Graphene massager control method, system and medium with multi-dimensional somatosensory feedback

Abstract

The present invention discloses a control method, system and medium for a graphene massager with multi-dimensional somatosensory feedback. By obtaining multi-dimensional features of the current user's occupation information, work and rest information, diet information, massage position information and personal information, and combining the usage data of other users in the database with the first user information of the current user, a personalized initial massage mode is generated, improving the intelligent level of the device. During the massage process, the user's physiological feedback information, manual adjustment information and feedback evaluation information are collected to dynamically adjust the massage parameters, providing a massage experience that better meets the individual needs of the current user, and dynamically optimizing the second reference massage information to enhance the intelligent perception ability of the device. The feedback information and usage mode of each current user's massage process are recorded in the personal usage database, providing data support for the optimization of subsequent massage modes and enhancing the overall user experience.

Description

Technical Field

[0001] The present invention relates to the field of computers, and in particular, to a control method, system and medium for a graphene massager with multi-dimensional somatosensory feedback. Background Art

[0002] Graphene has high electrical conductivity, high strength and high transparency. Taking advantage of this, a massager made of graphene has an efficient heat conduction effect and can quickly transfer heat to the parts that need to be massaged. Graphene also has good flexibility, enabling the graphene massager to better fit the skin, thereby improving the massage touch. The heat conduction and electrical conductivity of graphene can better control the temperature according to the control system, enhancing the massage effect of the massager.

[0003] There are many existing graphene massage products, but most of them use built-in massage modes to massage users, unable to achieve deeper dynamic adjustment, that is, fine-tuning of massage parameters and storage of memories based on personal usage habits or feelings, and unable to provide a more excellent customized massage function. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a control method, system and medium for a graphene massager with multi-dimensional somatosensory feedback.

[0005] In order to achieve the above technical objectives, the technical solutions adopted by the present invention are as follows:

[0006] In the first aspect, the present invention provides a control method for a graphene massager with multi-dimensional somatosensory feedback, including:

[0007] Obtain first user information, where the first user information includes information characteristics of different information categories;

[0008] Generate first initial characteristic information according to the first user information, and use the cosine similarity function in the database to match second characteristic information, where the second characteristic information is generated when other users use a graphene massager of the same massage device category;

[0009] Obtain massage parameter information associated with the massage mode of the second characteristic information with successful matching, and generate first reference massage information;

[0010] Perform weighted averaging on the massage parameters of the same massage point category in multiple first reference massage information to generate second reference massage information;

[0011] Obtain the initial massage mode of the current user according to the second reference massage information and massage the current user;

[0012] Collect the feedback information of the user during the massage process, where the feedback information includes physiological feedback information, manual adjustment information, and feedback evaluation information;

[0013] Generate the first usage characteristic information according to the feedback information, and adjust the second reference massage information according to the first usage characteristic information to obtain the third reference massage information;

[0014] Generate a usage massage mode according to the third reference massage information, and replace the current initial massage mode during the next batch of massages;

[0015] Moreover, construct a personal usage database, and store the feedback information of each batch of massages in the personal usage database in chronological order.

[0016] In some embodiments, generating the first initial characteristic information according to the first user information includes:

[0017] Perform one-hot encoding on the information characteristics in the first user information that are not numerically represented, and represent the first user information by formula (1), where formula (1) is as follows:

[0018] ;

[0019] In formula (1), is the vector representation of the first user information, is the numerical representation of the information characteristic of the th information category in the first user information, ;

[0020] Perform standardization processing on the vector representation of the first user information to obtain the first initial characteristic information, which is represented by formulas (2) to (4), and formula (2) is as follows:

[0021] ;

[0022] Formula (3) is as follows:

[0023] ;

[0024] Formula (4) is as follows:

[0025] ;

[0026] In formulas (2) to (4), is the first initial characteristic information, is the average value of is the th information characteristic value, is the th information characteristic average value, is the standard deviation vector of, is the standard deviation vector of the information feature of the th;

[0027] The second feature information is represented by formula (5), and formula (5) is as follows:

[0028] ;

[0029] In formula (5), is the second feature information, is the numerical representation of the information feature of the th information category in the second user information;

[0030] In the database, use the cosine similarity function to match the second feature information similar to the first initial feature information, obtain the massage parameter information associated with the massage mode corresponding to multiple successfully matched second feature information, and organize and generate the first reference massage information according to the massage mode, including:

[0031] Calculate the cosine similarity between each second feature information and the first initial feature information, which is represented by formula (6), and formula (6) is as follows:

[0032] ;

[0033] In formula (6), is the cosine similarity function, is the Euclidean norm of, is the Euclidean norm of, is the cosine similarity value;

[0034] Judging one by one whether the cosine similarity value corresponding to the second feature information is within the range of the preset similarity threshold;

[0035] If so, record the second feature information as the first reference feature information, and obtain the massage parameter information associated with the massage mode corresponding to the first reference feature information, and map and store the massage parameter information with the first reference feature information to obtain the first reference massage information;

[0036] If not, it means that the current second feature information does not match the first initial feature information successfully.

[0037] In some embodiments, perform weighted averaging on the massage parameters of the same massage point category in multiple first reference massage information, and generate the second reference massage information according to the multiple weighted average massage parameters, including:

[0038] The first reference massage information is represented by formula (7). Denote the multiple massage parameters in the first reference massage information as the first massage parameters. Formula (7) is as follows:

[0039] ;

[0040] In formula (7), is the th first reference massage information, is the th first massage parameter of the th massage point category in the

[0041] Perform weighted averaging on the first massage parameters of the same massage point category in multiple first reference massage information to obtain the weighted average massage parameters, denoted as the second massage parameters, which are represented by formula (8). Formula (8) is as follows:

[0042] ;

[0043] In formula (8), is the th weighted average second massage parameter, , is the th weight of the is the th first massage parameter of the th massage point category in the

[0044] Organize the second massage parameters to obtain the second reference massage information, which is represented by formula (9). Formula (9) is as follows:

[0045] ;

[0046] In formula (9), is the second reference massage information.

[0047] In some embodiments, the first usage feature information includes a first dimension feature, a second dimension feature, and a third dimension feature. The first dimension feature is configured to be generated by physiological feedback information, the second dimension feature is configured to be generated by manual adjustment information, and the third dimension feature is configured to be generated by feedback evaluation information;

[0048] Generating the first usage feature information according to the feedback information, and adjusting the second reference massage information according to the first usage feature information to obtain the third reference massage information includes:

[0049] Collect the feedback information of the current batch of users at the first preset frequency to obtain the first-dimensional feature, the second-dimensional feature, and the third-dimensional feature;

[0050] Input the first-dimensional feature, the second-dimensional feature, and the third-dimensional feature into the EKF model based on Kalman filtering for fusion to obtain a fused multi-dimensional feature vector;

[0051] Use the multi-dimensional feature vector as the input feature and input it into the DQN algorithm model to obtain the third reference massage information.

[0052] In some embodiments, the first-dimensional feature is configured to be generated through physiological feedback information, including:

[0053] Collect the pressure feedback information corresponding to the massage position information within a preset time period, the heart rate information of the current user within a preset time period, and the skin conductivity information of the current user within a preset time period;

[0054] Arrange the pressure feedback information in chronological order to obtain a pressure value change sequence, arrange the heart rate information in chronological order to obtain a heart rate value change sequence, and arrange the skin conductivity information in chronological order to obtain a skin conductivity change sequence;

[0055] Calculate the pressure feature corresponding to the pressure feedback information of the current user according to the pressure value change sequence. The pressure feature includes the mean of the pressure feedback information, the standard deviation of the pressure feedback information, and the peak value of the pressure feedback information;

[0056] Calculate the heart rate feature corresponding to the heart rate information of the current user according to the heart rate value change sequence. The heart rate feature includes the average of the heart rate information and the coefficient of variation of the heart rate information;

[0057] Calculate the skin conductivity feature corresponding to the skin conductivity information of the current user according to the skin conductivity change sequence. The skin conductivity feature includes the average of the skin conductivity information and the fluctuation amplitude of the skin conductivity information;

[0058] Perform feature splicing on the pressure feature, the heart rate feature, and the skin conductivity feature to obtain the first-dimensional feature;

[0059] The second-dimensional feature is configured to be generated through manual adjustment information, including:

[0060] Collect the massage intensity adjustment information, massage temperature adjustment information, and massage frequency adjustment information within a preset time period. The massage intensity adjustment information includes the number of adjustments of the massage intensity and the adjustment value of the massage intensity within the current preset time period. The massage temperature adjustment information includes the number of adjustments of the massage temperature and the adjustment value of the massage temperature within the current preset time period. The massage frequency adjustment information includes the number of adjustments of the massage frequency and the adjustment value of the massage frequency within the current preset time period;

[0061] Construct a massage intensity adjustment curve based on the massage intensity adjustment information, obtain the first change slope value of the massage intensity per unit time according to the massage intensity adjustment curve, and generate a massage intensity adjustment feature based on the first change slope value;

[0062] Construct a massage temperature adjustment curve based on the massage temperature adjustment information, obtain the second change slope value of the massage temperature per unit time according to the massage temperature adjustment curve, and generate a massage temperature adjustment feature based on the second change slope value;

[0063] Construct a massage frequency adjustment curve based on the massage frequency adjustment information, obtain the third change slope value of the massage frequency per unit time according to the massage frequency adjustment curve, and generate a massage frequency adjustment feature based on the third change slope value;

[0064] Perform feature splicing on the massage intensity adjustment feature, massage temperature adjustment feature, and massage frequency adjustment feature to obtain the second-dimensional feature.

