Rehabilitation interaction strategy optimization method of gamification digital human and intelligent chip

Abstract

The invention discloses a rehabilitation interaction strategy optimization method of a gamification digital human and an intelligent chip, and relates to the field of artificial intelligence, and the method comprises the steps: collecting rehabilitation features and user features of a target user, and configuring an interaction standard of the target user according to the rehabilitation features; carrying out standard weight and interaction weight distribution according to the rehabilitation features and the user features; calculating the difference among the first standard weight, the first interaction weight, the second standard weight and the second interaction weight to obtain a weight difference coefficient; according to the weight difference coefficient, the first standard weight, the first interaction weight, the second standard weight and the second interaction weight, in combination with the weight difference coefficient, rehabilitation interaction strategy optimization of the gamification digital human is carried out, an optimal rehabilitation interaction strategy is obtained, and interaction processing is carried out in combination with an interaction standard. The problems of insufficient individuation depth, inaccurate strategy adjustment and poor dynamic adaptability in a rehabilitation interaction strategy are solved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to a method for optimizing rehabilitation interaction strategies for gamified digital humans and an intelligent chip. Background Technology

[0002] With the development of artificial intelligence and virtual reality technologies, gamified digital humans, or virtual rehabilitation therapists, are being widely used to assist patients in their rehabilitation training.

[0003] Current interaction strategies are generally fixed, lack dynamic adaptability, have a single evaluation dimension, ignore user experience, affect user compliance and enthusiasm for rehabilitation interaction, and lack personalization depth, resulting in problems such as insufficient personalization depth, inaccurate strategy adjustment, and poor dynamic adaptability in rehabilitation interaction strategies.

[0004] In summary, there is a need for a method that can deeply integrate objective rehabilitation data with subjective user status, achieve real-time dynamic optimization of interaction strategies, and intelligently resolve conflicts between rehabilitation goals and user experience. Summary of the Invention

[0005] This application provides a gamified digital human rehabilitation interaction strategy optimization method and intelligent chip, aiming to solve the problems of insufficient personalization depth in the prior art, resulting in a lack of precise adjustment of rehabilitation interaction strategies, inability to achieve dynamic optimal balance according to real-time situation, and poor dynamic adaptability.

[0006] In view of the above problems, this application provides a method for optimizing rehabilitation interaction strategies for gamified digital humans and an intelligent chip.

[0007] Firstly, this application provides a method for optimizing rehabilitation interaction strategies for gamified digital humans, including: Collect the rehabilitation characteristics and user characteristics of the target users, and configure the interaction standards of the target users based on the rehabilitation characteristics; Based on the rehabilitation characteristics, standard weights and interaction weights are allocated to obtain a first standard weight and a first interaction weight. Based on the user characteristics, standard weights and interaction weights are allocated to obtain a second standard weight and a second interaction weight. Calculate the differences between the first standard weight, the first interaction weight, the second standard weight, and the second interaction weight to obtain the weight difference coefficient; Based on the weight difference coefficient, the first standard weight, the first interaction weight, the second standard weight, and the second interaction weight, and combined with the weight difference coefficient, the rehabilitation interaction strategy of the gamified digital human is optimized to obtain the optimal rehabilitation interaction strategy. The interaction is then processed in conjunction with the interaction standard, wherein the optimal rehabilitation interaction strategy includes the optimal interaction qualification threshold.

[0008] Secondly, this application provides a smart chip for optimizing rehabilitation interaction strategies for gamified digital humans. The smart chip stores a first computer program, which, when executed by a processor, implements a method for optimizing rehabilitation interaction strategies for gamified digital humans as described in Embodiment 1.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application provides a method for optimizing rehabilitation interaction strategies for gamified digital humans and an intelligent chip. First, it continuously collects and analyzes users' objective rehabilitation annotation values ​​and subjective emotion annotation values, achieving real-time data acquisition for rehabilitation interaction strategies. Second, it assigns standard weights and interaction weights to rehabilitation features and user features, quantifying the difference between rehabilitation needs and user capabilities, providing structured input for subsequent decision-making, and solving the problem of single evaluation dimensions in traditional systems. Third, it performs weighted calculations based on the fused standard weights and interaction weights to obtain a weight difference coefficient, outputting a local optimum solution, providing an intelligent multi-objective conflict fusion scheme to improve rehabilitation effectiveness. Finally, it optimizes the rehabilitation interaction strategies of the gamified digital human based on the weight difference coefficient, the first standard weight, the first interaction weight, the second standard weight, and the second interaction weight, dynamically adjusting thresholds and providing a quantifiable and assessable optimization process. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 A flowchart illustrating an optimization method for a gamified digital human rehabilitation interaction strategy; Figure 2 This is a schematic diagram of the structure of the smart chip provided in an embodiment of this application.

[0012] The labels in the attached diagram are explained as follows: Intelligent chip 200, first computer program 211. Detailed Implementation

[0013] This application provides a gamified digital human rehabilitation interaction strategy optimization method and a smart chip to address the problems in existing technologies where insufficient personalization depth leads to a lack of precise adjustment of rehabilitation interaction strategies, making it impossible to achieve dynamic optimal balance based on real-time situations and resulting in poor dynamic adaptability.

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0015] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0016] Example 1, as Figure 1 As shown, this application provides a method for optimizing rehabilitation interaction strategies for gamified digital humans and a smart chip, the method comprising: S10: Collect the rehabilitation characteristics and user characteristics of the target user, and configure the interaction standards of the target user based on the rehabilitation characteristics; In this embodiment, the target user refers to a patient undergoing rehabilitation training or an individual requiring rehabilitation assistance; rehabilitation characteristics are a set of parameters used to quantify the user's rehabilitation status and progress. These may include rehabilitation label values, which reflect the user's rehabilitation compliance and objective progress rate; user characteristics are a set of parameters used to describe the user's current subjective state, including user emotion label values; interaction standards refer to the standard requirements for interactive actions when the gamified digital human interacts with the user, which can be set according to the user's rehabilitation progress and disease type.

