A toy recommendation method and device, a storage medium and an electronic device

Abstract

The application provides a toy recommendation method and device, a storage medium and an electronic device, and relates to the technical field of computers. The method comprises the following steps: obtaining behavior video data and behavior observation data corresponding to a target object, wherein the behavior video data comprises the content of the interaction between the target object and a toy; the behavior observation data is the behavior performance data of the target object in a preset time period obtained through a user terminal; performing behavior analysis on the behavior video data to obtain behavior parameters of the target object in each capability dimension, wherein the behavior parameters comprise operation proficiency, interest focus, interaction duration and interaction frequency; generating a capability development graph of the target object according to the behavior parameters and the behavior observation data; determining a development lag dimension of the target object in each capability dimension based on the capability development graph; determining a candidate toy matched with the development lag dimension; and outputting recommendation information of the candidate toy. The target object can be quickly and accurately recommended a toy.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a toy recommendation method, apparatus, storage medium, and electronic device. Background Technology

[0002] Infant and toddler care and early development guidance are crucial for children's healthy growth. Scientific toy recommendations and interactive guidance can effectively promote the balanced development of children's abilities across multiple dimensions, including language, logic, motor skills, social skills, and exploration. Currently, methods for assessing children's abilities and recommending toys mainly rely on guardians' subjective observations or simple questionnaires, manually identifying children's developmental weaknesses and selecting corresponding toys accordingly.

[0003] However, existing methods still have significant shortcomings in terms of the accuracy of ability assessment and the relevance of recommendations. The assessment process is difficult to fully reflect the true developmental status of children, and there is a discrepancy between the recommendations and the actual needs of children, making it difficult for recommended toys to effectively promote the balanced development of children's abilities. Summary of the Invention

[0004] In view of this, this application provides a toy recommendation method, apparatus, storage medium, and electronic device, which can quickly and accurately recommend toys to a target audience. The specific solution is as follows: A toy recommendation method includes: The system acquires behavioral video data and behavioral observation data corresponding to the target object. The behavioral video data includes the content of the target object interacting with the toy. The behavioral observation data is the behavioral performance data of the target object within a preset time period, acquired through a user terminal. Behavioral analysis is performed on the behavioral video data to obtain behavioral parameters of the target object in various capability dimensions, including operational proficiency, focus of interest, interaction duration, and interaction frequency. Generate a capability development map of the target object based on the behavioral parameters and the behavioral observation data; Based on the capability development map, the development lag dimension of the target object is determined in each of the capability dimensions; Identify candidate toys that match the aforementioned developmental lag dimension; Output recommendation information for the candidate toys.

[0005] Optionally, in the above method, the step of performing behavioral analysis on the behavioral video data to obtain behavioral parameters of the target object across various capability dimensions includes: The behavioral video data is subjected to posture recognition to obtain the limb movements and hand operation trajectories of the target object; Based on the target object's limb movements and hand operation trajectories, determine the target object's operational proficiency in interacting with the toy; Eye tracking is performed on the behavioral video data to determine the target object's focus of interest during the interaction. The duration and number of interactions between the target object and the toy in the behavioral video data are statistically analyzed to determine the interaction duration and frequency.

[0006] Optionally, in the above method, determining the target object's operational proficiency in interacting with the toy based on the target object's limb movements and hand operation trajectories includes: Determine the degree of matching between the target object's limb movements and hand operation trajectories and preset standard operation data; The matching degree is used to determine the target object's operational proficiency in interacting with the toy.

[0007] Optionally, in the above method, generating the capability development map of the target object based on the behavioral parameters and the behavioral observation data includes: Establish a mapping relationship between each capability dimension and the behavioral parameters and indicators in the behavioral observation data; The behavioral parameters and the behavioral observation data are input into a preset cross-validation model, which is used to identify whether there is a conflict between the evaluation results of the behavioral parameters and the behavioral observation data for the same ability dimension. In response to the identification of a conflict, the behavioral parameters and the behavioral observation data are weighted according to a preset confidence rule; Based on the weighted behavioral parameters and the behavioral observation data, a capability development map of the target object is generated.