[0065] In some embodiments, the third-dimensional feature is configured to be generated through feedback evaluation information, including:

[0066] Collect multiple review texts within a preset time period, and perform the following steps on each review text:

[0067] Perform sentence segment recognition on the review text to obtain multiple basic sentence segments;

[0068] Input the basic sentence segments into a language recognition model to obtain emotion features, where the emotion features include emotion polarity and emotion intensity. The emotion polarity includes positive emotion and negative emotion, and the emotion intensity includes the intensity value of positive emotion and the intensity value of negative emotion;

[0069] Group the multiple basic sentence segments according to their relevance to obtain at least one first sentence segment group;

[0070] Arrange the multiple emotion features in the first sentence segment group in chronological order one by one to obtain the first emotion feature sequence of the current user, and construct an emotion fluctuation curve based on the first emotion feature sequence;

[0071] Calculate the fourth change slope value corresponding to each emotion fluctuation curve;

[0072] Concatenate multiple fourth change slope values to obtain a second emotional feature sequence corresponding to the current review text;

[0073] Concatenate multiple second emotional feature sequences to obtain a third-dimensional feature.

[0074] In some embodiments, input the first-dimensional feature, the second-dimensional feature, and the third-dimensional feature into an EKF model based on Kalman filtering for fusion, and the obtained fused multi-dimensional feature vector includes:

[0075] Represent the first-dimensional feature, the second-dimensional feature, and the third-dimensional feature by formula (10), and formula (10) is as follows:

[0076] ;

[0077] In formula (10), is the first vector expression of the first-dimensional feature, the second-dimensional feature, and the third-dimensional feature as the input features of the EKF model, is the first-dimensional feature, is the second-dimensional feature, is the third-dimensional feature;

[0078] Split into multiple second vector expressions according to the acquisition time within a preset time period, and the second vector expression is represented by formula (11), and formula (11) is as follows:

[0079] ;

[0080] In formula (11), is the second vector expression at the -th moment in the preset time period, is the first-dimensional feature at the -th moment, is the second-dimensional feature at the -th moment, is the third-dimensional feature at the -th moment;

[0081] Construct an EKF model based on Kalman filtering, use the second vector expression as the state vector, and define the state equation and the observation equation. The state equation is represented by formula (12), and formula (12) is as follows:

[0082] ;

[0083] The observation equation is represented by formula (13), and formula (13) is as follows:

[0084]

[0085] In Equations (12) and (13), is the state vector at the -th moment, is the parameter vector at the -th moment, is the state noise vector at the -th moment, is the state transition function, is the observation function, is the observation noise vector at the -th moment; is the observation vector at the -th moment;

[0086] Initialize the EKF model and execute the prediction step, which is represented by Equation (14) as follows:

[0087] ;

[0088] In Equation (14), is the state transition Jacobian matrix, is the transpose matrix of the state transition Jacobian matrix, is the state noise vector predicted based on the information at the previous moment at the current moment, is the state noise vector predicted based on the information at the previous moment at the moment, is the state estimation covariance obtained by predicting based on the state estimation and input at the previous moment at the current moment, is the state estimation covariance obtained by predicting based on the state estimation and input at the previous moment at the moment;

[0089] Execute the update step, which is represented by Equation (15) as follows:

[0090] ;

[0091] In Equation (15), is the Kalman gain matrix, is the state estimation covariance obtained by predicting based on the state estimation and input at the current moment, is the observation matrix, is the transpose matrix of the observation matrix, is the observation noise covariance matrix, is the state noise vector for information prediction at the current time instant, is the identity matrix;

[0092] After the acquisition time iteration within the preset time period is completed, the state vector finally estimated by EKF is obtained, denoted as the fused multi-dimensional feature vector .

[0093] In some embodiments, the DQN algorithm model includes a DQN online network model and a DQN target network model;

[0094] Taking the multi-dimensional feature vector as the input feature and inputting it into the DQN algorithm model, the obtained third reference massage information includes:

[0095] Inputting the multi-dimensional feature vector into the DQN online network model, which is represented by formula (16), and formula (16) is as follows:

[0096] ;

[0097] In formula (16), is the DQN online network model, is the DQN online network model function, is the th state corresponding to the th time instant in the DQN online network model, is the action value prediction of the second massage parameter for the th massage point category;

[0098] Inputting the output result of the DQN online network model into the DQN target network model, and calculating the prediction result of the maximum action value prediction of the second massage parameter in the future state as the third reference massage information, which is represented by formula (17), and formula (17) is as follows:

[0099] ;

[0100] In formula (17), is the prediction result of the maximum action value prediction of the second massage parameter for the th massage point category, that is, the third massage parameter for the th massage point category;

[0101] Repeat the foregoing steps until the third massage parameters corresponding to massage point categories are generated, and all the third massage parameters are sorted to form the third reference massage information, which is represented by formula (18), and formula (18) is as follows:

[0102] ;

[0103] In Formula (18), is the third reference massage information.

[0104] In a second aspect, the present invention further provides a graphene massager control system with multi-dimensional somatosensory feedback, which is applicable to the graphene massager control method with multi-dimensional somatosensory feedback described in the first aspect. The system includes an information acquisition module, a massage parameter configuration module, a feedback analysis module, and a personal data management module. The information acquisition module is used to obtain first user information, and the first user information includes information characteristics of five information categories: the occupation information, work and rest information, diet information, massage position information, and personal information of the current user; the massage parameter configuration module is used to generate first initial feature information according to the first user information, match second feature information similar to the first initial feature information in a database using a cosine similarity function, where the second feature information is second user information generated by other users when using a graphene massager of the same massage device category; obtain massage parameter information associated with the massage mode corresponding to multiple successfully matched second feature information, sort and generate first reference massage information according to the massage mode; perform weighted averaging on the massage parameters of the same massage point category in multiple first reference massage information, and generate second reference massage information according to multiple weighted average massage parameters; obtain the initial massage mode of the current user according to the second reference massage information and massage the current user;

[0105] The feedback analysis module is used to collect the feedback information of the user during the massage process. The feedback information includes physiological feedback information, manual adjustment information, and feedback evaluation information. The physiological feedback information includes at least one of pressure feedback information, heart rate information, and skin conductivity information. The manual adjustment information includes massage intensity adjustment information, massage temperature adjustment information, and massage frequency adjustment information. The feedback evaluation information includes the commentary text and commentary score input by the user during the massage process; the massage parameter configuration module is further used to generate first usage feature information according to the feedback information, and adjust the second reference massage information according to the first usage feature information to obtain third reference massage information; generate a usage massage mode according to the third reference massage information, and replace the current initial massage mode during the next batch of massage; the personal data management module is used to construct a personal usage database, map and store the feedback information of each batch of massage and the usage massage mode, and store them in the personal usage database in chronological order.

[0106] In a third aspect, the present invention further provides a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described in the first aspect is implemented.

[0107] Adopting the above technical solution, compared with the prior art, the present invention has the following beneficial effects:

[0108] Different from the prior art, the above technical solution obtains multi-dimensional features of the current user's occupation information, work and rest information, diet information, massage position information, and personal information, and matches them with the usage data of other users. Through the usage data of other users in the database, combined with the first user information of the current user, a personalized initial massage mode is generated, improving the intelligent level of the device. At the same time, during the massage process, the user's physiological feedback information, manual adjustment information, and feedback evaluation information are collected to dynamically adjust the massage parameters, providing a more personalized massage experience for the current user, dynamically optimizing the second reference massage information, and further enhancing the intelligent perception ability of the device. Moreover, the feedback information and usage mode of each current user's massage process are recorded in the personal usage database, providing data support for the optimization of subsequent massage modes and enhancing the overall user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0109] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0110] Figure 1 It is a schematic diagram of steps S101 to S107 of the graphene massager control method;

[0111] Figure 2 It is a schematic diagram of steps S201 to S206 of the graphene massager control method;

[0112] Figure 3 It is a schematic diagram of steps S301 to S305 of the graphene massager control method;

[0113] Figure 4 It is a schematic diagram of steps S401 to S407 of the graphene massager control method;

[0114] Figure 5 It is a schematic diagram of the graphene massager control system.