[0017] Specifically, the system collects users' rehabilitation characteristics and user features, and uses these objective rehabilitation characteristics to initially set interaction standards, providing an initial objective benchmark for subsequent interaction evaluation.

[0018] Step S10 in the method provided in this application embodiment includes: Collect the rehabilitation characteristics and user characteristics of the target users. The rehabilitation characteristics include rehabilitation type and rehabilitation label value. The rehabilitation label value is the ratio of actual rehabilitation time to planned rehabilitation time. The user characteristics include user emotion label value. Based on the rehabilitation characteristics, the interaction standards for the target user are configured, wherein the interaction standards include interaction action standards.

[0019] In this embodiment of the application, the rehabilitation characteristics and user characteristics of the target user are first collected. The rehabilitation characteristics include rehabilitation type and rehabilitation label value. The rehabilitation label value is the ratio of actual rehabilitation time to planned rehabilitation time. The user characteristics include user emotion label value.

[0020] Among them, rehabilitation type refers to the specific category or target area of ​​rehabilitation training required by the user. It is a qualitative description of the rehabilitation task and determines the core actions of the subsequent interactive game; rehabilitation label value is a key, quantitative progress indicator that reflects the user's rehabilitation compliance and stage completion; user emotion label value is a parameter used to quantify the user's subjective psychological state, which can be obtained and normalized into a standard value through various technical means.

[0021] Specifically, the process involves collecting rehabilitation characteristics and user features of the target users, including their rehabilitation type, rehabilitation label values, and user emotional label values. Rehabilitation types can include upper limb motor function rehabilitation, lower limb balance training, fine motor skills rehabilitation, speech and cognitive training, and spinal joint range of motion training. Rehabilitation label values ​​are the ratio of actual rehabilitation time to planned rehabilitation time. Actual rehabilitation time is the cumulative time it takes for the target user to effectively complete training movements that meet basic requirements within a single training cycle; planned rehabilitation time is the ideal training duration preset for the target user according to the rehabilitation plan.

[0022] In addition, the user's emotion label value is labeled according to the user's emotional reaction during the usual nursing and rehabilitation process. The more positive the emotion, the larger the user's emotion label value, ranging from 0 to 1. The larger the value, the more positive the emotion.

[0023] For example, through data collection, it is found that target user A is expected to recover in 30 days, and is currently 15 days old. Therefore, the recovery label value is 15 / 30=50%. Taking the rehabilitation type of upper limb motor function rehabilitation as an example in this application, the user's emotional reaction during the rehabilitation process is labeled, and the user's emotional label value is 0.3.

[0024] Secondly, based on the rehabilitation characteristics, the interaction standards for the target users are configured, wherein the interaction standards include interaction action standards.

[0025] Among them, the interaction action standard is the core part of the interaction standard. In a specific interaction round, the user's input must meet certain conditions, and only those interactions that include the interaction action standard can be judged as successful interactions.

[0026] Specifically, based on the collected rehabilitation features, specific interaction criteria are initialized, and the interaction action standards corresponding to the rehabilitation features are extracted from the historical data of rehabilitation types based on the rehabilitation features.

[0027] For example, in upper limb motor function rehabilitation, the standard for interactive actions of target user A may be that the user's arm abduction angle in the sagittal plane must be greater than or equal to 45 degrees, and the action must be held for more than 1.5 seconds.

[0028] In step S10 of the method provided in this application embodiment, configuring the interaction standards of the target user according to the rehabilitation characteristics includes: A rehabilitation interaction standard generator is obtained, wherein the rehabilitation interaction standard generator is constructed based on machine learning and is trained using a sample rehabilitation feature set and a sample interaction standard set; The rehabilitation features are input into the rehabilitation interaction standard generator, and the interaction standard is output.

[0029] In this embodiment of the application, the rehabilitation characteristics and user characteristics of the target user are first collected. The rehabilitation characteristics include rehabilitation type and rehabilitation label value. The rehabilitation label value is the ratio of actual rehabilitation time to planned rehabilitation time. The user characteristics include user emotion label value.

[0030] The rehabilitation interaction standard generator is a trained machine learning model that receives the user's rehabilitation features as input and outputs a matching interaction standard. The sample rehabilitation feature set is the input data used to train the generator. It is a dataset composed of historical data, where each sample contains a rehabilitation feature vector of a historical user. The sample interaction standard set is the dataset corresponding to the sample rehabilitation feature set.

[0031] Specifically, by acquiring historical data from different users' medical visits, a set of sample rehabilitation features and a set of sample interaction standards corresponding to the rehabilitation features are constructed. These two sets of sample rehabilitation features and sample interaction standards are then used as input to construct a rehabilitation interaction standard generator through machine learning.

[0032] For example, the collected sample rehabilitation feature set can be: {[upper limb motor function rehabilitation, rehabilitation label value: 60%], ...}, and the sample interaction standard set can be: {[angle > 40°, action holding time > 1s], ...}.

[0033] In this embodiment, a backpropagation (BP) neural network is used to construct a rehabilitation interaction standard generator. Besides the BP neural network, other machine learning models suitable for regression prediction can also be used to construct rehabilitation interaction standard generators. The BP neural network model is a feedforward neural network trained through error backpropagation and is commonly used to predict continuous values.