[0008] Optionally, in the above method, generating the capability development map of the target object based on the weighted behavioral parameters and the behavioral observation data includes: The weighted behavioral parameters are converted into a first feature vector; The weighted behavioral observation data is converted into a second feature vector; The first feature vector and the second feature vector are concatenated to generate a fused feature vector; The fused feature vector is input into a pre-trained evaluation model, which outputs the evaluation score of the target object in each capability dimension. The evaluation model is trained based on sample data labeled with the evaluation results of each capability dimension. Based on the assessment score, a capability development map of the target object is generated.

[0009] Optionally, in the above method, determining the developmental lag dimension of the target object in each of the capability dimensions based on the capability development map includes: Obtain the evaluation scores of the target object on each capability dimension, as output by the evaluation model; Obtain preset standard development reference data, which is determined based on the ability development distribution of the target object's age group, and includes the standard score range for each ability dimension. The assessment scores are compared with the standard score ranges for the corresponding ability dimensions. At least one capability dimension whose assessment score is below the lower limit of the standard score range is identified as a developmental lag dimension.

[0010] Optionally, in the above method, determining candidate toys that match the developmental lag dimension includes: Obtain the corresponding ability training tags for each toy from the toy resource library, and the ability training tags correspond one-to-one with the ability dimensions; Identify candidate toys whose ability training labels match the developmental lag dimension, and generate a first set; Remove toys from the first set whose ability training labels overlap with other ability dimensions besides the developmental lag dimension, and generate a second set; The toys in the second set are identified as candidate toys that match the developmental lag dimension.

[0011] A toy recommendation device, comprising: The acquisition unit is used to acquire behavioral video data and behavioral observation data corresponding to the target object. The behavioral video data includes the content of the target object interacting with the toy. The behavioral observation data is the behavioral performance data of the target object within a preset time period, acquired through a user terminal. The behavior analysis unit is used to perform behavior analysis on the behavior video data to obtain the behavior parameters of the target object in various ability dimensions. The behavior parameters include operation proficiency, interest focus, interaction duration and interaction frequency. A generation unit is configured to generate a capability development map of the target object based on the behavioral parameters and the behavioral observation data. The first determining unit is used to determine the development lag dimension of the target object in each of the capability dimensions based on the capability development map; The second determining unit is used to determine candidate toys that match the development lag dimension; The output unit is used to output recommendation information for the candidate toys.

[0012] A storage medium comprising storage instructions, wherein, when the instructions are executed, a device containing the storage medium is configured to perform the toy recommendation method as described above.

[0013] An electronic device includes a memory and one or more instructions, wherein one or more instructions are stored in the memory and configured to be executed by one or more processors as described above in the toy recommendation method.

[0014] Based on the above, this application provides a toy recommendation method, apparatus, storage medium, and electronic device. The method includes: acquiring behavioral video data and behavioral observation data corresponding to a target object; the behavioral video data includes content of the target object interacting with toys; the behavioral observation data is behavioral performance data of the target object within a preset time period obtained through a user terminal; performing behavioral analysis on the behavioral video data to obtain behavioral parameters of the target object in various ability dimensions, including operational proficiency, focus of interest, interaction duration, and interaction frequency; generating an ability development map of the target object based on the behavioral parameters and the behavioral observation data; determining the developmental lag dimension of the target object in each ability dimension based on the ability development map; determining candidate toys that match the developmental lag dimension; and outputting recommendation information for the candidate toys. This application, by acquiring behavioral video data and behavioral observation data, and generating an ability development map based on behavioral parameters obtained from behavioral analysis and behavioral observation data, can comprehensively depict the ability development status of the target object from different dimensions, overcoming the scene bias problem of evaluation based on a single data source. Furthermore, based on the ability development map, the development lag dimension is identified, and candidate toys that match the development lag dimension are selected for recommendation. This allows the recommended content to accurately target the ability shortcomings of the target audience, thereby improving the relevance of toy recommendations. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0016] Figure 1 A flowchart of a toy recommendation method provided in this application; Figure 2 A flowchart illustrating a process for obtaining behavioral parameters of a target object across various capability dimensions, as provided in this application; Figure 3A flowchart illustrating a process for generating a capability development map of a target object based on behavioral parameters and behavioral observation data, as provided in this application; Figure 4 A schematic diagram of a toy recommendation device provided in this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided in this application. Detailed Implementation

[0017] 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 some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] In this application, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0019] This application provides a toy recommendation method, which can be applied to electronic devices. The flowchart of the method is shown below. Figure 1 As shown, it specifically includes: S101: Obtain behavioral video data and behavioral observation data corresponding to the target object. The behavioral video data includes the content of the target object interacting with the toy. The behavioral observation data is the behavioral performance data of the target object within a preset time period obtained through the user terminal.