[0115] REFERENCE SIGNS:

[0116] 1. Graphene massager control system;

[0117] 11. Information acquisition module;

[0118] 12. Massage parameter configuration module;

[0119] 13. Feedback analysis module;

[0120] 14. Personal data management module. Specific implementation manners

[0121] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be specifically noted that the following embodiments are only used to illustrate the present invention, but do not limit the scope of the present invention. Similarly, the following embodiments are only partial embodiments of the present invention rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0122] Please refer to Figure 1 , in the first aspect, the present embodiment provides a control method for a graphene massager with multi-dimensional somatosensory feedback, including:

[0123] S101. Obtain first user information, where the first user information includes information characteristics of different information categories; the information categories include occupation information, work and rest information, diet information, massage position information, and personal information;

[0124] S102. Generate first initial feature information according to the first user information, and use the cosine similarity function in the database to match second feature information, where the second feature information is generated when other users use a graphene massager of the same massage device category;

[0125] S103. Obtain massage parameter information associated with the massage mode of the successfully matched second feature information, and generate first reference massage information;

[0126] S104. Perform weighted averaging on the massage parameters of the same massage point category in multiple first reference massage information to generate second reference massage information;

[0127] S105. Obtain the initial massage mode of the current user according to the second reference massage information and perform massage on the current user; and, during the massage process, collect feedback information of the user, where the feedback information includes physiological feedback information, manual adjustment information, and feedback evaluation information; the physiological feedback information includes at least one of pressure feedback information, heart rate information, and skin conductivity information, the manual adjustment information includes massage intensity adjustment information, massage temperature adjustment information, and massage frequency adjustment information, and the feedback evaluation information includes the commentary text and commentary score input by the user during the massage process;

[0128] S106. Generate first usage feature information according to the feedback information, and adjust the second reference massage information according to the first usage feature information to obtain third reference massage information;

[0129] S107. Generate a usage massage mode based on the third reference massage information and replace the current initial massage mode during the next batch of massages; and construct a personal usage database to store the feedback information of each batch of massages in chronological order in the personal usage database.

[0130] In step S101, the first user information is the information of the user currently using this massager. It can be understood that the first user information can be imported in the form of manual input or input through a third-party application software when the user first uses the current massager. Specifically, the first user information includes occupation information, work and rest information, diet information, massage position information, and personal information. Among them, the occupation information includes the occupation name, occupation place, and main duties. The occupation information can be used as one of the reference indicators for the distribution of the user's whole-body muscle status. For example, a user who is responsible for clerical work in the office is more likely to have lumbar muscle strain, lumbar disc herniation, and cervical spondylosis. In the express delivery profession, users are more likely to have leg muscle soreness and arm muscle soreness. The muscle strain groups of users in different occupations are different, and their massage needs are also different. The work and rest information includes the user's nap information and the night sleep time period, so that the time node when the user uses the massager can be more in line with the user's work and rest information. For example, before going to bed at night, the user's massage need is more inclined to a soothing massage mode, which helps to fall asleep. Another example is that after waking up from a nap, the user's massage need is more inclined to a massage mode for targeted muscle massage, and so on. As one of the data of the user's living habits, the diet information can infer the current user's diet category and further serve as a reference indicator for the health level of the user's living habits. For example, a user with diet information rich in protein and fiber can reflect a higher health level of the current user's living habits. On this basis, a reasonable prediction and inference can be further made on the muscle content in the user's body. That is, fitness personnel have a higher muscle content, and the massage parameters in the adapted massage mode may be stronger. Obese people have a higher fat content, and the massage parameters in the adapted massage mode will be adjusted repeatedly over time.

[0131] The massage position information specifically includes the body parts of the current user during massage. It can be understood that when constructing the personal usage database of the current user, a rough classification can be made according to the massage position information, and then the massage data of each batch of the user can be correspondingly stored in the same massage position information. Specifically, the massage position information can include the lower back, the lower back and abdomen, the upper buttocks, the lower buttocks, the thighs, the calves, the back of the neck, the shoulders, etc. The specific noun settings can refer to specialized medical terms or the image indications of body parts as an auxiliary input method for the user's massage position information. Personal information includes the user's name, contact number, age, medical history information, etc. The input of personal information can also be used as one of the reference indicators for setting subsequent massage parameters. That is, the massage needs of the elderly and middle-aged people are different. Similarly, the massage needs of middle-aged people and young people are also different. The input of personal information is also helpful for the establishment and distribution of the personal usage database.

[0132] In step S102, the first initial feature information is generated according to the first user information. Specifically, it can be understood that the numerical information and non-numerical information included in the first user information are uniformly represented by vectors, and on this basis, the second feature information is matched in the database. In this embodiment, each person will generate usage data when using the massager. On this basis, a cloud database is constructed, and the usage data of each person is uploaded to this database. The usage data includes the second feature information generated according to the second user information, as well as the fine-tuning data of other users during the use process, etc. These data can all be used as further reference data for the subsequent product improvement or research of the massager. On this basis, this embodiment uses the second feature information in the database to match the user group similar to the current user. This similarity is calculated through the cosine similarity function.

[0133] Further, in step S103, the massage parameter information recorded in the massage mode used by the successfully matched second feature information is sorted into the first reference massage information. For example, the number of second feature information that matches the first initial feature information is three. That is, there are three other users who are very similar to the current user. Then, obtain the massage modes corresponding to these three other users. There are a total of three massage modes, and the massage parameters in these three massage modes are sorted into three first reference massage information. The specific sorting method can be to sort the massage parameters in the massage mode into a set and record this set as the first reference massage information, or to sort the massage parameters in the massage mode into a vector expression and record it as the first reference massage information.

[0134] Further, in step S104, after obtaining multiple first reference massage information, the massage parameters of the same massage point category in the multiple first reference massage information are weighted and averaged. In this embodiment, the massage point category can be understood as follows: in the same massager, the massage execution end has temperature control parameters, strength control parameters, and frequency control parameters. Each first reference massage information has temperature control parameters, strength control parameters, and frequency control parameters. More specifically, these three types of parameters are further divided by massage points, that is, a massage execution end can have multiple massage points, and each massage point can be equipped with temperature control parameters, strength control parameters, frequency control parameters, etc. In some other alternative embodiments, one of the temperature control parameters, strength control parameters, and frequency control parameters can be retained, which is in line with the settings of various massage execution ends. On this basis, the same massage point category means that the massage parameters that are also temperature control parameters are selected from the multiple first reference massage information for weighted averaging. The strength control parameters and frequency control parameters are understood in this way. The obtained massage parameters are the weighted average massage parameters, which are recorded as the massage parameters in the second reference massage information. This method can eliminate the individual differences of other users who are similar to the current user, so as to give a general reference data, that is, the second reference massage information.

[0135] In step S105, the initial massage mode of the massager used by the current user is initialized according to the second reference massage information, and the current user is massaged. This method utilizes the screening, classification, and similarity matching of big data, so as to consider the individual needs of the current user at the initial stage, generate an initial massage mode that is more suitable for the current user to massage the current user, and give the user a more comfortable usage experience and happiness index. Further, during the massage process, the feedback information of the user is collected, and the collection method of the feedback information can be realized by arranging sensors, interacting with the user through voice or text, etc. Specifically, the feedback information includes physiological feedback information, manual adjustment information, and feedback evaluation information. Among them, the physiological feedback information is the physiological information naturally generated by the current user during the massage process, which can also be understood as the somatosensory feedback. The physiological feedback information specifically includes pressure feedback information, heart rate information, and skin conductivity information. The pressure feedback information specifically includes the pressure feedback information of the massage execution end of the massager. The heart rate information is the heart rate information of the user during the massage process. The degree of relaxation and stress level generated by the user due to the massage of the massager will be reflected by the change of the heart rate. Similarly, the skin conductivity information can reflect the skin humidity and blood circulation state of the human body. Different massage intensities will cause changes in skin humidity and blood circulation state. In this embodiment, the physiological information of these three categories is obtained as the physiological feedback information. The manual adjustment information is the adjustment data generated by the current user through interaction with a third-party application, the buttons or the screen on the massager during the massage process. For example, the massage parameters of a certain massage point category in the current massage mode are adjusted numerically, which can more intuitively reflect the adjustment needs of the current user for the massage. The aforementioned physiological feedback parameters quantify the adjustment needs of the user for the current massage mode from the perspective of somatosensation. Further, the feedback evaluation information includes the comment text and comment score input by the user during the massage process. Specifically, the interaction with the user can be carried out in the form of a pop-up window to input the comment text and comment score, or the form of voice inquiry can be used. The finally output feedback evaluation information is the comment text and comment score.

