has uniqueness, but also effectively hides the original information, providing a secure and reliable identity authentication basis for subsequent data encryption and storage.

[0127] Further, according to the element properties and mutual relations in the theory of five elements (gold, wood, water, fire, and earth) in traditional Chinese medicine, combined with the dot and dash coding rules of Morse code, a dynamic coding mapping table is constructed. The mapping table maps the acupoint names, massage methods, intensity levels, time length parameters, etc. in the personalized massage strategy to specific five-element symbols and Morse code combinations, respectively. For example, the "rubbing method" in the massage method is mapped to the five-element symbol representing the "wood" property and assigned a specific Morse code sequence; the intensity level is converted into different symbol combinations and code sequences through the principle of five-element generation and restriction. This design that combines traditional Chinese medicine theory and modern coding rules not only gives the coding semantic connotation, but also increases the complexity and uniqueness of the coding, so that the original parameters are converted into five-element symbol combinations and Morse code sequences with dual coding characteristics, effectively hiding the true meaning of the data.

[0128] Then, the five-element symbol combinations and Morse code sequences obtained in the previous step are digitally converted to construct a two-dimensional data matrix. The user's unique identification hash value generated in the first step is used as the random seed of the matrix permutation coding algorithm, which determines the specific rules and order of matrix row and column permutation. The matrix permutation coding algorithm rearranges the rows and columns of the matrix randomly, and performs cyclic shift and other operations to disrupt the arrangement order and structural characteristics of the original data. Due to the uniqueness of the hash value, the data matrix permutation method for each user is different, which completely confuses the original coding sequence and forms a highly chaotic confusion coding matrix, further enhancing the confidentiality and anti-cracking ability of the data.

[0129] Further, each element in the confusion coding matrix is subjected to bitwise XOR operation with the user's unique identification hash value. The XOR operation has the property that the same data XOR result is 0, and different data XOR result is 1. Through this operation, the data characteristics in the confusion coding matrix are changed again, and the uniqueness of the user's identity information is deeply integrated into the encryption process. After the XOR operation, the final encrypted coding data is generated, which contains not only the parameter information of the original personalized massage strategy, but also is highly encrypted through multiple coding and operations. Only the authorized party with the correct user's unique identification hash value can restore the original data through reverse operation, effectively ensuring the security of the data in the storage and transmission process.

[0130] Finally, to achieve efficient management and fast retrieval of encrypted data, a relational storage structure is constructed in the database. The unique identification hash value of the user is used as the index primary key to establish the association between it and the encrypted coded data. The uniqueness of the index primary key ensures that each user's encrypted data has a unique identifier in the database, facilitating fast positioning and querying. Through relational storage, user data can be managed in a standardized manner. Meanwhile, the database's permission control mechanism is used to further restrict access to encrypted data, ensuring that only authorized operations can access and process data, thereby providing a secure and reliable storage environment for personalized massage strategy data at the storage level.

[0131] In an embodiment, after obtaining the personalized massage strategy, the following steps are included:

[0132] Obtain the identity information of the user and a preset mapping table. The preset mapping table includes an original data column and a mapping data column. The original data column contains commonly used words and numbers in the massage strategy. The mapping data column contains sequentially arranged English alphabet characters.

[0133] Add the English alphabet characters in the mapping data column to the data table template sequentially one by one to obtain a character data table.

[0134] Extract multiple key numbers from the identity information and generate simulated curves based on each key number. Extract feature information from the identity information and generate corresponding graphics.

[0135] According to the rules, superimpose the graphics onto the character data table. Divide the character data table into multiple sub-regions by the edges of the graphics, and sort each sub-region.

[0136] Resort the characters in each sub-region based on the simulated curves to obtain sub-region character sequences. Combine the sub-region character sequences in order according to the sorting of the sub-regions to obtain a combined character sequence.

[0137] Replace the characters in the combined character sequence into the mapping data column of the preset mapping table sequentially one by one to obtain a new mapping table. Map the personalized massage strategy based on the new mapping table to obtain mapping data and store it to the management terminal.