[0034] For example, the steps to construct a rehabilitation interaction standard generator model based on a BP neural network are as follows: First, the sample rehabilitation feature set and sample interaction standard set are used as inputs and divided into training set, validation set and test set in a ratio of 7:2:1.

[0035] Secondly, the model is constructed, mainly consisting of an input layer, a hidden layer, and an output layer. The input layer takes in a set of sample rehabilitation features and a set of sample interaction criteria; the hidden layer uses an activation function to perform a non-linear transformation on these two sets of features; and the output layer outputs the interaction criteria.

[0036] Finally, model training involves setting the initial learning rate and weights, calculating the error between the predicted results and the sample rehabilitation feature set and sample interaction standard set using the mean squared error function, adjusting the weights and calculating the error, iterating and repeating until the error is minimized, generating parameters through forward propagation and updating them through backpropagation, and evaluating performance on a validation set after each training epoch to avoid overfitting. The model is considered successful when the MSE loss decreases by less than 1e over five consecutive training epochs. -5 When the MSE loss on the validation set stabilizes below 0.01, the model is considered converged, and the rehabilitation interaction standard generator is obtained.

[0037] Finally, the rehabilitation features are input into the rehabilitation interaction standard generator, and the interaction standard is output.

[0038] The rehabilitation features are the rehabilitation characteristics collected or acquired in real time from the current target user and input into the rehabilitation interaction standard generator. The rehabilitation interaction standard generator outputs a predicted interaction standard through a series of complex weighted calculations and nonlinear transformations.

[0039] By analyzing and predicting the rehabilitation characteristics of different target users, an interaction standard adapted to the target user can be obtained. If a target user has a low rehabilitation label value, it indicates that the target user's rehabilitation training is progressing poorly, and the interaction standard is increased accordingly to ensure the rehabilitation effect. If a target user has a high rehabilitation label value, it indicates that the target user's rehabilitation training is progressing well, and the interaction standard is decreased accordingly to reduce rehabilitation pressure.

[0040] For example, if the target user A is input: [Upper limb motor function rehabilitation, rehabilitation label value: 50%], the interaction standard is: [Angle > 45°, movement holding time > 1.5s].

[0041] In this embodiment, personalized initial interaction thresholds are provided for users of different rehabilitation stages and types by setting interaction standards. The main rehabilitation types are considered, and multi-dimensional features such as rehabilitation progress and user basic information are integrated. This reduces manual intervention, improves the efficiency and consistency of system deployment, ensures the safety and basic relevance of training, and realizes the automated and highly accurate personalized generation of initial interaction standards.

[0042] S20: Based on the rehabilitation characteristics, standard weights and interaction weights are allocated to obtain a first standard weight and a first interaction weight; based on the user characteristics, standard weights and interaction weights are allocated to obtain a second standard weight and a second interaction weight. In this embodiment, the standard weight refers to the coefficient used to measure the standard of interactive movements when evaluating rehabilitation effects or developing strategies. The higher the weight, the more emphasis is placed on hard indicators such as the accuracy and range of motion of the movements; the interaction weight refers to the importance coefficient of factors other than the standard during the interaction process when evaluating rehabilitation effects or developing strategies.

[0043] The first standard weight and the first interaction weight are weights derived from rehabilitation characteristics, representing the tendency to encourage interaction from the perspective of the objective rehabilitation process; the second standard weight and the second interaction weight are weight pairs derived from user characteristics, representing the tendency to encourage interaction from the perspective of the user's subjective state.

[0044] Specifically, the importance weights for achieving the interaction standard and the user's emotions during the interaction are assigned based on the rehabilitation characteristics. If the interaction does not meet the standard, the digital human's feedback may be considered an interaction failure, which could negatively impact the user's motivation for rehabilitation.

[0045] The more complete the recovery progress, the higher the standard weight, and the stronger the tendency to adhere to the interaction standard, thus improving the recovery effect. Conversely, the higher the emotion label value within the user characteristics, the lower the standard weight, the higher the interaction weight, and the stronger the tendency to engage in interaction, thus taking into account the user's emotions and promoting the recovery progress.

[0046] Step S20 in the method provided in this application embodiment includes: Obtain the rehabilitation label value within the rehabilitation feature; The ratio of the stated rehabilitation label value to the preset rehabilitation label value is calculated and used as the first standard weight. The first interaction weight is calculated based on the first standard weight.

[0047] In this embodiment of the application, the rehabilitation label value within the rehabilitation feature is first obtained.

[0048] Specifically, the system retrieves the corresponding rehabilitation label values ​​within the target user's current rehabilitation characteristics from stored or real-time calculated data structures. It then calculates the rehabilitation progress data for the first pair of weights, ensuring that the first standard weight and the first interactive weight represent a tendency from the perspective of rehabilitation effectiveness, unaffected by subjective factors.

[0049] For example, if user A extracts a rehabilitation annotation value of 50%, it means that the actual training time is only half of the planned time.

[0050] Secondly, the ratio of the rehabilitation label value to the preset rehabilitation label value is calculated and used as the first standard weight.

[0051] Among them, the preset rehabilitation standard value is a predefined benchmark value, which is usually set by rehabilitation medicine experts based on general principles.

[0052] Specifically, the first standard weight is calculated as: Rehabilitation Standard Value / Preset Rehabilitation Standard Value. If the ratio is greater than 1, it indicates that the user is ahead of schedule and should be given a higher first standard weight, with a greater inclination to maintain or even improve the precision of the movements. Conversely, if the ratio is less than 1, it indicates that the user is behind schedule and may be facing difficulties or lacking motivation. In this case, a lower first standard weight should be given, meaning that based on the objective reality of rehabilitation, the demand for absolute standards should be appropriately reduced to avoid the user giving up completely due to prolonged inability to meet the standards.