[0020] In this embodiment, the target object can be an infant or a preschool child, etc.

[0021] Optionally, the behavioral video data is video stream data collected by visual sensors deployed in children's activity areas, including but not limited to RGB cameras, depth cameras, or infrared cameras. The behavioral video data records a continuous sequence of images of the target object physically interacting with the toy, including actions such as touching, grasping, stacking, arranging, and throwing.

[0022] In this embodiment, the behavioral observation data is descriptive data obtained through a user terminal and recorded by the monitored individual regarding the target's daily behavioral performance. The preset time period can be set to 24 hours, one week, or one month, etc.

[0023] Optionally, the content of behavioral observation data includes the guardian's observation records of the target subject in the dimensions of language expression, social interaction, emotional response, interests and preferences and living habits, such as descriptive information such as "the child took the initiative to share toys with his friends today" and "the child showed a sustained interest in building blocks".

[0024] S102: Perform behavioral analysis on the behavioral video data to obtain the behavioral parameters of the target object in various ability dimensions. The behavioral parameters include operational proficiency, focus of interest, interaction duration, and interaction frequency.

[0025] In this embodiment, the ability dimension is an evaluation dimension used to assess the development level of the target object, including at least one of language ability, logical thinking ability, exploratory ability, social ability, and motor ability.

[0026] Optionally, by performing posture recognition and eye tracking on the behavioral video data, the target object's limb movement trajectory, hand operation trajectory, and eye gaze information can be extracted, and operation proficiency, focus of interest, interaction duration, and interaction frequency can be generated based on the extracted information.

[0027] S103: Generate a capability development map of the target object based on behavioral parameters and behavioral observation data.

[0028] In this embodiment, the capability development map is a data structure representing the development level of a target object across various capability dimensions. A mapping relationship can be established between each capability dimension and behavioral parameters and indicators in the behavioral observation data. The behavioral parameters and behavioral observation data are then input into a cross-validation model. The cross-validation model is used to identify any conflicts between the evaluation results of the behavioral parameters and behavioral observation data for the same capability dimension. In response to the identification of conflicts, the behavioral parameters and behavioral observation data are weighted according to a preset confidence rule, which pre-sets the credibility weight of different data sources when evaluating each capability dimension. Based on the weighted data, a capability development map containing a comprehensive score for each capability dimension is generated.

[0029] S104: Based on the capability development map, identify the development lag dimension of the target object in each capability dimension.

[0030] Specifically, the development lag dimension is at least one capability dimension in the capability development map whose overall score is lower than a preset reference threshold.

[0031] Optionally, standard development reference data corresponding to the target object's age in months can be obtained. The standard development reference data includes the standard score range corresponding to each capability dimension. The comprehensive score of each capability dimension in the capability development map is compared with the standard development reference data, and at least one capability dimension with a comprehensive score lower than the lower limit of the standard score range is identified as a developmental lag dimension.

[0032] S105: Identify candidate toys that match the development lag dimension.

[0033] In this embodiment, candidate toys are toys stored in the toy resource library that have the function of training abilities for developmental lag dimensions. When determining candidate toys, the ability training tags corresponding to each toy are obtained from the toy resource library, and the ability training tags correspond one-to-one with the ability dimensions. Candidate toys whose ability training tags match the developmental lag dimensions are selected. From the selection results, candidate toys whose ability training tags have training overlap with other ability dimensions besides the developmental lag dimension are removed, and the remaining candidate toys are determined as candidate toys that match the developmental lag dimensions.

[0034] S106: Output recommendation information for candidate toys.

[0035] Optional information may include illustrated descriptions of the candidate toys, instructions on how to develop skills, and interactive guidance.

[0036] In this embodiment, the recommendation information can be pushed to user terminals, including guardians' mobile terminals or home smart devices. The recommendation information can also be pushed to a management platform, which may include a management platform for community-based childcare integration spaces or a management system for childcare institutions. Upon receiving the recommendation information, the management platform triggers a corresponding toy allocation process, allocating candidate toys to the target child's activity area. This activity area may include a family activity area, a community toy library, or a play area in a childcare institution.