[0138] In this embodiment, the user's identity information is first collected, which includes unique identification data such as ID number, registered account code, etc. At the same time, a pre-set mapping table is called, which contains two columns of data structure: the original data column records the high-frequency Chinese characters (such as "jūn", "àn", "xué") in the personalized massage strategy, Arabic numerals (used to represent the intensity level, massage duration) and other original data elements; the mapping data column arranges a plurality of English alphabet characters (which can be uppercase or lowercase) in a fixed order to form the original character mapping relationship.

[0139] Further, a blank data table template is called, and based on the English alphabet order of the mapping data column in the pre-set mapping table, the characters are sequentially filled into the cells of the data table template. For example, starting from the letter "A", the letters "B", "C", etc. are sequentially added one by one until they are all added in the data table template, thereby constructing a character data table. The character data table presents the complete sequence of English letters in a two-dimensional table form, providing a basic carrier for subsequent data confusion and encryption operations. At this time, the character data table has not yet been associated with the specific content of the personalized massage strategy, but it has the structural basis for complex transformation.

[0140] Further, the user's identity information is deeply analyzed, and representative key numbers are selected, such as specific digits in the ID number, part of the timestamp in the registration, etc. Through mathematical interpolation algorithms (such as cubic spline interpolation) or chaotic mapping functions, these key numbers are converted into smooth analog curves, the shape and trend of which are determined by the numerical sequence of the key numbers, and have user identity specificity. At the same time, feature information is extracted from the identity information, such as the profile features of the head portrait on the ID card, the geographic coordinates of the registration address, etc. Using computer graphics algorithms, these feature information is converted into corresponding geometric figures, such as polygons, irregular curve figures, etc. These figures and curves are generated based on the user's identity information and have uniqueness and non-replicability, introducing user-specific personalized elements for data encryption.

[0141] According to the preset superposition rule, the figure generated in the previous step is accurately overlaid on the character data table. Each edge of the figure intersects with the rows and columns of the character data table, forming a plurality of closed sub-areas, each containing part of the English alphabet characters. Subsequently, according to the topological structure of the figure, the area size or relative position relationship of the sub-areas, the sub-areas are sequentially numbered. For example, according to the order from left to right and from top to bottom, or based on the clockwise contour direction, the sub-areas are sequentially marked, thereby establishing the logical order relationship between the sub-areas. This division and sorting method breaks the original structure of the character data table, creating conditions for subsequent data rearrangement.

[0142] For each English letter character in each sub-region, rearrangement is performed in a curve simulation oriented manner. In a specific operation, the numerical change of the simulation curve can be mapped to the weight or index of the character sorting, for example, the peak of the curve corresponds to the forward movement of the character, and the trough corresponds to the backward movement of the character, so as to disturb the original order of the characters in the sub-region and generate a new sub-region character sequence. It can also be other ways, which will not be described one by one here.

[0143] Then, according to the determined sub-region order, the character sequences of the sub-regions are sequentially spliced to form a complete combined character sequence. The combined character sequence integrates the graph and curve features generated by the user identity information, so that the original English letter sequence is transformed in a complex and targeted manner, greatly increasing the difficulty of data cracking.

[0144] Finally, each character in the combined character sequence is replaced with the original English letter in the mapping data column in the preset mapping table in sequence, thereby generating a new mapping table with user-specific encryption characteristics. At this time, the correspondence between the original data column and the mapping data column in the new mapping table has been completely changed, forming a unique encryption mapping rule. Based on the new mapping table, all original data elements (common words, numbers, etc.) in the personalized massage strategy are mapped and converted element by element to be converted into encrypted mapping data. Finally, these mapping data are securely transmitted and stored in the database of the management terminal. Since the mapping rule is deeply dependent on the graph and curve generated by the user identity information, only authorized users who have the same identity information and conversion rule can restore the original massage strategy data through reverse operation to realize high-strength encryption storage and protection of data.