[0136] In step S106, the first usage feature information is generated using the feedback information. It should be noted that the collection time of the feedback information can be further refined and standardized. Specifically, in this embodiment, a complete massage cycle of the user is regarded as a batch. In one batch, there can be multiple feedback information. For example, if the time for the user to massage one batch is 30 minutes, the collection frequency of the feedback information is 10 minutes / time or 5 minutes / time. By frequently collecting the feedback information, more detailed feedback content of the user during the massage process can be obtained, so as to optimize the massage parameters of subsequent batches. Optionally, in some embodiments, after setting the massage parameters for multiple batches, the collection frequency of the feedback information for each batch can be reduced, so as to achieve a more comfortable massage process and improve the user experience.

[0137] In this embodiment, generating the first usage feature information using the feedback information can be understood as reflecting the numerical data and non-numerical data in the feedback information in the form of vectors or matrices. On this basis, it is converted into the adjustment value of the second reference massage information, and the massage parameters in the second reference massage information are adjusted and modified through this adjustment value. Finally, the modified second reference massage information is obtained, denoted as the third reference massage information.

[0138] In step S107, the usage massage mode is generated according to the third reference massage information. It should be noted that the usage massage mode is used to distinguish from the initial massage mode. Moreover, step S106 has the effect of multiple iterations in the single-batch massage process, that is, the usage massage mode is updated once after each collection of the feedback information. The last usage massage mode is used as the initial massage mode for the next batch of massage. On this basis, when the user first uses this massager, a personal usage database is constructed, and after the massage operation of each batch is completed, the massage data generated in that batch will be stored, specifically including multiple feedback information corresponding to the massage operation of each batch and the usage massage mode at each iteration, and stored in the personal usage database in the order of the user's usage time of the current massager. This method is convenient for the subsequent product upgrade of the massager and the portrayal of the user's usage profile. Further, the data in the personal usage database can be used to synchronously analyze the change of the user's usage needs within a period of time, as the evaluation basis for the user's personal physical health.

[0139] This embodiment makes full use of big data and machine learning technologies to generate a personalized initial massage plan based on the first user information (such as occupation information, work and rest information, diet information, etc.) of the current user and the usage data of other users, providing the user with a massage experience more suitable for their own needs. On this basis, various feedback information of the user is collected in real time during the actual use process, and the feedback information is used to continuously optimize and adjust the massage parameters, thereby further improving the user's satisfaction and experience. This embodiment also establishes a personal usage database to record the usage history and feedback information of the current user, which can not only provide a more accurate personalized plan for the next use, but also be used to analyze the current user's physical health status and the changing trend of needs. This embodiment improves the intelligent adjustment level of the existing massager, realizes dynamic adjustment during the massage process, and improves the user's comfort during use.

[0140] In some embodiments, generating the first initial feature information according to the first user information includes:

[0141] Performing one-hot encoding on the information features in the first user information that are not numerically represented, and representing the first user information by formula (1), and formula (1) is as follows:

[0142] ;

[0143] In formula (1), is the vector representation of the first user information, is the th numerical representation of the information feature of the th information category in the first user information;

[0144] Performing standardization processing on the vector representation of the first user information to obtain the first initial feature information, which is represented by formula (2) to formula (4), and formula (2) is as follows:

[0145] ;

[0146] Formula (3) is as follows:

[0147] ;

[0148] Formula (4) is as follows:

[0149] ;

[0150] In formula (2) to formula (4), is the first initial feature information, is the average value of is the th eigenvalue of the information feature, is the The average value of the information features, is the standard deviation vector of the standard deviation vector of the th information feature, ;

[0151] The second feature information is represented by formula (5), and formula (5) is as follows:

[0152] ;

[0153] In formula (5), is the second feature information, is the numerical representation of the information feature of the th information category in the second user information;

[0154] In the database, use the cosine similarity function to match the second feature information similar to the first initial feature information, obtain the massage parameter information associated with the massage mode corresponding to multiple successfully matched second feature information, and organize and generate the first reference massage information according to the massage mode, including:

[0155] Calculate the cosine similarity between each second feature information and the first initial feature information, which is represented by formula (6), and formula (6) is as follows:

[0156] ;

[0157] In formula (6), is the cosine similarity function, is the Euclidean norm of is the Euclidean norm of is the cosine similarity value;

[0158] Judging one by one whether the cosine similarity value corresponding to the second feature information is within the range of the preset similarity threshold;

[0159] If so, record the second feature information as the first reference feature information, and obtain the massage parameter information associated with the massage mode corresponding to the first reference feature information, and map and store the massage parameter information with the first reference feature information to obtain the first reference massage information;

[0160] If not, it means that the current second feature information does not match the first initial feature information successfully.

[0161] In this embodiment, one-hot encoding is performed on the information features represented by non-numerical values in the first user information. For example, if the first user information contains non-numerical representation information such as sleep time, deep sleep, light sleep, and afternoon nap sleep, then binary vector representation can be performed through one-hot encoding, thereby converting it into a numericalized feature. For some quantifiable metrics, one-hot encoding is not required, and they can be directly used as a vector in the first user information. After one-hot encoding, the first user information can be represented by formula (1). It can be understood that ; that is, the foregoing information categories include five information categories: occupation information, work and rest information, diet information, massage position information, and personal information. For the convenience of description, in this listed order, the occupation information is recorded as the first information category, corresponding to ; the work and rest information is recorded as the second information category, corresponding to ; the diet information is recorded as the third information category, corresponding to ; the massage position information is recorded as the fourth information category, corresponding to ; the personal information is recorded as the fifth information category, corresponding to .

[0162] Furthermore, the first user information is standardized. The first user information after processing is denoted as the first initial feature information. Specifically, using the standardization formula, there are multiple feature values in each information category. After accumulating and averaging the feature values in each information category, the average value of the feature values in each information category is obtained. Further, the standard deviation vector corresponding to each information category is obtained. Finally, the first initial feature information is obtained. The first initial feature information includes the standardized features corresponding to each information category, that is, the meaning represented by formula (2).

[0163] Furthermore, the second feature information is represented by formula (5). Similar to the vector representation of the first user information, the difference is that the vector representation of the second user information is the version uploaded to the database. That is, the second user information here is the standardized feature. Therefore, the second user information is directly vectorized to obtain the second feature information.

[0164] In this embodiment, the cosine similarity function is used to calculate the similarity between the first initial feature information and the second feature information. By calculating the similarity between the second feature information and the first initial feature information one by one, the corresponding cosine similarity value between each second feature information and the first initial feature information is obtained.

[0165] It is necessary to determine whether the cosine similarity value is within the range of a preset similarity threshold. It should be noted that the range of the preset similarity threshold can be achieved through a form of manual pre - setting. In this embodiment, if the cosine similarity value is within the range of the preset similarity threshold, it indicates that the current second feature information has a certain similarity with the first initial feature information. This second feature information is recorded as the first reference feature information, and then the massage parameter information associated with the first reference feature information is obtained and mapped and stored with it to obtain the first reference massage information.

[0166] By repeating the above - mentioned steps, the second feature information within the range of the preset similarity threshold can be screened out from the database, and then multiple first reference massage information can be obtained.

[0167] In this embodiment, the non - numerical features in the first user information are one - hot encoded to obtain the first initial feature information, and the second user information is represented as the second feature information. The cosine similarity between the two is calculated through the cosine similarity function, and it is determined whether the cosine similarity is within the preset threshold range. If so, the second feature information is screened out from the database, and the corresponding massage parameter information is obtained to form the first reference massage information. Through standardization processing in this embodiment, the dimensional difference between different feature values in the same information category is eliminated, and using the cosine similarity can more accurately measure the similarity degree between the first initial feature information and the second feature information. Associating and storing the first reference feature information and the massage parameter information facilitates the generation of the first reference massage information.

[0168] In some embodiments, performing weighted averaging on the massage parameters of the same massage point category in multiple first reference massage information, and generating the second reference massage information according to the multiple weighted - averaged massage parameters includes:

[0169] Represent the first reference massage information by formula (7). Denote the multiple massage parameters in the first reference massage information as the first massage parameters. Formula (7) is as follows:

[0170] ;

[0171] In formula (7), is the th first reference massage information, is the th first reference massage information, and the th first massage parameter of the

[0172] Perform weighted averaging on the first massage parameters of the same massage point category in multiple first reference massage information to obtain the weighted - averaged massage parameters, denoted as the second massage parameters, which are represented by formula (8). Formula (8) is as follows:

[0173] ;

[0174] In formula (8), is the th second massage parameter after weighted average, , is the weight of the th first reference massage information, is the th first massage parameter of the th massage point category in the

[0175] The second massage parameters are organized to obtain the second reference massage information, which is represented by formula (9). Formula (9) is as follows:

[0176] ;

[0177] In formula (9), is the second reference massage information.

