Artificial Intelligence-Based Elderly Turning and Back-Patting Assistance System

Abstract

The present invention discloses an artificial intelligence-based turning and back-patting assistance system for the elderly, including a data acquisition module, a data verification module, a module for constructing a turning and back-patting requirement prediction model, and a turning and back-patting assistance module. The present invention relates to the technical field of medical assistance, specifically an artificial intelligence-based turning and back-patting assistance system for the elderly. In this solution, abnormal data is cleared by defining enhanced weights, constructing an enhanced hybrid feature model, probability calculation, and parameter update, more accurately describing the distribution characteristics of the data, reducing the impact of noise data on the model, and improving the quality and reliability of the data; by introducing differential entropy and oscillation factors, designing an interleaved kernel, optimizing the objective function, performing parameter search and optimization, the complex relationships between data are more comprehensively described, enabling the model to more accurately capture the distribution characteristics and laws of the data, improving the efficiency and accuracy of parameter search, making the model parameters more appropriately set, and improving the accuracy of the model.

Description

Technical Field

[0001] The present invention relates to the field of medical assistance technologies, and specifically refers to an elderly turning and back-patting assistance system based on artificial intelligence. Background Art

[0002] The elderly turning and back-patting assistance system based on artificial intelligence uses artificial intelligence technology and data processing technology to construct a demand prediction model for the elderly's turning and back-patting needs, accurately predicts the needs, and provides timely turning and back-patting assistance services. However, traditional data verification methods have problems such as inaccurate verification of data related to the elderly's turning and back-patting needs and difficulty in effectively identifying abnormal data; traditional turning and back-patting demand prediction models have problems such as low prediction accuracy, difficulty in effectively capturing complex relationships and dynamic changes between data, and inappropriate model parameter settings. Summary of the Invention

[0003] In view of the above situation, to overcome the defects of the prior art, the present invention provides an elderly turning and back-patting assistance system based on artificial intelligence. Aiming at the problems of inaccurate verification of data related to the elderly's turning and back-patting needs and difficulty in effectively identifying abnormal data in traditional data verification methods, this solution clears abnormal data by constructing feature vectors, defining enhancement weights, constructing an enhanced hybrid feature model, probability calculation, and parameter update, more accurately describes the distribution characteristics of the data, improves the model's fitting ability and generalization ability for the data, reduces the influence of noise data on the model, and improves the quality and reliability of the data; aiming at the problems of low prediction accuracy, difficulty in effectively capturing complex relationships and dynamic changes between data, and inappropriate model parameter settings in traditional turning and back-patting demand prediction models, this solution more comprehensively describes the complex relationships between data by introducing differential entropy and oscillation factors, designing an interleaved kernel, optimizing the objective function, performing parameter search and optimization, enables the model to more accurately capture the distribution characteristics and laws of the data, improves the efficiency and accuracy of parameter search, makes the model parameter settings more appropriate, and improves the accuracy of the model.

[0004] The technical solution adopted by the present invention is as follows: The elderly turning and back-patting assistance system based on artificial intelligence provided by the present invention includes a data acquisition module, a data verification module, a module for constructing a turning and back-patting demand prediction model, and a turning and back-patting assistance module;

[0005] The data acquisition module acquires historical physiological data, environmental data, and turning and back-patting demand levels of the elderly; the physiological data includes heart rate, respiratory rate, body temperature, blood pressure, and bedridden time; the environmental data refers to the temperature, humidity, and noise intensity of the turning and back-patting environment; the turning and back-patting demand levels include two levels: no demand and having demand;

[0006] The data verification module clears abnormal data by constructing feature vectors, defining enhancement weights, constructing an enhanced hybrid feature model, calculating probabilities, and updating parameters;

[0007] The module for constructing a demand prediction model for turning over and back patting constructs a demand prediction model for turning over and back patting by setting label data, defining differential entropy, defining an oscillation factor, designing an interleaved kernel, determining a target, calculating the optimal hyperplane weight, calculating the optimal hyperplane bias, preparing for parameter optimization, generating a jump factor, generating an initial parameter point, defining a kinetic energy control factor, designing a parameter search function, introducing a global reset mechanism, and customizing start and stop strategies;

[0008] The turning over and back patting assistance module predicts the turning over and back patting demand level of the elderly through the turning over and back patting demand prediction model and assists the elderly in turning over and back patting.