[0145] Referring to Figure 2 In another embodiment of the present application, a massage device based on intelligent AI recognition of back acupoints is also provided, comprising:

[0146] An extraction module is configured to collect a back image of a user, perform feature extraction on the back image, obtain back muscle texture and skeletal contour features in the image, highlight the area features of the acupoints through an attention mechanism, and obtain an acupoint candidate feature vector.

[0147] A generation module is configured to input the acupoint candidate feature vector into a trained acupoint positioning model to obtain an acupoint positioning result, and generate an optimization strategy based on the acupoint positioning result.

[0148] An optimization module is configured to obtain a massage requirement input by a user, match a corresponding initial massage strategy based on the massage requirement by using a decision tree algorithm, and optimize the initial massage strategy based on the optimization strategy to obtain a personalized massage strategy.

[0149] The control module is configured to generate a control instruction of the massage device based on the personalized massage strategy, so as to control the massage device to complete the back acupoint massage.

[0150] In the embodiment, the specific implementation of each module in the device embodiment is described above in the method embodiment, and will not be repeated here.

[0151] Referring to Figure 3 The computer device can be a server, and the internal structure thereof can be as shown in Figure 3 The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. The processor of the computer device is configured to provide calculation and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store corresponding data in the embodiment. The network interface of the computer device is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement the above method.

[0152] Those skilled in the art can understand Figure 3 The structure shown in the embodiment is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied.

[0153] The computer readable storage medium in the embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0154] In summary, the control method for the massage device based on intelligent AI recognition of back acupoints provided in the embodiments of the present application includes: collecting a back image of a user, performing feature extraction on the back image to obtain back muscle texture and skeletal contour features in the image, highlighting the region features of the existing acupoints through an attention mechanism to obtain an acupoint candidate feature vector; inputting the acupoint candidate feature vector into a trained acupoint positioning model to obtain an acupoint positioning result; generating an optimization strategy based on the acupoint positioning result; obtaining a massage requirement input by the user, matching a corresponding initial massage strategy based on the massage requirement by using a decision tree algorithm; optimizing the initial massage strategy based on the optimization strategy to obtain a personalized massage strategy; and generating a control instruction of the massage device based on the personalized massage strategy to control the massage device to complete back acupoint massage. In the present application, the region features of the existing acupoints are highlighted through the back image of the user to obtain an acupoint candidate feature vector, and an optimization strategy is generated accordingly; a corresponding initial massage strategy is generated in combination with the massage requirement input by the user, and then the initial massage strategy is optimized based on the optimization strategy to obtain a personalized massage strategy, thereby overcoming the defect that the current massage strategy cannot be quickly generated based on the differences of users.

[0155] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiments. Any reference to memory, storage, database or other medium provided by the present application and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM) and memory bus dynamic RAM, etc.

[0156] It is to be understood that the terminology "including", "comprising", or any other variation thereof, is intended to cover a non-exclusive inclusion such that process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises a... " does not, without more constraints, exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0157] The above description is merely the preferred embodiments of the present application, and is not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made according to the content of the present application specification and drawings, or directly or indirectly applied to other related technical fields, is also included in the patent protection scope of the present application.