[0178] In this embodiment, for the convenience of description, the multiple massage parameters in the first reference massage information are denoted as the first massage parameters, and the parameters in the second reference massage information are denoted as the second massage parameters. The weighted average of the first massage parameters in the massage point category shown in the foregoing step S104 can be specifically understood by combining formula (7) and formula (8). Regarding the weight of the first reference massage information, it can be set according to actual needs. After weighted average, the second massage parameters corresponding to the first massage parameters of the same massage point category are obtained, and then the second massage parameters are organized according to the massage point category to form the second reference massage information.

[0179] In some embodiments, the first usage feature information includes a first dimension feature, a second dimension feature, and a third dimension feature. The first dimension feature is configured to be generated through physiological feedback information, the second dimension feature is configured to be generated through manual adjustment information, and the third dimension feature is configured to be generated through feedback evaluation information;

[0180] Generating the first usage feature information according to the feedback information, and adjusting the second reference massage information according to the first usage feature information to obtain the third reference massage information includes:

[0181] Collecting the feedback information of the current batch of users at the first preset frequency to obtain the first dimension feature, the second dimension feature, and the third dimension feature;

[0182] Input the first - dimension feature, second - dimension feature, and third - dimension feature into the EKF model based on Kalman filtering for fusion to obtain a fused multi - dimensional feature vector;

[0183] Take the multi - dimensional feature vector as the input feature and input it into the DQN algorithm model to obtain the third reference massage information.

[0184] In this embodiment, the feedback information of the current batch of users is collected at the first preset frequency. It can be understood that during the massage process of the current batch, the feedback information is collected at intervals, and the adjacent interval time periods can be calculated through the first preset frequency, that is, the preset time section mentioned later. That is, the collection time of the feedback information has a certain interval, so as to realize timed feedback and adjustment during the massage operation of the same batch, and realize the adjustment and change of the dynamic massage intensity.

[0185] Furthermore, in this embodiment, the EKF model is used to fuse the first - dimension feature, second - dimension feature, and third - dimension feature in the feedback information to obtain a multi - dimensional feature vector, and then input it into the DQN algorithm model to obtain the third reference massage information. For the convenience of description, the massage parameters in the third reference massage information can be denoted as the third massage parameters.

[0186] Furthermore, please refer to Figure 2 , in some embodiments, the first - dimension feature is configured to be generated through physiological feedback information, including:

[0187] S201. Collect the pressure feedback information corresponding to the massage position information within the preset time section, the heart rate information of the current user within the preset time section, and the skin conductivity information of the current user within the preset time section;

[0188] S202. Arrange the pressure feedback information in chronological order to obtain a pressure value change sequence, arrange the heart rate information in chronological order to obtain a heart rate value change sequence, and arrange the skin conductivity information in chronological order to obtain a skin conductivity change sequence;

[0189] S203. Calculate the pressure feature corresponding to the pressure feedback information of the current user according to the pressure value change sequence. The pressure feature includes the mean value of the pressure feedback information, the standard deviation of the pressure feedback information, and the peak value of the pressure feedback information;

[0190] S204. Calculate the heart rate feature corresponding to the heart rate information of the current user according to the heart rate value change sequence. The heart rate feature includes the average value of the heart rate information and the coefficient of variation of the heart rate information;

[0191] S205. Calculate the skin conductivity characteristics corresponding to the skin conductivity information of the current user according to the skin conductivity change sequence. The skin conductivity characteristics include the average value of the skin conductivity information and the fluctuation range of the skin conductivity information.

[0192] S206. Concatenate the pressure characteristics, heart rate characteristics, and skin conductivity characteristics to obtain the first-dimensional characteristics.

[0193] In step S201, the preset time period is the time interval period in the foregoing first preset acquisition frequency. The first-dimensional characteristics include three characteristics: pressure feedback information, heart rate information, and skin conductivity information. Specifically, the heart rate value, pressure value, and skin conductivity of the user are collected in real time within this preset time period. The heart rate value is sorted into heart rate information, the pressure value is sorted into pressure feedback information, and the skin conductivity is sorted into skin conductivity information.

[0194] In step S202, arrange the pressure feedback information in chronological order to obtain the pressure value change sequence. Specifically, the pressure value change sequence can be used to construct a two-dimensional curve graph for visualization operations, or the pressure value change sequence can be subjected to logical operations and data processing to obtain the mean value, standard deviation, and peak value corresponding to the pressure value change sequence shown in step S203, which are denoted as the mean value, standard deviation, and peak value of the pressure feedback information, that is, the pressure characteristics.

[0195] In step S202, arrange the heart rate values in chronological order to obtain the heart rate value change sequence. In step S204, further calculate the average value of the heart rate values of the current user and the coefficient of variation of the heart rate values, which are denoted as the average value and coefficient of variation of the heart rate information, that is, the heart rate characteristics.

[0196] In step S202, arrange the skin conductivity in chronological order to obtain the skin conductivity change sequence. In step S205, further calculate the average value and fluctuation range of the skin conductivity, which are denoted as the average value and fluctuation range of the skin conductivity information, that is, the skin conductivity characteristics.

[0197] In step S206, concatenate the pressure characteristics, heart rate characteristics, and skin conductivity characteristics to obtain the first-dimensional characteristics. The first-dimensional characteristics reflect the physiological feelings of the user for the massage mode within the preset time period, that is, the body feeling, and will be used as one of the input characteristics for adjusting the massage parameters of the subsequent massage mode.

[0198] Further, please refer to Figure 3 , the second-dimensional characteristics are configured to be generated through manual adjustment information, including:

[0199] S301. Collect the massage intensity adjustment information, massage temperature adjustment information, and massage frequency adjustment information within a preset time period. The massage intensity adjustment information includes the number of adjustments and the adjustment values of the massage intensity within the current preset time period. The massage temperature adjustment information includes the number of adjustments and the adjustment values of the massage temperature within the current preset time period. The massage frequency adjustment information includes the number of adjustments and the adjustment values of the massage frequency within the current preset time period;

[0200] S302. Construct a massage intensity adjustment curve based on the massage intensity adjustment information, obtain the first change slope value of the massage intensity per unit time according to the massage intensity adjustment curve, and generate a massage intensity adjustment feature based on the first change slope value;

[0201] S303. Construct a massage temperature adjustment curve based on the massage temperature adjustment information, obtain the second change slope value of the massage temperature per unit time according to the massage temperature adjustment curve, and generate a massage temperature adjustment feature based on the second change slope value;

[0202] S304. Construct a massage frequency adjustment curve based on the massage frequency adjustment information, obtain the third change slope value of the massage frequency per unit time according to the massage frequency adjustment curve, and generate a massage frequency adjustment feature based on the third change slope value;

[0203] S305. Concatenate the massage intensity adjustment feature, the massage temperature adjustment feature, and the massage frequency adjustment feature to obtain the second-dimensional feature.

[0204] In step S301, the manual adjustment information is divided into massage intensity adjustment information, massage temperature adjustment information, and massage frequency adjustment information. Among them, the massage intensity adjustment information is the number of adjustments of the massage intensity and the adjustment values of the adjustment intensity input by the current user during the massage process. The massage temperature adjustment information is the number of adjustments of the massage temperature and the adjustment values of the adjustment temperature input by the current user during the massage process. The massage frequency adjustment information is the number of adjustments of the massage frequency and the adjustment values of the massage frequency input by the current user during the massage process.

[0205] In step S302, a massage intensity adjustment curve is constructed. Specifically, it can be formed by performing hyperbolic fitting on the adjustment times of the discretely distributed massage intensity and the adjustment values of the massage intensity. On this basis, according to the massage intensity adjustment curve, the first change slope value of the massage intensity within a unit time is obtained, which is also the adjustment trend of the user for the massage intensity. Furthermore, the first change slope value is represented as a vector to form the massage intensity adjustment feature. It should be noted that the unit time is the minimum recording time. For example, when the preset time period is 5 minutes, the unit time can be refined to 10 seconds or 1 second. When the preset time period is 1 hour, the unit time can be refined to 1 minute, etc. This embodiment does not limit this, and the order of magnitude of the unit time can be specifically defined according to time requirements.

[0206] In step S303, a massage temperature adjustment curve is constructed. Specifically, it can be formed by performing hyperbolic fitting on the adjustment times of the discretely distributed massage temperature and the adjustment values of the massage temperature. On this basis, according to the massage temperature adjustment curve, the second change slope value of the massage temperature within a unit time is obtained, which is also the adjustment trend of the user for the massage temperature. Furthermore, the second change slope value is represented as a vector to form the massage temperature adjustment feature.