[0053] For example, the preset rehabilitation label value is set to 80% based on the rehabilitation stage, representing a higher tolerance for early rehabilitation. User A's ratio is 50% / 80% = 0.625. Therefore, User A's first standard weight is 0.625.

[0054] Finally, the first interaction weight is calculated based on the first standard weight.

[0055] Specifically, the first interaction weight and the first standard weight are negatively correlated and their sum is 1. Therefore, the first interaction weight is 1 minus the first standard weight, i.e., the first interaction weight = 1 - the first standard weight.

[0056] For example, the weight of user A's first interaction is 1 - 0.635 = 0.375.

[0057] In step S20 of the method provided in this application embodiment, standard weights and interaction weights are allocated according to the user characteristics to obtain second standard weights and second interaction weights, including: Obtain the user sentiment annotation value within the user features; The ratio of the user emotion label value to the preset user emotion label value is calculated and used as the second standard weight; The second interaction weight is calculated based on the second standard weight.

[0058] In this embodiment of the application, the user emotion label value within the user features is first obtained.

[0059] Specifically, the system first extracts the target user's emotional state label values ​​from continuously monitored user characteristic data within the current or most recent time window. This is separated from objective rehabilitation goals, using the emotional state label values ​​to reflect the subjective user experience. This allows for immediate responses to changes in the target user's internal state, rather than relying solely on delayed objective progress reports.

[0060] For example, user A's sentiment label value is 0.3.

[0061] Secondly, the ratio of the user's sentiment label value to the preset user sentiment label value is calculated as the second standard weight.

[0062] Among them, the preset user emotion label value is a predefined value that represents the user's emotional state in an ideal state. It can be obtained by calculating the historical average value and represents the emotional level that can maintain rehabilitation training.

[0063] Specifically, the second standard weight is calculated as: user emotion label value / preset user emotion label value. If the user emotion value is greater than the preset user emotion label value, it indicates that the user is in a good emotional state and can accept higher requirements, so the second standard weight should be higher. Conversely, if the user emotion value is low, it indicates that the user is in a fragile emotional state and the requirements should be lowered to ensure user experience, so the second standard weight should be lower.

[0064] For example, if the historical average value is calculated to obtain a preset user emotion label value of 0.8, which represents an idealized user emotion state, then the ratio of user A's user emotion label value to the preset user emotion label value is 0.3 / 0.8 = 0.375, and the second standard weight of user A is 0.375.

[0065] Finally, the second interaction weight is calculated based on the second standard weight.

[0066] Specifically, the second standard weight and the second interaction weight are negatively correlated and their sum is 1. Therefore, the second interaction weight is 1 - the second standard weight, that is, the second interaction weight = 1 - the second standard weight.

[0067] For example, the second interaction weight for user A is 1-0.375=0.625.

[0068] In this embodiment, a dual-channel weight quantization generation mechanism is constructed to calculate the first standard weight, the first interaction weight, the second standard weight, and the second interaction weight, respectively. The evaluation is carried out from two independent dimensions: objective rehabilitation progress and subjective emotional state. This reveals that under the same training scenario, objective rehabilitation needs and subjective user state may have different or even opposite strategy tendencies, providing clear input for subsequent strategy fusion and optimization.

[0069] S30: Calculate the difference between the first standard weight, the first interaction weight, the second standard weight, and the second interaction weight to obtain the weight difference coefficient; In this embodiment of the application, the weight difference coefficient is a comprehensive scalar value used to quantify the overall difference between the first standard weight and the first interaction weight, as well as between the second standard weight and the second interaction weight, reflecting the degree of complexity between rehabilitation needs and the user's current willingness or ability.

[0070] Specifically, the deviations between the first standard weight and the second standard weight, as well as the deviations between the first interaction weight and the second interaction weight, are calculated separately. Then, a weight difference coefficient is calculated based on these two deviations. The larger the value, the greater the difference between the objective requirements and the subjective state, and the higher the complexity of the current user's situation; the smaller the value, the smaller the difference, the more consistent the two are, and the lower the complexity of the current user's situation.

[0071] Step S30 in the method provided in this application embodiment includes: Calculate the deviation between the first standard weight and the second standard weight, and use it as the standard weight difference coefficient; Calculate the deviation between the first interaction weight and the second interaction weight, and use it as the interaction weight difference coefficient; The weight difference coefficient is calculated based on the standard weight difference coefficient and the interaction weight difference coefficient.

[0072] In this embodiment of the application, the deviation between the first standard weight and the second standard weight is first calculated as the standard weight difference coefficient.

[0073] Among them, the deviation range refers to the degree of inconsistency or deviation between two values; the standard weight difference coefficient is a coefficient obtained by calculating the deviation range between the first standard weight and the second standard weight.

[0074] Specifically, the deviation magnitude is the absolute value of the difference between the first standard weight and the second standard weight, i.e., the standard weight difference coefficient = |first standard weight - second standard weight|. A smaller standard weight difference coefficient indicates a smaller difference between rehabilitation needs and user experience; conversely, a larger standard weight difference coefficient indicates a greater difference between rehabilitation needs and user experience at the standard level.

[0075] For example, user A's first standard weight is 0.625, user A's second standard weight is 0.375, and the standard weight difference coefficient is 0.625-0.375=0.25, which is relatively small, indicating that the difference at the standard level is small.

[0076] Next, the deviation between the first interaction weight and the second interaction weight is calculated and used as the interaction weight difference coefficient.