Claims

1. A control method of a massage device based on intelligent AI recognition of back acupoints, characterized in that, The method comprises the following steps: Collecting a back image of a user, performing feature extraction on the back image to obtain back muscle texture and skeletal contour features in the image, highlighting the features of the region where the acupoint exists through an attention mechanism, and obtaining an acupoint candidate feature vector; Inputting the acupoint candidate feature vector into a trained acupoint positioning model to obtain an acupoint positioning result; and generating an optimization strategy based on the acupoint positioning result; Obtaining a massage requirement input by a user, and matching a corresponding initial massage strategy based on the massage requirement by using a decision tree algorithm; the method comprises the following steps: performing semantic feature extraction on a text corresponding to the massage requirement, identifying the intensity level of the requirement by combining a sentiment analysis technique, and constructing a multi-dimensional requirement feature vector; obtaining a traditional Chinese medicine massage knowledge graph; in the traditional Chinese medicine massage knowledge graph, entities corresponding to acupoints, meridians, diseases, massage techniques and massage intensity are connected through association relationships to form a semantic network; performing semantic alignment on the multi-dimensional requirement feature vector and the traditional Chinese medicine massage knowledge graph, calculating the semantic similarity between the massage requirement and each acupoint, and screening out candidate acupoints with a similarity greater than a threshold; obtaining a pre-constructed decision tree forest, the decision tree forest comprising multi-dimensional decision trees; inputting the candidate acupoints into the decision tree forest for parallel reasoning to obtain initial parameter suggestions in each dimension; resolving conflicts of the initial parameter suggestions in each dimension to construct a parameter compatibility matrix; based on the parameter compatibility matrix, calculating the synergistic effect scores of different parameter combinations, and selecting a parameter combination with the highest synergistic effect score as the initial massage strategy; optimizing the initial massage strategy based on the optimization strategy to obtain a personalized massage strategy; Generating a control instruction of a massage device based on the personalized massage strategy; Performing a hash operation on biological feature data of a user and a registered mobile phone number to generate a user unique identification hash value; constructing a dynamic coding mapping table, the dynamic coding mapping table being designed based on a combination of the five-element theory of traditional Chinese medicine and the Morse code rule; based on the dynamic coding mapping table, converting parameters in the personalized massage strategy into a five-element symbol combination and a Morse code sequence; using a matrix permutation coding algorithm, taking the user unique identification hash value as a random seed, and performing matrix row and column permutation on the five-element symbol combination and the Morse code sequence to form a confusion coding matrix; performing an exclusive OR operation on the confusion coding matrix and the user unique identification hash value to generate encrypted coding data; and establishing a relational storage structure in a database, taking the user unique identification hash value as an index primary key, and associatively storing the encrypted coding data. 2.The control method of the massage device based on the smart AI recognition of back acupoints according to claim 1, characterized in that, The personalized massage strategy comprises parameters of a massage technique, a massage intensity and a massage duration. 3.The control method of the massage device based on the smart AI recognition of back acupoints according to claim 1, characterized in that, Based on the acupoint positioning result, an optimization strategy is generated, comprising: Performing acupoint correlation analysis on the acupoint positioning result to construct an acupoint neighborhood relationship graph, calculate the spatial distance, meridian correlation and functional synergy between acupoints, and obtain an acupoint correlation matrix; Based on the acupoint correlation matrix, performing acupoint clustering analysis by using a graph neural network algorithm to identify functionally related acupoint groups, extract core acupoints and edge acupoints of each acupoint group, and obtain an acupoint clustering result; The acupoint clustering result is evaluated in terms of massage priority, and a priority ranking result is obtained; Based on the priority ranking result, an optimization strategy including acupoint group massage order, time allocation ratio and intensity adjustment coefficient is generated; wherein the intensity adjustment coefficient is dynamically adjusted according to the position of the acupoint in the cluster, and the intensity adjustment coefficient of the core acupoint is higher than that of the edge acupoint. 4.The control method of the massage device based on the smart AI recognition of back acupoints according to claim 1, characterized in that, The multi-dimensional decision tree includes a technique decision tree, an intensity decision tree, and a time length decision tree; The technique decision tree generates decision rules based on the meridian attribute and demand type of the candidate acupoint; The intensity decision tree integrates user constitution data and acupoint sensitivity coefficient, and uses a random forest algorithm to optimize the decision boundary; The time length decision tree determines the optimal massage time through gradient boosting algorithm according to the demand intensity level and acupoint tolerance model. 5.The control method of the massage device based on the smart AI recognition of back acupoints according to claim 1, characterized in that, After selecting the parameter combination with the highest synergistic effect score as the initial massage strategy, including: Based on the preset traditional Chinese medicine knowledge base, the initial massage strategy is reverse reasoned to check whether there is a meridian compatibility taboo or a technique conflict, and if there is, a backtracking correction mechanism is started to adjust the conflict parameters until it passes the verification. 6.The control method of the massage device based on the smart AI recognition of back acupoints according to claim 1, characterized in that, Based on the optimization strategy, the initial massage strategy is optimized to obtain a personalized massage strategy, including: The acupoint massage order in the initial massage strategy is analyzed, compared with the acupoint group massage priority in the optimization strategy, and the matching degree score of the acupoint massage order and the acupoint group massage priority is calculated; if the matching degree score is lower than the set threshold, the acupoint massage order is adjusted to obtain the adjusted acupoint massage order; According to the time allocation ratio of each acupoint group in the optimization strategy, combined with the user's real-time physiological data, the massage time of each acupoint in the initial massage strategy is dynamically adjusted based on the adaptive weighted fusion algorithm; Based on the intensity adjustment coefficient in the optimization strategy, the intensity parameter in the initial massage strategy is calibrated to obtain the calibrated massage intensity; The adjusted acupoint massage order, the massage time of each acupoint after adjustment, and the calibrated massage intensity are integrated to form the optimized massage parameter combination; The optimized massage parameter combination and the initial massage strategy are compared and verified again, if there is a parameter conflict, the constraint satisfaction algorithm is used to resolve the conflict based on the traditional Chinese massage theory and the equipment performance constraint, and finally the personalized massage strategy is obtained.