[0009] Furthermore, the data verification module specifically includes the following:

[0010] Construct a feature vector by forming a feature vector from the feature data: heart rate, respiratory rate, body temperature, blood pressure, bedridden time, temperature, humidity, noise intensity, and turning over and back patting demand level;

[0011] Define the enhancement weight, expressed as follows:

[0012] ;

[0013] Where, represents the input feature vector, u and represent the indices of the Gaussian components, represents the enhancement weight of the input feature vector , e represents the natural constant, U represents the total number of Gaussian components, represents the mean of the u-th Gaussian component, represents the -th Gaussian component's mean, represents taking the modulus length;

[0014] Construct an enhanced hybrid feature model, expressed as follows:

[0015] ;

[0016] Where, represents the parameter set of the enhanced hybrid feature model, including: the mixing weight , the mean vector and the covariance matrix , represents the probability density of the enhanced hybrid feature model when the input feature vector is under the condition that the parameter is , Represents the input feature vector For a Gaussian distribution with a mean of and a variance of Probability density;

[0017] Probability calculation, expressed as follows:

[0018] ;

[0019] where g represents the index of the feature vector, represents the probability that the g-th feature vector comes from the u-th Gaussian component, represents the g-th feature vector;

[0020] Parameter update, updating the mixing weights, mean vectors, and covariance matrices, expressed as follows:

[0021] ;

[0022] where G represents the total number of feature vectors, represents the transpose operation;

[0023] Abnormal data cleaning, repeating probability calculation and parameter update until the parameters converge, setting a probability density threshold, calculating the probability density of each feature vector's enhanced mixture feature model, and setting the feature vectors with probability density lower than the probability density threshold as abnormal vectors and cleaning them up.

[0024] Furthermore, the module for constructing the prediction model for the need of turning over and back patting specifically includes the following:

[0025] Set the label data, and set the need level of turning over and back patting as the label data of the prediction model for the need of turning over and back patting;

[0026] Define the differential entropy, expressed as follows:

[0027] ;

[0028] where x1 and x2 represent the input feature vectors, represents the differential entropy between x1 and x2, i represents the dimension index of the feature vector, and respectively represent the probabilities obtained by mapping the differences between the values of the feature vectors x1 and x2 in the i-th dimension and the mean to the interval [0,1], represents the logarithmic function;

[0029] Define the oscillation factor, expressed as follows:

[0030] ;

[0031] where, represents the oscillation factor between the feature vectors x1 and x2, and respectively represent the second-order derivatives of the feature vectors x2 and x2 in the i-th dimension, and respectively represent the first-order derivatives of the feature vectors x2 and x2 in the i-th dimension, represents taking the absolute value, represents removing the zero factor;

[0032] Design the interleaving kernel, which is expressed as follows:

[0033] ;

[0034] where, represents the interleaving kernel between the feature vectors x1 and x2, D represents the maximum dimension of the feature vectors, represents the eigenvalue of the feature vector x1 in the i-th dimension, represents the eigenvalue of the feature vector x2 in the i-th dimension, represents the width parameter of the i-th dimension, and respectively represent the cross weight and the oscillation weight;

[0035] Determine the objective, which is expressed as follows:

[0036] ;

[0037] where, w is the hyperplane weight vector, represents the discrete weight vector of the i-th dimension, represents taking the maximum value, represents taking the square of the L2 norm, j and k represent the indices of the feature vectors, represents the adjustment weight, and respectively represent the Lagrange multipliers of the j-th and k-th feature vectors, and respectively represent the labels of the j-th and k-th feature vectors;

[0038] Calculate the optimal hyperplane weight, which is expressed as follows:

[0039] ;

[0040] where, represents the weight of the optimal hyperplane, represents the optimal solution of the Lagrange multiplier of the j-th feature vector, represents the feature mapping of the interleaving kernel for the j-th feature vector;

[0041] Calculate the optimal hyperplane bias, which is expressed as follows:

[0042] ;

[0043] Among them, represents the bias of the optimal hyperplane, represents the total number of support vectors, represents the feature vector index belonging to the support vector, represents the optimal solution of the Lagrange multiplier of the k-th feature vector;