[0207] In step S304, a massage frequency adjustment curve is constructed. Specifically, it can be formed by performing hyperbolic fitting on the adjustment times of the discretely distributed massage frequency and the adjustment values of the massage frequency. On this basis, according to the massage frequency adjustment curve, the third change slope value of the massage frequency within a unit time is obtained, which is also the adjustment trend of the user for the massage frequency. Furthermore, the third change slope value is represented as a vector to form the massage frequency adjustment feature.

[0208] In step S305, the massage intensity adjustment feature, the massage temperature adjustment feature, and the massage frequency adjustment feature are concatenated to obtain the second-dimensional feature. The second-dimensional feature reflects the user's intuitive requirements for the massage mode within the preset time period and will be used as one of the input features for adjusting the massage parameters of the subsequent massage mode.

[0209] Please refer to Figure 4 , in some embodiments, the third-dimensional feature is configured to be generated through feedback evaluation information, including:

[0210] Collect multiple review texts within the preset time period, and perform the following steps for each review text:

[0211] S401. Perform sentence segment recognition on the review text to obtain multiple basic sentence segments;

[0212] S402. Input the basic sentence segments into a language recognition model to obtain sentiment features, where the sentiment features include sentiment polarity and sentiment intensity. The sentiment polarity includes positive sentiment and negative sentiment, and the sentiment intensity includes the intensity value of positive sentiment and the intensity value of negative sentiment;

[0213] S403. Group multiple basic sentence segments according to their relevance to obtain at least one first sentence segment group;

[0214] S404. Arrange the multiple sentiment features in the first sentence segment group in chronological order one by one to obtain the first sentiment feature sequence of the current user, and construct a sentiment fluctuation curve based on the first sentiment feature sequence;

[0215] S405. Calculate the fourth change slope value corresponding to each sentiment fluctuation curve;

[0216] S406. Concatenate multiple fourth change slope values to obtain the second sentiment feature sequence corresponding to the current review text;

[0217] S407. Concatenate multiple second sentiment feature sequences to obtain the third-dimensional feature.

[0218] In this embodiment, the third-dimensional feature is for generating feedback evaluation information, and the feedback evaluation information is collected in the form of manual input or voice input by the user. Based on this, it is necessary to perform text processing on the feedback evaluation information.

[0219] Specifically, collect multiple review texts within a preset time period, that is, the feedback evaluation information includes multiple review texts within the preset time period. On this basis, in step S401, perform sentence segment recognition on the review text to obtain multiple basic sentence segments, which can be specifically achieved by recognizing punctuation marks in the sentence segments.

[0220] In step S402, input the basic sentence segments into a language recognition model. The language recognition model can specifically be a pre-trained language model such as BERT or RoBERTa. This language recognition model is also connected to a sentiment classifier, and after encoding the sentence segments, it will further perform sentiment classification to obtain sentiment features, where the sentiment features include sentiment polarity and sentiment intensity, that is, each basic sentence segment has a sentiment feature.

[0221] In step S403, group multiple basic sentence segments according to their relevance. The relevance shown in this embodiment can be understood as the relevance in the statement of the basic sentence segments, that is, the distribution of the main and secondary sentence segments in a review text. Specifically, the relevance grouping of the basic sentence segments can be performed by constructing a neural network or a tree diagram. In this way, at least one first sentence segment group can be obtained, that is, the number of first sentence segment groups can be multiple.

[0222] In step S404, multiple sentiment features in the first sentence segment group are arranged in chronological order to construct a first sentiment feature sequence. Then, the sentiment features in the discrete first sentiment feature sequence are fitted to obtain the sentiment fluctuation curve corresponding to the first sentence segment group.

[0223] In step S405, the fourth change slope value of each sentiment fluctuation curve is calculated. In step S406, multiple fourth change slope values are concatenated to obtain a second sentiment feature sequence.

[0224] Further, in step S407, multiple second sentiment feature sequences are concatenated to obtain a third-dimensional feature.

[0225] The above embodiments illustrate the specific generation processes of the first-dimensional feature, the second-dimensional feature, and the third-dimensional feature, facilitating subsequent further operations and analyses.

[0226] In some embodiments, the first-dimensional feature, the second-dimensional feature, and the third-dimensional feature are input into an EKF model based on Kalman filtering for fusion. The fused multi-dimensional feature vector includes:

[0227] The first-dimensional feature, the second-dimensional feature, and the third-dimensional feature are represented by formula (10), and formula (10) is as follows:

[0228] ;

[0229] In formula (10), is the first vector expression of the first-dimensional feature, the second-dimensional feature, and the third-dimensional feature as the input features of the EKF model, is the first-dimensional feature, is the second-dimensional feature, is the third-dimensional feature;

[0230] will be split into multiple second vector expressions according to the acquisition time within a preset time period. The second vector expression is represented by formula (11), and formula (11) is as follows:

[0231] ;

[0232] In formula (11), is the second vector expression at the th moment in the preset time period, is the first-dimensional feature at the th moment, is the second-dimensional feature at the th moment, is the The third-dimensional feature at a certain moment;

[0233] Construct an EKF model based on the Kalman filter, take the second vector expression as the state vector, and define the state equation and the observation equation. The state equation is represented by formula (12), and formula (12) is as follows:

[0234] ;

[0235] The observation equation is represented by formula (13), and formula (13) is as follows:

[0236]

[0237] In formula (12) and formula (13), is the state vector at the -th moment, is the parameter vector at the -th moment, is the state noise vector at the -th moment, is the state transition function, is the observation function, is the observation noise vector at the -th moment, is the -th moment of the observation vector;

[0238] Initialize the EKF model and execute the prediction step. The prediction step is represented by formula (14), and formula (14) is as follows:

[0239] ;

[0240] In formula (14), is the state transition Jacobian matrix, is the transpose matrix of the state transition Jacobian matrix, is the state noise vector predicted based on the information at the previous moment at the current moment, is at the moment based on the information at the previous moment predicted state noise vector, is at the current moment based on the state estimate and input prediction at the previous moment obtained state estimate covariance, at the moment based on the state estimate and input prediction at the previous moment obtained state estimate covariance;

[0241] Execute the update step, represented by Equation (15), and Equation (15) is as follows:

[0242] ;

[0243] In Equation (15), is the Kalman gain matrix, is the state estimation covariance obtained from the state estimation and input prediction at the current time, is the observation matrix, is the transpose matrix of the observation matrix, is the observation noise covariance matrix, is the state noise vector of the information prediction at the current time, is the identity matrix;

[0244] After the acquisition time iteration within the preset time period is completed, the state vector finally estimated by EKF is obtained, denoted as the fused multi-dimensional feature vector .

[0245] In this embodiment, the first-dimensional feature, the second-dimensional feature, and the third-dimensional feature are used as the input features of the EKF model, making full use of the complementary information of various sensing data to improve the richness of the expression of the input features. Through the prediction and update mechanism of EKF, the state information at historical times is used to better estimate the fused features at the current time. Moreover, the EKF model introduces noise terms in the state equation and the observation equation, which can effectively suppress the noise interference in each dimension of the features and improve the accuracy of the fusion result.

[0246] In some embodiments, the DQN algorithm model includes a DQN online network model and a DQN target network model;

[0247] Taking the multi-dimensional feature vector as the input feature and inputting it into the DQN algorithm model, the third reference massage information obtained includes:

[0248] Inputting the multi-dimensional feature vector into the DQN online network model, represented by Equation (16), and Equation (16) is as follows:

[0249] ;

[0250] In Equation (16), is the DQN online network model, is the DQN online network model function, is the th state corresponding to the th moment in the DQN online network model, is the Action value prediction of the second massage parameter for a massage point category;

[0251] Input the output result of the DQN online network model into the DQN target network model, and calculate the prediction result of the maximum action value prediction of the second massage parameter in the future state, which is used as the third reference massage information and is represented by formula (17). Formula (17) is as follows:

[0252] ;

[0253] In formula (17), is the prediction result of the maximum action value prediction of the second massage parameter for the th massage point category, that is, the third massage parameter for the th massage point category;

[0254] Repeat the above steps until the third massage parameters corresponding to massage point categories are generated. Organize all the third massage parameters to form the third reference massage information, which is represented by formula (18). Formula (18) is as follows:

[0255] ;

[0256] In formula (18), is the third reference massage information.

[0257] In this embodiment, the DQN model can dynamically learn and predict the optimal massage parameter scheme through the interaction and update of the online network and the target network, rather than simply making decisions based on static rules or experience, and better adapts to the complexity and dynamics of user needs. The DQN model has good learning and transfer ability. By continuously optimizing the online network and the target network, it can continuously improve the accuracy and robustness of decisions. The finally obtained third reference massage information specifically gives the optimal massage parameters for each massage point category, which can provide a more personalized massage scheme for users and improve the user's experience satisfaction.