[0037] In one optional embodiment, after outputting the recommendation information for candidate toys, the method further includes: after a preset period, acquiring updated behavioral video data of the target object interacting with the candidate toys; updating the capability development map based on the updated behavioral video data; and determining whether the development lag dimension has changed based on the updated capability development map. If the development lag dimension has changed, the development lag dimension is redefined and recommendations continue; if the development lag dimension has not changed, the screening criteria for candidate toys are adjusted, including relaxing or tightening the matching threshold, changing the matching strategy, or increasing the number of recommended toys.

[0038] This application embodiment constructs a capability development map by fusing behavioral video data and behavioral observation data, enabling the assessment of the developmental level of target objects based on multi-source data, and recommending toys based on the developmental lag dimension. This provides targeted capability training resources for target objects and improves the accuracy and effectiveness of toy recommendations.

[0039] In one embodiment provided in this application, based on the above-described solution, optionally, the process of performing behavioral analysis on behavioral video data to obtain behavioral parameters of the target object in various capability dimensions, such as... Figure 2 As shown, it includes: S201: Perform posture recognition on behavioral video data to obtain the limb movements and hand operation trajectories of the target object.

[0040] In this embodiment, a pose recognition algorithm based on skeletal keypoint detection can be used to infer the coordinates of multiple skeletal keypoints of the target object in the image coordinate system for each frame of the behavioral video data. The skeletal keypoints include, but are not limited to, shoulder keypoints, elbow keypoints, wrist keypoints, hand keypoints, hip keypoints, knee keypoints, and ankle keypoints.

[0041] Optionally, limb movements are represented by the positional sequence of trunk and limb key points among the aforementioned skeletal key points, used to describe the overall motion state of the target object, including but not limited to posture changes such as standing, squatting, walking, and bending over. Hand operation trajectories are represented by the positional sequence of hand and wrist key points, used to describe the movement path of the target object's hand in space, including fine operation trajectories related to toy interaction such as grasping, moving, rotating, stacking, and releasing.

[0042] The output of pose recognition is the skeleton structure change information of the target object in a continuous time series. This information includes both the limb movement features and the hand operation trajectory features of the target object.

[0043] S202: Determine the target object's operational proficiency in interacting with the toy based on the target object's limb movements and hand operation trajectories.

[0044] In one embodiment provided in this application, based on the above-described solution, optionally, the operational proficiency of the target object in interacting with the toy is determined based on the target object's limb movements and hand operation trajectories, including: Determine the degree of matching between the target object's limb movements and hand operation trajectories and preset standard operation data; The matching degree is used to determine the target object's operational proficiency in interacting with the toy.

[0045] In this embodiment, the standard operation data is typical operation path data pre-collected for a specific toy. This data includes the sequence of limb movements and hand operation trajectories required to complete the expected gameplay of the toy. Taking stacking toys as an example, the standard operation data may include trajectory information corresponding to operation sequences such as grasping blocks, moving blocks to the stacking position, and releasing blocks to make them fall into the target position.

[0046] Optionally, the matching degree characterizes the similarity between the target object's actual operation trajectory and standard operation data. In practice, the target object's limb movement sequence and hand operation trajectory can be temporally aligned, and the Euclidean distance or dynamic time-warped distance between the aligned trajectory and the standard operation data can be calculated. A matching degree score is generated based on the calculated distance value. The higher the matching degree, the closer the target object's operation trajectory is to the standard operation path.

[0047] Specifically, the matching degree can be directly mapped to an operational proficiency score, or the matching degree can be converted into an operational proficiency score through a preset non-linear mapping function. The operational proficiency score is positively correlated with the matching degree; the higher the matching degree, the higher the operational proficiency score, indicating that the target object is more proficient in operating the toy.

[0048] S203: Perform eye tracking on behavioral video data to determine the target object's focus of interest during the interaction process.

[0049] In this embodiment, a gaze tracking algorithm based on pupil detection and corneal reflection can be used to process the facial region of the target object in the behavioral video data, extract the pupil center position and the corneal reflection point position, calculate the gaze direction vector based on the relative positional relationship between the two, and intersect the gaze direction vector with the three-dimensional structure of the scene in the behavioral video data to determine the gaze point position of the target object on the toy surface or toy parts.