[0044] Parameter optimization preparation: Set the accuracy of the turning and back patting demand prediction model as the performance value of the parameter individual, and determine the optimization parameters, including the width parameter, cross weight, oscillation weight, and adjustment weight;

[0045] Generate a jump factor, which is expressed as follows:

[0046] ;

[0047] Among them, q represents the number of times of generating the initial parameter point, represents the jump factor when generating the initial parameter point for the (q + 1)-th time, represents the jump factor when generating the initial parameter point for the q-th time, where the initial jump factor is a random number with a value range between 0 and 1;

[0048] Generate the initial parameter point, which is expressed as follows:

[0049] ;

[0050] Among them, represents the position of the initial parameter point generated for the (q + 1)-th time, represents the upper bound of the parameter space, represents the lower bound of the parameter space;

[0051] Define the kinetic energy control factor, which is expressed as follows:

[0052] ;

[0053] Among them, t represents the current parameter search times, represents the kinetic energy control factor during the t-th parameter search, represents the maximum parameter search times, represents a random number between 0 and 1, represents a random number between 0 and 2, represents the sign function, represents the average position of the generated initial parameter search points;

[0054] The design parameter search function is expressed as follows:

[0055] ;

[0056] Among them, represents the parameter position obtained from the (t + 1)-th parameter search, represents the parameter position obtained from the t-th parameter search, represents the position with the highest global parameter performance during the t-th parameter search, and r3 represents a random number with a value range between 0 and 1, represents the position with the lowest global parameter performance during the t-th parameter search;

[0057] The global reset mechanism is introduced and expressed as follows:

[0058] ;

[0059] Among them, represents the parameter position during the parameter search process, represents the parameter position after reset, represents taking the parameter position with the closest distance;

[0060] Customize the start-stop strategy, set the parameter performance threshold and the maximum number of parameter searches, and use the parameter search function to search for the initial parameter search point. During the search process, if the parameter performance at a parameter position is greater than the parameter performance threshold, stop the search and set the parameter at the position with the highest global parameter performance at this time as the model parameter; if the number of parameter searches reaches the maximum number of parameter searches, restart the search; otherwise, continue the search.

[0061] Furthermore, the turning and back-patting assistance module collects the physiological data and environmental data of the elderly in real time, inputs the data into the turning and back-patting demand prediction model, and the model predicts the turning and back-patting demand level of the elderly to assist the elderly in turning and back-patting in real time.

[0062] The beneficial effects achieved by the present invention using the above solution are as follows:

[0063] (1) Aiming at the problems of inaccurate inspection of data related to the turning and back-patting needs of the elderly and difficulty in effectively identifying abnormal data in traditional data inspection methods, this solution clears abnormal data by constructing feature vectors, defining enhanced weights, constructing an enhanced hybrid feature model, probability calculation, and parameter update, more accurately describes the distribution characteristics of the data, improves the fitting ability and generalization ability of the model to the data, reduces the influence of noise data on the model, and improves the quality and reliability of the data.

[0064] (2) Aiming at the problems of the traditional turning and back-patting demand prediction model, such as low prediction accuracy, difficulty in effectively capturing the complex relationships and dynamic changes between data, and inappropriate model parameter settings, this solution introduces differential entropy and oscillation factors, designs an interleaved kernel, optimizes the objective function, conducts parameter search and optimization, more comprehensively describes the complex relationships between data, enables the model to more accurately capture the distribution characteristics and laws of data, improves the efficiency and accuracy of parameter search, makes the model parameter settings more appropriate, and improves the accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a schematic diagram of the elderly turning and back-patting assistance system based on artificial intelligence provided by the present invention;

[0066] Figure 2 It is a schematic diagram of the data verification module;

[0067] Figure 3 It is a schematic diagram of the module for constructing the turning and back-patting demand prediction model;

[0068] Figure 4 It is a schematic diagram of the customized start-stop strategy.

[0069] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, but do not constitute a limitation to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0070] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0071] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation to the present invention.