[0258] Please refer to Figure 5, in a second aspect, the present embodiment further provides a graphene massager control system 1 with multi-dimensional somatosensory feedback, which is applicable to the graphene massager control method with multi-dimensional somatosensory feedback described in the first aspect. The system includes an information acquisition module 11, a massage parameter configuration module 12, a feedback analysis module 13, and a personal data management module 14. The information acquisition module 11 is used to obtain the first user information, and the first user information includes the information characteristics of five information categories: the occupation information, work and rest information, diet information, massage position information, and personal information of the current user; the massage parameter configuration module 12 is used to generate the first initial characteristic information according to the first user information, use the cosine similarity function in the database to match the second characteristic information similar to the first initial characteristic information, and the second characteristic information is the second user information generated by other users when using the graphene massager of the same massage device category; obtain the massage parameter information associated with the massage mode corresponding to multiple successfully matched second characteristic information, sort and generate the first reference massage information according to the massage mode; perform weighted averaging on the massage parameters of the same massage point category in multiple first reference massage information, and generate the second reference massage information according to multiple weighted-averaged massage parameters; obtain the initial massage mode of the current user according to the second reference massage information and massage the current user.

[0259] The feedback analysis module 13 is used to collect the feedback information of the user during the massage process. The feedback information includes physiological feedback information, manual adjustment information, and feedback evaluation information. The physiological feedback information includes at least one of pressure feedback information, heart rate information, and skin conductivity information. The manual adjustment information includes massage intensity adjustment information, massage temperature adjustment information, and massage frequency adjustment information. The feedback evaluation information includes the commentary text and commentary score input by the user during the massage process; the massage parameter configuration module 12 is further used to generate the first usage characteristic information according to the feedback information, and adjust the second reference massage information according to the first usage characteristic information to obtain the third reference massage information; generate the usage massage mode according to the third reference massage information, and replace the current initial massage mode during the next batch of massages; the personal data management module 14 is used to construct a personal usage database, map and store the feedback information of each batch of massages and the usage massage mode, and store them in the personal usage database in chronological order.

[0260] Corresponding to the foregoing text, the system provided in the present embodiment is used to implement the method recorded in the foregoing text, and the specific steps involved can be understood with reference to the foregoing text, and will not be elaborated herein.

[0261] In a third aspect, the present embodiment further provides a computer-readable storage medium, on which computer program instructions are stored, and the computer program instructions, when executed by a processor, implement the method described in the first aspect.

[0262] The computer program involved in this embodiment can be stored in a computer device-readable storage medium, which includes but is not limited to magnetic disks, magnetic tapes, magnetic cards, floppy disks, flash memories, optical discs, optical cards, read-only memories (ROMs), random access memories (RAMs), erasable programmable ROMs (EPROMs), and electrically erasable programmable ROMs (EEPROMs), etc. It also includes other biological, physical, or chemical structures that can achieve the same or equivalent functions as the above-listed storage media, such as units with information storage capabilities like DNA, RNA, proteins, etc. In a specific embodiment, the involved storage medium can be one of the above medium types or a combination of the above medium types. In different embodiments, the computer program involved in the embodiment can be centrally stored in a single medium or distributedly stored in multiple media. The memory containing the computer device-readable storage medium can be a non-volatile memory or a random access memory. These computer device-readable storage media can be built into the device or connected to the device involved in the embodiment as an external device or a part of an external device. In some embodiments, the memory with the computer device-readable storage medium is deployed locally; in other embodiments, a scheme of deploying the memory away from the processor can also be adopted, such as a network-attached memory accessed via an RF circuit or an external port and a communication network, where the communication network can be the Internet, one or more intranets, local area networks (LANs), wide area wireless networks (WLANs), storage area networks (SANs), etc., or a suitable combination thereof, as long as the computer device can access the memory. In addition, the computer program involved in the embodiment can be stored in plaintext / ciphertext form or designed as training data and integrated and recombinantly stored implicitly in the parameter states of a deep neural network or other machine learning models through model training.

[0263] Adopting the above technical solution, compared with the prior art, the beneficial effects of the present invention are as follows:

[0264] Different from the prior art, the above technical solution obtains multi-dimensional features of the current user's occupation information, work and rest information, diet information, massage position information, and personal information, and matches them with the usage data of other users. Through the usage data of other users in the database and combining the first user information of the current user, a personalized initial massage mode is generated, improving the intelligent level of the device. At the same time, during the massage process, the user's physiological feedback information, manual adjustment information, and feedback evaluation information are collected to dynamically adjust the massage parameters, providing a more personalized massage experience for the current user, dynamically optimizing the second reference massage information, and further enhancing the intelligent perception ability of the device. Moreover, the feedback information and usage pattern of each current user's massage process are recorded in the personal usage database, providing data support for the optimization of subsequent massage modes and improving the overall usage experience of the user.

[0265] In addition, each functional unit in various embodiments of the present invention may be integrated into one processing unit, may exist separately as individual physical units, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware or in the form of software functional units.

[0266] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, may be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0267] The above are only partial embodiments of the present invention, and thus do not limit the protection scope of the present invention. Any equivalent device or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be similarly included in the patent protection scope of the present invention.

Claims

1. A control method for a graphene massager with multi-dimensional somatosensory feedback, characterized in that: include: Acquire first user information, where the first user information includes information features of different information categories; Generate first initial feature information according to the first user information, and match second feature information in the database using a cosine similarity function, where the second feature information is generated when other users use a graphene massager of the same massage device category; Acquire massage parameter information associated with the massage mode of the second characteristic information that matches successfully, and generate first reference massage information; Taking a weighted average of massage parameters of the same massage point category in a plurality of first reference massage information to generate second reference massage information; obtaining an initial massage mode of the current user according to the second reference massage information and performing massage on the current user; Collecting user feedback information during the massage process, wherein the feedback information includes physiological feedback information, manual adjustment information, and feedback evaluation information; generating first usage characteristic information according to the feedback information, and adjusting the second reference massage information according to the first usage characteristic information to obtain third reference massage information; Generate a massage mode for use according to the third reference massage information, and replace the current initial massage mode in the next batch of massages; and, constructing a personal use database, storing the feedback information of each batch of massages in the personal use database in chronological order; The first usage feature information includes a first dimension feature, a second dimension feature, and a third dimension feature, the first dimension feature is configured to be generated by the physiological feedback information, the second dimension feature is configured to be generated by the manual adjustment information, and the third dimension feature is configured to be generated by the feedback evaluation information; Generating first usage characteristic information according to the feedback information, and adjusting the second reference massage information according to the first usage characteristic information to obtain third reference massage information includes: Collect feedback information of users of the current batch according to a first preset frequency to obtain the first dimensional features, the second dimensional features, and the third dimensional features; Inputting the first dimensional features, the second dimensional features and the third dimensional features into an EKF model based on Kalman filtering for fusion, to obtain a fused multi-dimensional feature vector; Inputting the multidimensional feature vector as an input feature into the DQN algorithm model to obtain the third reference massage information; The first dimensional features, the second dimensional features and the third dimensional features are input into the EKF model based on Kalman filtering for fusion, and the fused multi-dimensional feature vector includes: The first dimensional feature, the second dimensional feature and the third dimensional feature are expressed by formula (10), and the formula (10) is as follows: ； In formula (10), The first dimension feature, the second dimension feature, and the third dimension feature are used as the first vector expression of the input feature of the EKF model. is the first dimension feature, is the second dimension feature, It is the third dimension feature; Will The acquisition time in the preset time segment is divided into multiple second vector expressions, and the second vector expressions are expressed by formula (11), which is as follows: ； In formula (11), The preset time period The second vector expression of the moment, For the The first dimension feature of the moment, For the The second dimension feature of the moment, For the The third dimension characteristics of each moment; Construct an EKF model based on Kalman filtering, use the second vector expression as the state vector, define the state equation and the observation equation, and the state equation is expressed by formula (12), which is as follows: ； The observation equation is expressed by formula (13), which is as follows: In formula (12) and formula (13), For the The state vector at the moment, For the The parameter vector at each moment, For the The state noise vector at the moment is is the state transfer function, is the observation function, For the The observation noise vector at time instant, For the The observation vector at each moment; The EKF model is initialized and a prediction step is performed. The prediction step is represented by formula (14), which is as follows: ； In formula (14), is the state transfer Jacobian matrix, is the transposed matrix of the state transfer Jacobian matrix, For the current Based on the previous moment The state noise vector predicted by the information, For Based on the previous moment The state noise vector predicted by the information, For the current Based on the previous moment The state estimate covariance obtained by the state estimate and input prediction is, exist Based on the previous moment The state estimate covariance between the state estimate of and the state estimate obtained by input prediction; The update step is performed, which is expressed by formula (15), which is as follows: ； In formula (15), is the Kalman gain matrix, For the current The state estimation covariance obtained by the state estimation at time and the input prediction is, is the observation matrix, is the transposed matrix of the observation matrix, is the observation noise covariance matrix, For the current The state noise vector of the information prediction at the moment, is the identity matrix; After the acquisition time iteration within the preset time segment is completed, the final estimated state vector of the EKF is obtained, which is recorded as the fused multidimensional feature vector .