[0077] Specifically, the deviation between the first interaction weight and the second interaction weight is the absolute value of the difference between the first interaction weight and the second interaction weight, that is, the interaction weight difference coefficient = |first interaction weight - second interaction weight|.

[0078] For example, user A's first interaction weight is 0.375. User A's second interaction weight is 0.625, and the interaction weight difference coefficient is |0.375-0.6265|=0.25, which is relatively small, indicating that the difference at the interaction level is small.

[0079] Finally, the weight difference coefficient is calculated based on the standard weight difference coefficient and the interaction weight difference coefficient.

[0080] Among them, the standard weight difference coefficient and the interaction weight difference coefficient represent the two aspects of standard and interaction, respectively. If more attention is paid to the standard of movement, it may reduce the user experience. If more attention is paid to interaction, that is, to the user experience, it may affect the rehabilitation progress. The two are contradictory. Therefore, it is necessary to consider both the degree of interaction standard and the rehabilitation progress of interaction at the same time to comprehensively reflect the degree of difference between the two.

[0081] The standard weight difference coefficient and the interactive weight difference coefficient are used as the weight difference coefficient for comprehensive evaluation. The weight difference coefficient = standard weight difference coefficient + interactive weight difference coefficient.

[0082] Specifically, the larger the weight difference coefficient, the greater the difference, and the more suitable the solution needs to be found; the smaller the weight difference coefficient, the more unified the opinions.

[0083] For example, the weight difference coefficients for user A are 0.25 + 0.25 = 0.5.

[0084] In this embodiment, the complex multi-objective trade-off between rehabilitation progress and user experience is transformed into a computable single objective. The weight difference coefficient dynamically reflects the directionality of personalized strategy adjustment at the current moment, which is conducive to improving the adaptive adjustment of search intensity in the optimization process.

[0085] S40: Based on the weight difference coefficient, the first standard weight, the first interaction weight, the second standard weight, and the second interaction weight, and combined with the weight difference coefficient, optimize the rehabilitation interaction strategy of the gamified digital human to obtain the optimal rehabilitation interaction strategy, and perform interaction processing in combination with the interaction standard, wherein the optimal rehabilitation interaction strategy includes the optimal interaction qualification threshold.

[0086] In this embodiment, the threshold for optimizing the interaction is the threshold at which the standard of the action meets the requirements during the action interaction; the optimal rehabilitation interaction strategy is the best set of strategy parameters obtained after optimization calculation; and the interaction processing refers to the process by which the gamified digital human application interacts with the user in real time using the optimized strategy.

[0087] Specifically, based on the weight difference coefficient, the first standard weight, the first interaction weight, the second standard weight, and the second interaction weight, the optimal rehabilitation interaction strategy is sought to find the best outcome. Finally, the optimal interaction qualification threshold that best balances rehabilitation effect and user experience is output and immediately applied to the current interaction.

[0088] Step S40 in the method provided in this application embodiment includes: Obtain the interaction qualification threshold range, wherein the interaction qualification threshold range is from 0 to 1; Randomly set the first acceptable threshold for interaction; Based on the first standard weight, the first interaction weight, the second standard weight, the second interaction weight, and the rehabilitation characteristics, the first strategy fitness is calculated to obtain the first interaction qualification threshold. The interaction qualification threshold is iteratively optimized. During the optimization process, the configuration is optimized according to the weight difference coefficient until convergence, and the optimal interaction qualification threshold with the greatest strategy fitness is obtained, which is used as the optimal rehabilitation interaction strategy.

[0089] In this embodiment of the application, the interaction qualification threshold range is first obtained, wherein the interaction qualification threshold range is from 0 to 1.

[0090] Specifically, the interaction qualification threshold to be optimized is constrained to a continuous interval between 0 and 1, so that the optimization algorithm can be calculated independently of the specific rehabilitation action type, ensuring that strategy adjustments can be represented by the same threshold searched in [0,1].

[0091] Secondly, a first qualified threshold for interaction is randomly set.

[0092] Specifically, within a predefined range of interaction qualification thresholds, an initial value is selected without preference according to a uniform distribution or other probability distribution, and the first interaction qualification threshold is used as the starting point for subsequent calculations and iterative improvements.

[0093] The first interaction qualification threshold is the starting point or initial solution of the iterative optimization process.

[0094] For example, the number 0.7 is randomly generated and set as the first interaction qualification threshold.

[0095] When digital human recognition reaches the acceptable threshold during interaction, it can be considered to meet the acceptable interaction requirements.

[0096] For example, through sensor interaction, the interaction standard is that the grip force duration reaches N, and the set interaction qualification threshold is 0.7, that is, 0.7N. Then, if the grip force reaches 0.7N during the interaction, the digital human will give the feedback result that the interaction is qualified.

[0097] Next, based on the first standard weight, the first interaction weight, the second standard weight, the second interaction weight, and the rehabilitation characteristics, the first strategy fitness that obtains the first interaction qualification threshold is calculated.

[0098] Among them, strategy fitness is used to quantitatively evaluate the overall merits of any interaction threshold. The higher the fitness value, the better the strategy corresponding to that threshold is, and the better it can balance rehabilitation effectiveness and user experience.

[0099] Finally, the interaction qualification threshold is iteratively optimized. During the optimization process, the configuration is optimized according to the weight difference coefficient until convergence, and the optimal interaction qualification threshold with the greatest strategy fitness is obtained, which is used as the optimal rehabilitation interaction strategy.

[0100] Iterative optimization refers to a progressively improving search process, which starts from the first interaction qualification threshold, generates new candidate thresholds, and evaluates their fitness in a loop.