[0072] Embodiment 1, refer to Figure 1 , the elderly turning and back-patting assistance system based on artificial intelligence provided by the present invention includes a data acquisition module, a data verification module, a module for constructing a turning and back-patting demand prediction model, and a turning and back-patting assistance module;

[0073] The data acquisition module collects historical physiological data, environmental data, and the level of turning and back-patting needs of the elderly, and sends the data to the data verification module;

[0074] The data verification module clears abnormal data by constructing feature vectors, defining enhancement weights, constructing an enhanced hybrid feature model, probability calculation, and parameter update, and sends the data to the module for constructing a turning and back-patting needs prediction model;

[0075] The module for constructing a turning and back-patting needs prediction model constructs a turning and back-patting needs prediction model by setting label data, defining differential entropy, defining an oscillation factor, designing an interleaved kernel, determining the target, calculating the optimal hyperplane weight, calculating the optimal hyperplane bias, preparing for parameter optimization, generating a jump factor, generating an initial parameter point, defining a kinetic energy control factor, designing a parameter search function, introducing a global reset mechanism, and customizing start-stop strategies, and sends the data to the turning and back-patting assistance module;

[0076] The turning and back-patting assistance module predicts the level of turning and back-patting needs of the elderly through the turning and back-patting needs prediction model and assists the elderly in turning and back-patting.

[0077] Embodiment 2, refer to Figure 1 and Figure 2 , based on the above embodiment, the data verification module specifically includes the following content:

[0078] Construct a feature vector by forming a feature vector from the feature data: heart rate, respiratory rate, body temperature, blood pressure, lying time, temperature, humidity, noise intensity, and the level of turning and back-patting needs;

[0079] Define the enhancement weight, which is expressed as follows:

[0080] ;

[0081] Among them, represents the input feature vector, u and represent the indices of Gaussian components, represents the enhancement weight of the input feature vector , e represents the natural constant, U represents the total number of Gaussian components, represents the mean of the u-th Gaussian component, represents the -th Gaussian component mean, represents taking the modulus length;

[0082] Construct an enhanced hybrid feature model, which is expressed as follows:

[0083] ;

[0084] Among them, Denote the parameter set of the enhanced mixture feature model, including: mixture weights , mean vectors and covariance matrices , Denote the input feature vector The probability density of the enhanced mixture feature model under the condition that the parameter is , Denote the input feature vector For a Gaussian distribution with a mean of and a variance of ;

[0085] Probability calculation, which is expressed as follows:

[0086] ;

[0087] Among them, g represents the index of the feature vector, Denote the probability that the g-th feature vector comes from the u-th Gaussian component, Denote the g-th feature vector;

[0088] Parameter update, update the mixture weights, mean vectors and covariance matrices, which is expressed as follows:

[0089] ;

[0090] Among them, G represents the total number of feature vectors, Denote the transpose operation;

[0091] Abnormal data cleaning, repeat probability calculation and parameter update until the parameters converge, set the probability density threshold, calculate the probability density of the enhanced mixture feature model of each feature vector, and set the feature vectors with probability density lower than the probability density threshold as abnormal vectors and clean them up.

[0092] By performing the above operations, aiming at the problems that the traditional data inspection method has inaccurate inspection of data related to the need for turning over and back patting of the elderly and is difficult to effectively identify abnormal data, this solution clears abnormal data by constructing feature vectors, defining enhanced weights, constructing an enhanced mixture feature model, probability calculation and parameter update, more accurately describes the distribution characteristics of the data, improves the fitting ability and generalization ability of the model to the data, reduces the influence of noise data on the model, and improves the quality and reliability of the data.

[0093] Example 3, refer to Figure 1 , Figure 3 and Figure 4 , this example is based on the above example, and the module for constructing a turning over and back patting demand prediction model specifically includes the following contents:

[0094] Set label data, and set the turning and back-patting requirement level as the label data of the turning and back-patting requirement prediction model;

[0095] Define differential entropy, which is expressed as follows:

[0096] ;

[0097] where x1 and x2 represent input feature vectors, represents the differential entropy between x1 and x2, i represents the dimension index of the feature vector, and respectively represent the probabilities obtained by mapping the differences between the values of the feature vectors x1 and x2 in the i-th dimension and the mean to the interval [0, 1], represents the logarithmic function;

[0098] Define the oscillation factor, which is expressed as follows:

[0099] ;