2. The control method of the graphene massager with multi-dimensional somatosensory feedback according to claim 1, characterized in that: Generating first initial characteristic information according to the first user information includes: The non-numerical information features in the first user information are one-hot encoded, and the first user information is represented by formula (1), which is as follows: ； In formula (1), is the vector representation of the first user information, The first user information The numerical representation of the information characteristics of each information category, ; The first user information represented by the vector is normalized to obtain the first initial feature information, which is expressed by formula (2) to formula (4). Formula (2) is as follows: ； Formula (3) is as follows: ； Formula (4) is as follows: ； In formula (2) to formula (4), is the first initial feature information, for The average value of For the The eigenvalues ​​of the information features, For the The average value of the information features, for The standard deviation vector of For the The standard deviation vector of the information features, ; The second characteristic information is expressed by formula (5), and the formula (5) is as follows: ； In formula (5), is the second characteristic information, The second user information Numerical representation of information characteristics of each information category; Using a cosine similarity function to match second feature information similar to the first initial feature information in a database, obtaining massage parameter information associated with massage modes corresponding to a plurality of successfully matched second feature information, and arranging and generating first reference massage information according to the massage modes includes: The cosine similarity between each of the second feature information and the first initial feature information is calculated and expressed by formula (6). Formula (6) is as follows: ； In formula (6), is the cosine similarity function, for The Euclidean norm of , for The Euclidean norm of , is the cosine similarity value; Determining one by one whether the cosine similarity values ​​corresponding to the second feature information are within a range of a preset similarity threshold; If yes, the second feature information is recorded as the first reference feature information, massage parameter information associated with the massage mode corresponding to the first reference feature information is obtained, the massage parameter information is mapped and stored with the first reference feature information, and the first reference massage information is obtained; If not, it means that the second feature information is currently unsuccessful in matching with the first initial feature information.

3. The control method of the graphene massager with multi-dimensional somatosensory feedback according to claim 1, characterized in that: Performing weighted averaging on massage parameters of the same massage point category in a plurality of first reference massage information, and generating second reference massage information according to the plurality of weighted averaged massage parameters comprises: The first reference massage information is represented by formula (7), and the multiple massage parameters in the first reference massage information are recorded as first massage parameters. The formula (7) is as follows: ； In formula (7), For the The first reference massage information, For the The first reference massage information The first massage parameter of each massage point category; The first massage parameters of the same massage point category in the plurality of first reference massage information are weighted averaged to obtain a weighted average massage parameter, which is recorded as a second massage parameter and is expressed by formula (8). Formula (8) is as follows: ； In formula (8), For the The second massage parameter after weighted average, , For the The weight of the first reference massage information, For the The first reference massage information The first massage parameter of each massage point category; The second massage parameters are sorted to obtain the second reference massage information, which is expressed by formula (9). The formula (9) is as follows: ； In formula (9), Massage information for the second reference.

4. The control method of the graphene massager with multi-dimensional somatosensory feedback according to claim 1, characterized in that: The first dimension feature is configured to be generated by the physiological feedback information and includes: Collecting pressure feedback information corresponding to massage position information within a preset time period, heart rate information of the current user within a preset time period, and skin conductivity information of the current user within a preset time period; Arranging the pressure feedback information in chronological order to obtain a pressure value change sequence, arranging the heart rate information in chronological order to obtain a heart rate value change sequence, and arranging the skin conductivity information in chronological order to obtain a skin conductivity change sequence; Calculating the pressure characteristics corresponding to the pressure feedback information of the current user according to the pressure value change sequence; Calculating the heart rate characteristics corresponding to the heart rate information of the current user according to the heart rate value change sequence; Calculating a skin conductivity feature corresponding to the skin conductivity information of the current user according to the skin conductivity change sequence; Concatenating the pressure feature, the heart rate feature, and the skin conductivity feature to obtain the first dimensional feature; The second dimension feature is configured to be generated by the manual adjustment information and includes: Collecting massage intensity adjustment information, massage temperature adjustment information and massage frequency adjustment information within a preset time period; Constructing a massage intensity adjustment curve according to the massage intensity adjustment information, obtaining a first change slope value of the massage intensity within a unit time according to the massage intensity adjustment curve, and generating a massage intensity adjustment feature according to the first change slope value; Constructing a massage temperature adjustment curve according to the massage temperature adjustment information, obtaining a second change slope value of the massage temperature in a unit time according to the massage temperature adjustment curve, and generating a massage temperature adjustment feature according to the second change slope value; Constructing a massage frequency adjustment curve according to the massage frequency adjustment information, obtaining a third change slope value of the massage frequency in unit time according to the massage frequency adjustment curve, and generating a massage frequency adjustment feature according to the third change slope value; The massage intensity adjustment feature, the massage temperature adjustment feature and the massage frequency adjustment feature are concatenated to obtain the second dimension feature.

5. The control method of the graphene massager with multi-dimensional somatosensory feedback according to claim 4, characterized in that: The third dimension feature is configured to be generated through the feedback evaluation information and includes: Collect multiple review texts within a preset time period, and perform the following steps on each review text: Performing sentence segment recognition on the review text to obtain a plurality of basic sentence segments; Inputting the basic sentence segment into a language recognition model to obtain a sentiment feature, wherein the sentiment feature includes sentiment polarity and sentiment intensity, wherein the sentiment polarity includes positive sentiment and negative sentiment, and the sentiment intensity includes a positive sentiment intensity value and a negative sentiment intensity value; Grouping the plurality of basic segments according to relevance to obtain at least one first segment group; Arranging the plurality of emotion features in the first sentence group one by one in chronological order to obtain a first emotion feature sequence of the current user, and constructing an emotion fluctuation curve according to the first emotion feature sequence; Calculating a fourth change slope value corresponding to each of the emotion fluctuation curves; splicing a plurality of the fourth change slope values ​​to obtain a second emotion feature sequence corresponding to the current review text; The third dimension feature is obtained by concatenating multiple second emotion feature sequences.

6. The control method of the graphene massager with multi-dimensional somatosensory feedback according to claim 1, characterized in that: The DQN algorithm model includes a DQN online network model and a DQN target network model; Inputting the multidimensional feature vector as an input feature into the DQN algorithm model, and obtaining the third reference massage information includes: The multi-dimensional feature vector is input into the DQN online network model and expressed by formula (16), which is as follows: ； In formula (16), is the DQN online network model, is the DQN online network model function, is the first The corresponding moment status, For the Action value prediction of the second massage parameter for each massage point category; The output result of the DQN online network model is input into the DQN target network model, and the prediction result of the maximum action value prediction of the second massage parameter in the future state is calculated as the third reference massage information, which is expressed by formula (17). The formula (17) is as follows: ； In formula (17), For the The prediction result of the maximum action value prediction of the second massage parameter of the massage point category, that is, The third massage parameter of each massage point category; Repeat the above steps until the generated The third massage parameters corresponding to the massage point categories are sorted to form the third reference massage information, which is expressed by formula (18). The formula (18) is as follows: ； In formula (18), The third reference massage information.

7. A graphene massager control system with multi-dimensional somatosensory feedback, characterized in that: The method according to any one of claims 1 to 6 comprises: An information collection module, used to obtain first user information, wherein the first user information includes information characteristics of five information categories: occupation information, work and rest information, diet information, massage location information and personal information of the current user; A massage parameter configuration module is used to generate first initial feature information according to the first user information, match second feature information in a database using a cosine similarity function, where the second feature information is generated when other users use a graphene massager of the same massage device category; obtain massage parameter information associated with massage modes corresponding to multiple successfully matched second feature information, sort and generate first reference massage information according to the massage mode; perform weighted averaging on massage parameters of the same massage point category in multiple first reference massage information, and generate second reference massage information according to multiple weighted averaged massage parameters; obtain the initial massage mode of the current user according to the second reference massage information and massage the current user; A feedback analysis module, used to collect user feedback information during the massage process, wherein the feedback information includes physiological feedback information, manual adjustment information and feedback evaluation information; The massage parameter configuration module is further used to generate first usage characteristic information according to the feedback information, and adjust the second reference massage information according to the first usage characteristic information to obtain third reference massage information; generate a usage massage mode according to the third reference massage information, and replace the current initial massage mode in the next batch of massages; The personal data management module is used to construct a personal usage database and store the feedback information of each batch of massages in the personal usage database in chronological order.

8. A computer-readable storage medium storing computer program instructions, characterized in that: The computer program instructions implement the method according to any one of claims 1 to 6 when executed by a processor.

Images