[0101] Specifically, in each iteration, several new thresholds are randomly generated with small increments based on the current weight difference coefficients, and the fitness of the new thresholds is calculated. This process continues until the algorithm determines that it has found a threshold under the given conditions that cannot be further improved, thus obtaining the optimal rehabilitation interaction strategy.

[0102] In step S40 of the method provided in this application embodiment, the first strategy fitness, which calculates the first interaction qualification threshold based on the first standard weight, the first interaction weight, the second standard weight, the second interaction weight, and the rehabilitation features, includes: Calculate the ratio of the first interaction qualification threshold to the rehabilitation label value within the rehabilitation feature to obtain the first rehabilitation parameter; Calculate the ratio of the sentiment annotation value within the user feature to the first interaction qualification threshold to obtain the first sentiment parameter; Calculate the average of the first standard weight and the second standard weight, and the average of the first interaction weight and the second interaction weight to obtain the standard weight and interaction weight. Then, perform a weighted calculation on the first rehabilitation parameter and the first emotional parameter to obtain the first strategy fitness.

[0103] In this embodiment, the ratio of the first interaction qualification threshold to the rehabilitation label value within the rehabilitation feature is first calculated to obtain the first rehabilitation parameter.

[0104] Specifically, the first rehabilitation parameter = first interaction qualification threshold / rehabilitation annotation value within the rehabilitation feature. If the first interaction qualification threshold is approximately equal to the rehabilitation annotation value within the rehabilitation feature (i.e., the ratio is close to 1), it indicates that the difficulty adjustment proposed by the system is highly matched with the user's actual ability. If the first interaction qualification threshold is greater than the rehabilitation annotation value within the rehabilitation feature (i.e., the ratio is greater than 1), it indicates that the rehabilitation standard is too strict for the user, which may mean unrealistically high requirements.

[0105] If the first interaction qualification threshold is less than the rehabilitation label value within the rehabilitation feature (i.e., the ratio is less than 1), it indicates that the standard suggested by the system is more lenient than the user's current ability, which may mean that the training dose is insufficient. Therefore, the closer the value of the first rehabilitation parameter is to 1, the higher the rehabilitation efficiency.

[0106] For example, user A's rehabilitation feature has a rehabilitation label value of 50%, and the first rehabilitation parameter is 0.7 / 50%=1.4.

[0107] Secondly, the ratio of the sentiment annotation value within the user's features to the first interaction qualification threshold is calculated to obtain the first sentiment parameter.

[0108] Among them, the first emotion parameter is a quantitative assessment of the quality of the current interaction qualification threshold from the perspective of user experience and emotional response.

[0109] Specifically, the first sentiment parameter is calculated as: user feature sentiment annotation value / first interaction pass threshold. A higher user sentiment (E) results in a larger ratio, indicating that the user is currently better able to handle the difficulty and has a better emotional experience. Therefore, a larger first sentiment parameter value leads to a better user experience and higher levels of positive emotion during user training.

[0110] For example, user A's sentiment label value within user features is 0.3, and the first sentiment parameter is 0.3 / 0.7≈0.43.

[0111] Finally, the average of the first standard weight and the second standard weight, as well as the average of the first interaction weight and the second interaction weight, are calculated to obtain the standard weight and interaction weight. The first rehabilitation parameter and the first emotional parameter are then weighted and calculated to obtain the first strategy fitness.

[0112] Among them, the standard weight is the importance of the action standard level after taking into account both objective needs and subjective state; the interaction weight is the importance of optimizing the interaction experience after taking into account both objective needs and subjective state.

[0113] Specifically, the standard weight and interaction weight are calculated by taking the mean values ​​of each. Then, the standard weight and interaction weight are weighted and summed with the first rehabilitation parameter and the first emotional parameter, respectively, to calculate the fitness of the first strategy. The fitness of the first strategy is the sum of the product of the first rehabilitation parameter and the standard weight, and the product of the first emotional parameter and the interaction weight. The fitness of the first strategy = the product of the first rehabilitation parameter and the standard weight + the product of the first emotional parameter and the interaction weight. The higher the fitness of the first strategy, the better the strategy.

[0114] For example, user A's first interaction weight is 0.375, second interaction weight is 0.625, first standard weight is 0.625, and second standard weight is 0.375. User A's interaction weight and standard weight are (0.375+0.625) / 2=0.5 and (0.62+0.375) / 2=0.5, respectively, and the fitness of the first strategy is 1.4×0.5+0.43×0.5=0.915.

[0115] In step S40 of the method provided in this application embodiment, iterative optimization of the interaction qualification threshold is performed. During the optimization process, the optimization configuration is performed according to the weight difference coefficient until convergence, including: Get the preset adjustment quantity; Based on the weight difference coefficient, the preset adjustment quantity is adaptively calculated to obtain the adjustment quantity; Within the neighborhood of the first interaction qualification threshold, the first interaction qualification threshold is randomly adjusted by the number of adjustments to obtain multiple second interaction qualification thresholds. Continue iterative optimization until convergence.

[0116] In this embodiment of the application, the preset adjustment quantity is first obtained.

[0117] The preset adjustment quantity is a basic parameter initially set. It represents the baseline number of new candidate solutions generated during each iteration of optimization, exploring around the current optimal candidate solution.

[0118] Specifically, the preset adjustment quantity can be obtained by using the average adjustment quantity in historical data.

[0119] For example, the preset adjustment quantity is calculated to be 5.

[0120] Secondly, based on the weight difference coefficient, the preset adjustment quantity is adapted and calculated to obtain the adjustment quantity.

[0121] Specifically, the preset adjustment quantity is calculated based on the magnitude of the weight difference coefficient: if the weight difference coefficient is large, it indicates a large degree of difference, and the adjustment quantity will be increased; if the weight difference coefficient is small, it indicates a small degree of difference, and the adjustment quantity will be reduced.