[0100] where, represents the oscillation factor between the feature vectors x1 and x2, and respectively represent the second-order derivatives of the feature vectors x2 and x2 in the i-th dimension, and respectively represent the first-order derivatives of the feature vectors x2 and x2 in the i-th dimension, represents taking the absolute value, represents removing the zero factor;

[0101] Design the interleaving kernel, which is expressed as follows:

[0102] ;

[0103] where, represents the interleaving kernel between the feature vectors x1 and x2, D represents the maximum dimension of the feature vector, represents the eigenvalue of the feature vector x1 in the i-th dimension, represents the eigenvalue of the feature vector x2 in the i-th dimension, represents the width parameter of the i-th dimension, and respectively represent the cross weight and the oscillation weight;

[0104] Determine the objective, which is expressed as follows:

[0105] ;

[0106] where w is the hyperplane weight vector, represents the discrete weight vector of the i-th dimension, denotes taking the maximum value, denotes taking the square of the L2 norm, where j and k represent the indices of the eigenvectors, denotes adjusting the weight, and denote the Lagrange multipliers of the j-th and k-th eigenvectors respectively, and denote the labels of the j-th and k-th eigenvectors respectively;

[0107] Calculate the optimal hyperplane weight, which is expressed as follows:

[0108] ;

[0109] where, denotes the weight of the optimal hyperplane, denotes the optimal solution of the Lagrange multiplier of the j-th eigenvector, denotes the feature mapping of the intertwining kernel for the j-th eigenvector;

[0110] Calculate the optimal hyperplane bias, which is expressed as follows:

[0111] ;

[0112] where, denotes the bias of the optimal hyperplane, denotes the total number of support vectors, denotes the index of the eigenvector belonging to the support vector, denotes the optimal solution of the Lagrange multiplier of the k-th eigenvector;

[0113] Prepare for parameter optimization. Set the accuracy of the turning and back patting demand prediction model as the performance value of the parameter individual, and determine the optimization parameters, including the width parameter, cross weight, oscillation weight, and adjustment weight;

[0114] Generate a jump factor, which is expressed as follows:

[0115] ;

[0116] where q represents the number of times of generating the initial parameter point, denotes the jump factor when generating the initial parameter point for the (q + 1)-th time, denotes the jump factor when generating the initial parameter point for the q-th time, where the initial jump factor is a random number with a value range between 0 and 1;

[0117] Generate the initial parameter point, which is expressed as follows:

[0118] ;

[0119] Among them, represents the position of the initial parameter point generated at the (q + 1)-th time, represents the upper bound of the parameter space, represents the lower bound of the parameter space;

[0120] Define the kinetic energy control factor, which is expressed as follows:

[0121] ;

[0122] Among them, t represents the current parameter search times, represents the kinetic energy control factor during the t-th parameter search, represents the maximum parameter search times, represents a random number between 0 and 1, represents a random number between 0 and 2, represents the sign function, represents the average position of the generated initial parameter search points;

[0123] Design the parameter search function, which is expressed as follows:

[0124] ;

[0125] Among them, represents the parameter position obtained from the (t + 1)-th parameter search, represents the parameter position obtained from the t-th parameter search, represents the position with the highest global parameter performance during the t-th parameter search, r3 represents a random number with a value range between 0 and 1, represents the position with the lowest global parameter performance during the t-th parameter search;

[0126] Introduce the global reset mechanism, which is expressed as follows:

[0127] ;

[0128] Among them, represents the parameter position during the parameter search process, represents the parameter position after reset, represents taking the parameter position closest in distance;

[0129] Customize the start-stop strategy, set the parameter performance threshold and the maximum number of parameter searches, use the parameter search function to perform parameter search on the initial parameter search points. During the search process, if there is a parameter position whose parameter performance is greater than the parameter performance threshold, stop the search and set the parameter at the position with the highest global parameter performance at this time as the model parameter; if the number of parameter searches reaches the maximum number of parameter searches, restart the search; otherwise, continue the search.

[0130] By performing the above operations, aiming at the problems existing in the traditional turning and back-patting demand prediction model, such as low prediction accuracy, difficulty in effectively capturing the complex relationships and dynamic changes between data, and inappropriate model parameter settings, this solution improves the accuracy of the model by introducing differential entropy and oscillation factors, designing an interleaved kernel, optimizing the objective function, performing parameter search and optimization, more comprehensively describing the complex relationships between data, enabling the model to more accurately capture the distribution characteristics and laws of data, improving the efficiency and accuracy of parameter search, making the model parameter settings more appropriate.