7. A massage device based on intelligent AI recognition of back acupoints, characterized by, Including: The extraction module is used to collect the back image of the user, extract the features of the back image, obtain the muscle texture and skeletal contour features of the back in the image, highlight the regional features of the existing acupoints through the attention mechanism, and obtain the acupoint candidate feature vector; The generation module is used to input the acupoint candidate feature vector into the trained acupoint positioning model to obtain an acupoint positioning result; and generate an optimization strategy based on the acupoint positioning result; An optimization module is configured to obtain a massage requirement input by a user, match an initial massage strategy corresponding to the massage requirement based on the massage requirement by using a decision tree algorithm, and include the following steps: performing semantic feature extraction on a text corresponding to the massage requirement, identifying an intensity level of the requirement by combining a sentiment analysis technique, and constructing a multi-dimensional requirement feature vector; obtaining a traditional Chinese medicine massage knowledge graph; in the traditional Chinese medicine massage knowledge graph, entities corresponding to acupoints, meridians, diseases, massage techniques and force are connected by association relationships to form a semantic network; performing semantic alignment on the multi-dimensional requirement feature vector and the traditional Chinese medicine massage knowledge graph, calculating semantic similarity between the massage requirement and each acupoint, and screening candidate acupoints with a similarity greater than a threshold; obtaining a pre-constructed decision tree forest, the decision tree forest including multi-dimensional decision trees; inputting the candidate acupoints into the decision tree forest for parallel reasoning to obtain initial parameter suggestions in each dimension; resolving conflicts of the initial parameter suggestions in each dimension to construct a parameter compatibility matrix; based on the parameter compatibility matrix, calculating a synergistic effect score of different parameter combinations, and selecting a parameter combination with the highest synergistic effect score as an initial massage strategy; optimizing the initial massage strategy based on the optimization strategy to obtain a personalized massage strategy; A control module is configured to generate a control instruction of a massage device based on the personalized massage strategy. Hashing operation is performed on biological feature data of a user and a registered mobile phone number to generate a user unique identification hash value; a dynamic coding mapping table is constructed, the dynamic coding mapping table being designed based on a combination of a traditional Chinese medicine five-element theory and Morse code rules; based on the dynamic coding mapping table, parameters in the personalized massage strategy are converted into five-element symbol combinations and Morse code sequences; a matrix permutation coding algorithm is used to take the user unique identification hash value as a random seed to perform matrix row and column permutation on the five-element symbol combinations and the Morse code sequences to form a confusion coding matrix; exclusive OR operation is performed on the confusion coding matrix and the user unique identification hash value to generate encrypted coding data; a relational storage structure is established in a database, the user unique identification hash value is taken as an index primary key, and the encrypted coding data is stored in association. 8.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-7. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.

Images