[0122] When the degree of difference is large, more extensive sampling is performed to avoid getting trapped in local optima. When the degree of difference is small, the optimal solution may be nearby, so fewer exploration points are needed for more efficient local optimization.

[0123] The adjustment quantity is the product of the preset adjustment quantity and the weight difference coefficient. The adjustment quantity = preset adjustment quantity × weight difference coefficient, and the product is rounded to obtain the corresponding adjustment quantity.

[0124] For example, user A's weight difference coefficient is 0.5. Therefore, the adjustment amount for user A is: 5 × 0.5 = 0.25, rounded up to 3.

[0125] Next, within the neighborhood of the first interaction qualification threshold, the first interaction qualification threshold is randomly adjusted by the aforementioned adjustment amount to obtain multiple second interaction qualification thresholds.

[0126] The first interaction qualification threshold is the candidate solution with the highest fitness in the current iteration, which is the center point of the local search; the neighborhood is a small range around the current best solution, which limits the range of random adjustment and ensures that the search is locally refined; random adjustment is to randomly generate a new value within the neighborhood; the second interaction qualification threshold is a new candidate solution obtained by randomly adjusting the first interaction qualification threshold a number of times.

[0127] Specifically, first, the current optimal solution, i.e., the first interaction qualification threshold and the search neighborhood, are determined. Then, the adjustment quantity is calculated, and within the neighborhood, a corresponding number of second interaction qualification thresholds are randomly and independently generated. Finally, policy fitness is calculated, and the maximum policy fitness is updated.

[0128] The neighborhood is defined as ±0.1 of the first interaction qualification threshold.

[0129] For example, user A's current policy fitness is 0.915, and the neighborhood is [0.815, 1]. Three interaction qualification thresholds are generated: {0.75, 0.95, 0.93}. User A: The evaluation finds that the interaction qualification threshold of 0.95 is better than the current one, so the interaction qualification threshold is updated to 0.95 after iteration.

[0130] Finally, iterative optimization continues until convergence.

[0131] Specifically, the process continues to repeat. After several iterations, the optimal threshold hovers within a certain range, and the corresponding optimal fitness improves very little in the most recent iterations. At this point, it is determined that the optimization process has converged, the iteration stops, and the threshold that obtained the highest fitness in the most recent iteration is officially output as the optimal interaction qualified threshold.

[0132] For example, if the value remains almost unchanged after 10 iterations, the iteration is considered to have converged, and the optimal interaction qualification threshold of 0.95 is output.

[0133] In this embodiment, the first rehabilitation parameter ensures that the optimization process does not completely deviate from the user's actual ability level, avoiding blindly setting the difficulty. By responding to the user's negative emotions and lowering the difficulty based on the first emotional parameter, the long-term efficacy of rehabilitation training is significantly improved, increasing the user's training motivation. Simultaneously, by dynamically adjusting rehabilitation goals through thresholds, training is not completely sacrificed for the sake of user experience, thus promoting improved final rehabilitation results.

[0134] The embodiments of this application, through the above specific implementation methods, achieve the following technical effects: In this embodiment, personalized initial interaction thresholds are first provided for users of different rehabilitation stages and types by setting interaction standards. The main rehabilitation types are considered, and multi-dimensional features such as rehabilitation progress and user basic information are integrated. This reduces manual intervention, improves the efficiency and consistency of system deployment, ensures the safety and basic relevance of training, and realizes the automated and highly accurate personalized generation of initial interaction standards.

[0135] Secondly, by constructing a dual-channel weight quantification generation mechanism, the first standard weight, the first interaction weight, the second standard weight, and the second interaction weight are calculated respectively. The evaluation is carried out from two independent dimensions: objective rehabilitation progress and subjective emotional state. This reveals that under the same training scenario, there may be different or even opposite strategy tendencies between objective rehabilitation needs and subjective user state, providing clear input for subsequent strategy fusion and optimization.

[0136] Furthermore, by calculating the weight difference coefficient, the complex multi-objective trade-off between rehabilitation progress and user experience is transformed into a calculable single objective. The weight difference coefficient dynamically reflects the directionality of personalized strategy adjustments at the current moment, which is conducive to improving the adaptive adjustment of search intensity during the optimization process.

[0137] Ultimately, the first rehabilitation parameter and the first emotional parameter were calculated. The first rehabilitation parameter ensured that the optimization process did not completely deviate from the user's actual ability level, avoiding blindly setting the difficulty. The first emotional parameter responded to the user's negative emotions and adjusted the difficulty accordingly, significantly improving the long-term efficacy of rehabilitation training and increasing the user's training motivation. At the same time, by dynamically adjusting rehabilitation goals through thresholds, training was not completely sacrificed for the sake of user experience, thus promoting the improvement of the final rehabilitation effect.

[0138] Example 2, as Figure 2 As shown, this embodiment of the invention also provides a computer intelligent chip 200, which stores a first computer program 211. When the first computer program 211 is executed by a processor, it implements a gamified digital human rehabilitation interaction strategy optimization method as described in Embodiment 1.