[0131] Example 4, refer to Figure 1 , based on the above example, the turning and back-patting assistance module collects the physiological data and environmental data of the elderly in real time, inputs the data into the turning and back-patting demand prediction model, and the model predicts the turning and back-patting demand level of the elderly, and assists the elderly in turning and back-patting in real time.

[0132] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0133] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention.

[0134] The above describes the present invention and its embodiments, and this description is not restrictive. What is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. All in all, if those of ordinary skill in the art are inspired by it and design similar structural ways and embodiments without creative efforts without departing from the purpose of the present invention, they shall fall within the protection scope of the present invention.

Claims

1. An artificial intelligence-based elderly turning-over and back-patting assistance system, characterized by: It includes data collection module, data verification module, module for building prediction model of turning over and patting back demand and turning over and patting back auxiliary module; The data collection module collects historical physiological data, environmental data and the level of demand for turning over and patting the back of the elderly; The data verification module removes abnormal data by constructing feature vectors, defining enhancement weights, constructing enhanced hybrid feature models, probability calculations, and parameter updates; The module for constructing a prediction model for the demand for turning over and patting the back is constructed by setting label data, defining difference entropy, defining oscillation factors, designing interleaving kernels, determining targets, calculating optimal hyperplane weights, calculating optimal hyperplane biases, preparing parameter optimization, generating jump factors, generating initial parameter points, defining kinetic energy control factors, designing parameter search functions, introducing global reset mechanisms, and customizing start-stop strategies to construct a prediction model for the demand for turning over and patting the back; The definition of difference entropy is expressed as follows: ； Among them, x1 and x2 represent the input feature vectors, represents the difference entropy between x1 and x2, i represents the dimension index of the feature vector, and They represent the probability of mapping the difference between the value of the feature vector x1 and x2 in the i-th dimension and the mean to the interval [0,1], represents the logarithmic function; The turning over and back-patting auxiliary module predicts the turning over and back-patting demand level of the elderly through the turning over and back-patting demand prediction model, and assists the elderly in turning over and back-patting.

2. The artificial intelligence-based elderly back-patting assistance system according to claim 1 is characterized by: The module for predicting the demand for turning over and patting the back includes the following contents: Set label data, and set the turning over and patting back demand level as label data of the turning over and patting back demand prediction model; Define differential entropy; Define the oscillation factor as follows: ； in, represents the oscillation factor between eigenvectors x1 and x2, and Respectively represent the second-order derivatives of the eigenvectors x2 and x2 in the i-th dimension, and Represent the first-order derivatives of the eigenvectors x2 and x2 in the i-th dimension, respectively. Indicates taking the absolute value, represents the zero removal factor; Design the interleaving kernel, expressed as follows: ； in, represents the interleaved kernel between feature vectors x1 and x2, D represents the maximum dimension of the feature vector, represents the eigenvalue of the eigenvector x1 in the i-th dimension, represents the eigenvalue of the eigenvector x2 in the i-th dimension, Represents the width parameter of the i-th dimension, and denote the cross-weight and oscillation weight respectively; Determine the goal, expressed as follows: ； Where w is the hyperplane weight vector, represents the discrete weight vector of the i-th dimension, Indicates taking the maximum value, represents the square of the L2 norm, j and k represent the index of the feature vector, represents the adjustment weight, and denote the Lagrange multipliers of the j-th and k-th eigenvectors, respectively, and Represent the labels of the j-th and k-th feature vectors respectively; Calculate the optimal hyperplane weight, expressed as follows: ； in, represents the weight of the optimal hyperplane, represents the optimal solution of the Lagrange multiplier of the jth eigenvector, represents the feature mapping of the interleaved kernel for the jth eigenvector; Calculate the optimal hyperplane bias, expressed as follows: ； in, represents the bias of the optimal hyperplane, represents the total number of support vectors, denotes the eigenvector index belonging to the support vector, represents the optimal solution of the Lagrange multiplier of the kth eigenvector; Parameter optimization preparation, setting the accuracy of the turning over and patting back demand prediction model as the performance value of the parameter individual, and determining the optimization parameters, including width parameter, cross weight, oscillation weight and adjustment weight; Generate the jump factor, expressed as follows: ； Where q represents the number of times the initial parameter points are generated. Indicates the jump factor when generating the initial parameter point for the q+1th time, Indicates the jump factor when generating the initial parameter point for the qth time, where the initial jump factor is a random number ranging from 0 to 1; Generate initial parameter points, expressed as follows: ； in, represents the position of the initial parameter point generated for the q+1th time, represents the upper bound of the parameter space, represents the lower bound of the parameter space; The kinetic energy control factor is defined as follows: ； Among them, t represents the number of current parameter searches, represents the kinetic energy control factor during the t-th parameter search, Indicates the maximum number of parameter searches, Represents a random number between 0 and 1. represents a random number between 0 and 2. represents the symbolic function, represents the average position of the generated initial parameter search points; Design parameter search function, expressed as follows: ； in, Indicates the parameter position obtained by the t+1th parameter search, represents the parameter position obtained by the tth parameter search, represents the position with the highest global parameter performance during the tth parameter search, r3 represents a random number ranging from 0 to 1, Indicates the position with the lowest global parameter performance during the t-th parameter search; A global reset mechanism is introduced, which is expressed as follows: ； in, Indicates the parameter position during the parameter search process, Indicates the parameter position after reset, Indicates taking the parameter position with the closest distance; Customize the start-stop strategy, set the parameter performance threshold and the maximum number of parameter searches, and use the parameter search function to perform parameter search on the initial parameter search point. During the search process, if there is a parameter position with parameter performance greater than the parameter performance threshold, stop the search and set the parameter at the position with the highest global parameter performance at this time as the model parameter; if the parameter search number reaches the maximum number of parameter searches, search again; otherwise, continue searching.