[0050] Interest focus refers to the toy area or toy part where the target object concentrates its attention during interaction. To determine interest focus, a series of gaze points of the target object within a preset time window are acquired. Cluster analysis is performed on each gaze point location, and areas where the gaze point distribution density exceeds a preset density threshold are identified as interest focus. Furthermore, the dwell time for each gaze point can be recorded, and areas where the dwell time exceeds a preset duration threshold are identified as interest focus.

[0051] S204: Analyze the duration and number of interactions between the target object and the toy in the statistical behavioral video data to determine the interaction duration and frequency.

[0052] In this embodiment, the duration of a single interaction can be obtained by identifying the start and end frames of the physical contact or eye contact between the target object and the toy in the behavioral video data, and calculating the time difference between the start and end frames. When it is necessary to calculate the total interaction duration, the durations of all single interactions within a preset statistical period can be accumulated.

[0053] Optionally, the interaction count is the number of events in which the target object interacts with the toy within a preset statistical period. In implementation, event detection is performed on the behavioral video data. When the start event of physical contact or eye contact between the target object and the toy is detected, the interaction count is incremented by one, and the start time of the interaction is recorded. When the end event of the interaction is detected, the end time of the interaction is recorded.

[0054] Optionally, interaction duration and interaction frequency are two independent dimensions among the behavioral parameters, representing the target object's attention persistence and interaction frequency, respectively. The longer the interaction duration and the more interactions, the higher the target object's interest in the toy, or the stronger the toy's appeal to the target object.

[0055] In one embodiment provided in this application, based on the above-described scheme, optionally, the process of generating a capability development map of the target object based on behavioral parameters and behavioral observation data, such as... Figure 3 As shown, it includes: S301: Establish the mapping relationship between each capability dimension and behavioral parameters, as well as various indicators in the behavioral observation data.

[0056] The ability dimensions include at least one of the following: language ability, logical thinking ability, exploratory ability, social ability, and motor ability. Behavioral parameters include operational proficiency, focus of interest, interaction duration, and interaction frequency. Behavioral observation data includes descriptive records from the monitored individuals regarding the target individual's language expression, social interaction, emotional responses, interests, and lifestyle habits.

[0057] Optionally, the mapping relationship is used to define the degree of correlation between different ability dimensions and behavioral parameters, as well as various indicators in the behavioral observation data. For example, the motor ability dimension is strongly correlated with operational proficiency and interaction duration among the behavioral parameters, and weakly correlated with interaction frequency; the language ability dimension is strongly correlated with language expression records in the behavioral observation data, and weakly correlated with focus of interest among the behavioral parameters. The mapping relationship can be represented in the form of a weight matrix, where the matrix elements represent the contribution of the corresponding indicator to the ability dimension.

[0058] S302: Input the behavioral parameters and behavioral observation data into a preset cross-validation model. The cross-validation model is used to identify whether there is a conflict between the evaluation results of the behavioral parameters and behavioral observation data for the same ability dimension.

[0059] In this embodiment, the cross-validation model is a pre-built rule engine or machine learning model used to verify the consistency of multi-source evaluation results for the same capability dimension. In practice, the cross-validation model extracts evaluation information for the target capability dimension from behavioral parameters and behavioral observation data, respectively, generating a first evaluation result and a second evaluation result.

[0060] Specifically, conflict identification is based on the degree of difference between the first and second assessment results. When the degree of difference exceeds a preset conflict threshold, the cross-validation model determines that a conflict exists. Taking the motor ability dimension as an example, a high score in the operational proficiency score in the behavioral parameters indicates that the target subject is proficient in toy manipulation, while the guardian's records in the behavioral observation data show that the target subject has insufficient motor coordination in daily life. In this case, the assessment results from the two sources conflict.

[0061] S303: In response to the identification of a conflict, the behavioral parameters and behavioral observation data are weighted according to a preset confidence rule.

[0062] In this embodiment, the confidence rule is a pre-defined set of rules used to determine the reliability of different data sources in the assessment of each ability dimension. The confidence rule is set based on factors such as the objectivity of the data source, the controllability of the collection environment, and the clarity of the assessment indicators. For example, for the motor ability dimension, behavioral parameters are derived from objective records from visual sensors, resulting in a high confidence level; for the language ability dimension, behavioral observation data are derived from the daily observations of guardians, also resulting in a high confidence level.