[0139] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0140] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for optimizing rehabilitation interaction strategies for gamified digital humans, characterized in that, The method includes: Collect the rehabilitation characteristics and user characteristics of the target users, and configure the interaction standards of the target users based on the rehabilitation characteristics; Based on the rehabilitation characteristics, standard weights and interaction weights are allocated to obtain a first standard weight and a first interaction weight. Based on the user characteristics, standard weights and interaction weights are allocated to obtain a second standard weight and a second interaction weight. Calculate the differences between the first standard weight, the first interaction weight, the second standard weight, and the second interaction weight to obtain the weight difference coefficient; Based on the weight difference coefficient, the first standard weight, the first interaction weight, the second standard weight, and the second interaction weight, and combined with the weight difference coefficient, the rehabilitation interaction strategy of the gamified digital human is optimized to obtain the optimal rehabilitation interaction strategy. The interaction is then processed in conjunction with the interaction standard, wherein the optimal rehabilitation interaction strategy includes the optimal interaction qualification threshold.

2. The method for optimizing rehabilitation interaction strategies for gamified digital humans according to claim 1, characterized in that, Collect the rehabilitation characteristics and user characteristics of the target user, and configure the interaction standards of the target user based on the rehabilitation characteristics, including: Collect the rehabilitation characteristics and user characteristics of the target users. The rehabilitation characteristics include rehabilitation type and rehabilitation label value. The rehabilitation label value is the ratio of actual rehabilitation time to planned rehabilitation time. The user characteristics include user emotion label value. Based on the rehabilitation characteristics, the interaction standards for the target user are configured, wherein the interaction standards include interaction action standards.

3. The method for optimizing rehabilitation interaction strategies for gamified digital humans according to claim 1, characterized in that, Based on the rehabilitation characteristics, configure the interaction standards for the target user, including: A rehabilitation interaction standard generator is obtained, wherein the rehabilitation interaction standard generator is constructed based on machine learning and is trained using a sample rehabilitation feature set and a sample interaction standard set; The rehabilitation features are input into the rehabilitation interaction standard generator, and the interaction standard is output.

4. The method for optimizing rehabilitation interaction strategies for gamified digital humans according to claim 1, characterized in that, Based on the rehabilitation characteristics, standard weights and interaction weights are assigned to obtain first standard weights and first interaction weights, including: Obtain the rehabilitation label value within the rehabilitation feature; The ratio of the stated rehabilitation label value to the preset rehabilitation label value is calculated and used as the first standard weight. The first interaction weight is calculated based on the first standard weight.

5. The method for optimizing rehabilitation interaction strategies for gamified digital humans according to claim 1, characterized in that, Based on the user characteristics, standard weights and interaction weights are assigned to obtain second standard weights and second interaction weights, including: Obtain the user sentiment annotation value within the user features; The ratio of the user emotion label value to the preset user emotion label value is calculated and used as the second standard weight; The second interaction weight is calculated based on the second standard weight.

6. The method for optimizing rehabilitation interaction strategies for gamified digital humans according to claim 1, characterized in that, Calculate the differences between the first standard weight, the first interaction weight, the second standard weight, and the second interaction weight to obtain the weight difference coefficient, including: Calculate the deviation between the first standard weight and the second standard weight, and use it as the standard weight difference coefficient; Calculate the deviation between the first interaction weight and the second interaction weight, and use it as the interaction weight difference coefficient; The weight difference coefficient is calculated based on the standard weight difference coefficient and the interaction weight difference coefficient.

7. The method for optimizing rehabilitation interaction strategies for gamified digital humans according to claim 1, characterized in that, Based on the weight difference coefficient, the first standard weight, the first interaction weight, the second standard weight, and the second interaction weight, combined with the weight difference coefficient, the rehabilitation interaction strategy of the gamified digital human is optimized to obtain the optimal rehabilitation interaction strategy, including: Obtain the interaction qualification threshold range, wherein the interaction qualification threshold range is from 0 to 1; Randomly set the first acceptable threshold for interaction; Based on the first standard weight, the first interaction weight, the second standard weight, the second interaction weight, and the rehabilitation characteristics, the first strategy fitness is calculated to obtain the first interaction qualification threshold. The interaction qualification threshold is iteratively optimized. During the optimization process, the configuration is optimized according to the weight difference coefficient until convergence, and the optimal interaction qualification threshold with the greatest strategy fitness is obtained, which is used as the optimal rehabilitation interaction strategy.

8. The method for optimizing rehabilitation interaction strategies for gamified digital humans according to claim 7, characterized in that, Based on the first standard weight, the first interaction weight, the second standard weight, the second interaction weight, and the rehabilitation characteristics, the first strategy fitness to obtain the first interaction qualification threshold is calculated, including: Calculate the ratio of the first interaction qualification threshold to the rehabilitation label value within the rehabilitation feature to obtain the first rehabilitation parameter; Calculate the ratio of the sentiment annotation value within the user feature to the first interaction qualification threshold to obtain the first sentiment parameter; Calculate the average of the first standard weight and the second standard weight, and the average of the first interaction weight and the second interaction weight to obtain the standard weight and interaction weight. Then, perform a weighted calculation on the first rehabilitation parameter and the first emotional parameter to obtain the first strategy fitness.

9. The method for optimizing rehabilitation interaction strategies for gamified digital humans according to claim 1, characterized in that, Iterative optimization of the interaction qualification threshold is performed, and the optimization configuration is adjusted according to the weight difference coefficient during the optimization process until convergence, including: Get the preset adjustment quantity; Based on the weight difference coefficient, the preset adjustment quantity is adaptively calculated to obtain the adjustment quantity; Within the neighborhood of the first interaction qualification threshold, the first interaction qualification threshold is randomly adjusted by the number of adjustments to obtain multiple second interaction qualification thresholds. Continue iterative optimization until convergence.

10. A smart chip for optimizing rehabilitation interaction strategies in gamified digital humans, characterized in that, The smart chip stores a first computer program, which, when executed by a processor, implements a gamified digital human rehabilitation interaction strategy optimization method as described in any one of claims 1 to 9.

Images