3. The artificial intelligence-based elderly back-patting assistance system according to claim 1 is characterized by: The data verification module specifically includes the following contents: Construct a feature vector by combining the feature data: heart rate, respiratory rate, body temperature, blood pressure, bed rest time, temperature, humidity, noise intensity, and the level of need for turning over and patting the back; Define the enhancement weight as follows: ； in, represents the input feature vector, u and represents the index of the Gaussian component, Represents the input feature vector The enhancement weight, e represents the natural constant, U represents the total number of Gaussian components, represents the mean of the u-th Gaussian component, Indicates The mean of the Gaussian components, Indicates the modulus length; Construct an enhanced mixed feature model, which is expressed as follows: ； in, Represents the parameter set of the enhanced hybrid feature model, including: hybrid weight , mean vector and the covariance matrix , The feature vector representing the input In the parameter The probability density of the enhanced mixed feature model under the condition of, The feature vector representing the input For a mean , the variance is The probability density of the Gaussian distribution of The probability calculation is expressed as follows: ； Among them, g represents the index of the feature vector, represents the probability that the g-th eigenvector comes from the u-th Gaussian component, represents the g-th eigenvector; Parameter update, update the mixing weights, mean vector and covariance matrix, expressed as follows: ； Where G represents the total number of eigenvectors, Represents the transpose operation; Abnormal data is cleared, probability calculation and parameter update are repeated until the parameters converge, the probability density threshold is set, the probability density of the enhanced mixed feature model of each feature vector is calculated, and the feature vector with probability density lower than the probability density threshold is set as an abnormal vector and cleared.

4. The artificial intelligence-based elderly back-patting assistance system according to claim 1 is characterized by: The data acquisition module collects historical physiological data, environmental data and the level of demand for turning over and patting the back of the elderly; the physiological data includes heart rate, respiratory rate, body temperature, blood pressure and bed rest time; the environmental data refers to the temperature, humidity and noise intensity of the environment for turning over and patting the back; the level of demand for turning over and patting the back includes two levels: no demand and demand.

5. The artificial intelligence-based elderly back-patting assistance system according to claim 1 is characterized by: The turning over and back-patting auxiliary module collects the elderly's physiological data and environmental data in real time, and inputs the data into the turning over and back-patting demand prediction model. The model predicts the elderly's turning over and back-patting demand level, and assists the elderly in turning over and back-patting in real time.

Images