[0063] During weighted processing, behavioral parameters and behavioral observation data are multiplied by their respective weighting coefficients, with the sum of these coefficients being 1. When no conflicts are identified, default weighting coefficients can be used; for example, both behavioral parameters and behavioral observation data can have a weighting coefficient of 0.5. When conflicts are identified, the weighting coefficients are adjusted according to confidence rules: data sources with higher confidence have their weighting coefficients increased, while data sources with lower confidence have their weighting coefficients decreased.

[0064] S304: Generate a capability development map of the target object based on weighted behavioral parameters and behavioral observation data.

[0065] In some embodiments, the weighted behavioral parameters and behavioral observation data are mapped to a comprehensive score for each capability dimension. The comprehensive score can be calculated using a weighted summation method, which involves multiplying each index value by its corresponding mapping weight and summing the results to obtain an initial score for each capability dimension. The initial score is then multiplied by the weight coefficients determined by the confidence rule to obtain the comprehensive score.

[0066] Optionally, the capability development map is a visual data structure representing the development level of the target object across various capability dimensions. It can be presented as a radar chart, where each axis corresponds to a capability dimension, and the scale value on the axis represents the overall score for that capability dimension. The overall scores of different capability dimensions are connected on the radar chart to form a closed polygon. The shape and area of ​​the polygon visually reflect the balanced development of the target object across various capability dimensions.

[0067] In one embodiment provided in this application, based on the above-described scheme, optionally, a capability development map of the target object is generated based on weighted behavioral parameters and behavioral observation data, including: The weighted behavioral parameters are converted into the first feature vector; The weighted behavioral observation data is converted into a second feature vector; The first feature vector and the second feature vector are concatenated to generate a fused feature vector; The fused feature vector is input into the pre-trained evaluation model, which outputs the evaluation score of the target object in each capability dimension. The evaluation model is trained based on sample data labeled with the evaluation results of each capability dimension. Based on the assessment scores, a capability development map of the target object is generated.

[0068] In this embodiment, the feature vector is a representation of structured data converted into a numerical vector form. The first feature vector consists of the values ​​of each indicator in the weighted behavioral parameters arranged in a preset order, such as [operational proficiency score, interest focus intensity, interaction duration, interaction frequency]. The second feature vector consists of the quantified values ​​obtained after semantic parsing of each dimension description in the weighted behavioral observation data, arranged in a preset order, such as [language ability quantified value, social ability quantified value, emotional response quantified value, interest preference quantified value, lifestyle habit quantified value].

[0069] Optionally, the fused feature vector is obtained by concatenating the first and second feature vectors along a vector dimension. If the first feature vector has a dimension of 4 and the second feature vector has a dimension of 5, then the fused feature vector has a dimension of 9. The fused feature vector integrates all information from the behavioral parameters and behavioral observation data, serving as input to the evaluation model.

[0070] Optionally, the evaluation model is a pre-trained deep learning model, and the model structure can be a multilayer perceptron or a convolutional neural network. The training process of the evaluation model is as follows: collect a sample dataset, which includes fused feature vector samples and their corresponding evaluation score labels for each capability dimension; input the fused feature vector samples into the evaluation model to be trained, and output the predicted evaluation score; calculate the loss value between the predicted evaluation score and the label; update the model parameters based on the loss value using the backpropagation algorithm, and iterate the training until the model converges.

[0071] In this embodiment, the evaluation score is obtained by the evaluation model through reasoning on the fused feature vector, and the output form is a numerical vector with the same number of ability dimensions, such as [language ability score, logical thinking ability score, exploratory ability score, social ability score, and motor ability score]. The evaluation score can be normalized to a range of 0 to 1, or 0 to 100, to facilitate subsequent comparison and visualization.

[0072] Specifically, the assessment scores can be directly mapped to numerical values ​​for each capability dimension in the capability development map, and presented in the form of radar charts or bar charts. The numerical values ​​in the capability development map maintain a linear correspondence with the assessment scores; the higher the assessment score, the higher the development level of the target object in that capability dimension.

[0073] In one optional embodiment, the evaluation model also incorporates timestamp features as auxiliary input during the training phase. Timestamp features represent the time window information corresponding to the fused feature vector. By concatenating the timestamp features with the fused feature vector and inputting the result into the evaluation model, the model can learn the patterns of ability development over time. This setup allows the evaluation model to incorporate the historical development trajectory of the target object during the inference phase, outputting more temporally consistent evaluation scores.

[0074] In one embodiment provided in this application, based on the above-described scheme, optionally, the developmental lag dimension of the target object is determined in each capability dimension based on the capability development map, including: Obtain the assessment scores of the target object across each capability dimension, as output by the assessment model; Obtain preset standard development reference data. The standard development reference data is determined based on the ability development distribution of the target group corresponding to the age of the target object. The standard development reference data includes the standard score range corresponding to each ability dimension. The assessment scores were compared with the standard score ranges for the corresponding competency dimensions. At least one capability dimension whose assessment score is below the lower limit of the standard score range is identified as a developmental lag dimension.

[0075] In this embodiment, the evaluation model performs inference based on the fused feature vector generated from the weighted behavioral parameters and behavioral observation data, and outputs an evaluation score vector with the same number of ability dimensions. For example, if the ability dimensions include five dimensions: language ability, logical thinking ability, exploratory ability, social ability, and motor ability, the evaluation model outputs a vector containing five evaluation scores. The value range of each evaluation score is preset to be from 0 to 100. The higher the score, the higher the development level of the target object in that ability dimension.

[0076] Optionally, the standard developmental reference data is a pre-constructed benchmark data used to measure whether the developmental level of the target subject is within the normal range. The standard developmental reference data is constructed as follows: Sample data is collected from a group of healthy children of the same age as the target subject. This sample data includes assessment scores for each ability dimension obtained through the aforementioned method of this application. Statistical analysis is performed on the sample data to calculate the score distribution characteristics of each ability dimension, including statistical measures such as mean, standard deviation, and percentiles. Based on these statistical measures, a standard score interval is set, which is used to define the range of values ​​for the normal developmental level. The standard score interval includes an upper limit and a lower limit. The lower limit is typically set as the 10th percentile or mean minus 1.5 times the standard deviation in the sample data, and the upper limit is typically set as the 90th percentile or mean plus 1.5 times the standard deviation. When the evaluation score falls within the standard score range, it indicates that the target object's development level in that ability dimension is within the normal range; when the evaluation score is below the lower limit of the range, it indicates that the target object's development in that ability dimension is lagging behind; when the evaluation score is above the upper limit of the range, it indicates that the target object's development in that ability dimension is advanced.

[0077] Furthermore, the standard developmental reference data is dynamically linked to the age of the target subjects. The standard developmental reference data differs for different ages because children's developmental abilities change systematically with age. In implementation, a standard developmental reference database covering ages 0 to 72 months can be pre-built. This database stores the standard score ranges for each ability dimension corresponding to each age. After obtaining the age information of the target subjects, the standard developmental reference data corresponding to their age is retrieved from the database. For any ability dimension, the assessment score for that dimension output by the assessment model is obtained, along with the corresponding standard score range from the standard developmental reference data. The relationship between the assessment score and the lower and upper limits of the standard score range is then determined.

[0078] Taking motor skills as an example, assuming the target subject's motor skills assessment score is 68, and the standard score range for motor skills corresponding to the target subject's age in months is [70, 95], then the comparison result is that the assessment score of 68 is lower than the lower limit of the standard score range of 70. Taking language skills as an example, assuming the target subject's language skills assessment score is 85, and the standard score range for language skills corresponding to the target subject's age in months is [60, 80], then the comparison result is that the assessment score of 85 is higher than the upper limit of the standard score range of 80.

[0079] The lagging development dimension refers to the capability dimensions in the capability development map whose assessment scores have not reached the lower limit of the normal development level. In the determination process, all capability dimensions are traversed, and capability dimensions whose assessment scores are lower than the lower limit of the corresponding standard score range are selected as lagging development dimensions.

[0080] When only one competency dimension has an assessment score below the lower limit of its standard score range, the developmental lag dimension is that single competency dimension. When the assessment scores of multiple competency dimensions are all below the lower limit of their respective standard score ranges, the developmental lag dimension is the set of multiple competency dimensions. When the assessment scores of all competency dimensions are not below the lower limit of their respective standard score ranges, there is no developmental lag dimension; in this case, the competency dimension with the lowest assessment score can be considered the developmental lag dimension.

[0081] In an optional embodiment, when multiple developmental lag dimensions exist, the method further includes a step of prioritizing each developmental lag dimension. The prioritization is based on the degree of lag and / or the developmental sensitive period for each developmental lag dimension. The degree of lag is represented by the absolute value of the difference between the assessment score and the lower limit of the standard score range; the larger the absolute value of the difference, the more severe the lag. The developmental sensitive period is determined based on the principles of child development psychology; different age groups have different sensitivities to different ability dimensions, and intervention during the sensitive period is more effective than during non-sensitive periods. The prioritization results are used to prioritize matching developmental lag dimensions with severe lag or those in the developmental sensitive period when subsequently identifying candidate toys.

[0082] Furthermore, when the assessment score exceeds the upper limit of the standard score range, it indicates that the target individual is developing advanced in that ability dimension. In an optional embodiment, the ability dimension in which the assessment score exceeds the upper limit of the standard score range is also defined as the developmental advancement dimension, and a prompt message is output to help the guardian understand the developmental advantages of the target individual.

[0083] In one embodiment provided in this application, based on the above-described solution, optionally, determining candidate toys that match the developmental lag dimension includes: Obtain the corresponding ability training tags for each toy from the toy resource library; the ability training tags correspond one-to-one with the ability dimensions. Identify candidate toys that match ability training labels with developmental lag dimensions, and generate the first set; Remove toys from the first set that have overlapping training labels with other ability dimensions (excluding the developmental lag dimension) to generate the second set; The toys in the second set were identified as candidate toys that matched the developmental lag dimension.

[0084] In this embodiment, the toy resource library is a pre-built collection of data storing toy information and its associated tags. Each toy entry in the toy resource library includes at least a toy identifier, toy name, toy description, applicable age range, and skills training tag.

[0085] The ability training label is an identifier that represents the ability dimension that the toy can promote development. The ability training label corresponds one-to-one with the ability dimension.

[0086] For example, if the ability dimensions include language ability, logical thinking ability, exploratory ability, social ability, and motor ability, then the ability training tags will correspondingly include language ability tags, logical thinking ability tags, exploratory ability tags, social ability tags, and motor ability tags. A toy can correspond to one ability training tag or multiple ability training tags. Taking a building block toy as an example, the toy can be labeled with both a motor ability tag (promoting the development of fine motor skills) and a logical thinking ability tag (promoting the development of spatial reasoning ability).

[0087] Specifically, for cases where the developmental lag dimension is a single ability dimension, toys with ability training labels matching that developmental lag dimension are identified as candidate toys. For cases where the developmental lag dimension is multiple ability dimensions, the matching strategy can be set as follows: toys with ability training labels matching any one of the developmental lag dimensions are identified as candidate toys, or toys with ability training labels matching all of the developmental lag dimensions are identified as candidate toys.

[0088] In this embodiment, the first set is a set of candidate toys selected through the matching operation described above. All toys in the first set have training capabilities for at least one developmental lag dimension.

[0089] Training overlap refers to the fact that the toy's ability training labels contain at least one label corresponding to a non-developmental lag dimension. A non-developmental lag dimension is any ability dimension other than the developmental lag dimension.

[0090] Specifically, for each toy in the first set, all ability training labels corresponding to that toy are obtained; it is determined whether there is at least one label in the toy whose corresponding ability dimension does not belong to the developmental lag dimension; if so, the toy is removed from the first set; if not, the toy is retained.

[0091] Taking the developmental lag dimension as the motor ability dimension as an example, toy A's ability training label is a motor ability label, toy B's ability training labels are both motor ability and language ability labels, and toy C's ability training label is a language ability label. In the matching operation, toy A and toy B are selected and included in the first set. In the elimination operation, toy B is eliminated because its ability training label includes a language ability label (not a developmental lag dimension), while toy A is retained because its ability training label only includes a motor ability label. This elimination mechanism is used to avoid overtraining of already achieved ability dimensions by recommended content, ensuring that recommended resources are focused and targeted.

[0092] The second set is the set of candidate toys remaining after the elimination operation.

[0093] In one optional embodiment, when the second set is empty, the elimination strategy can be adjusted by raising the elimination threshold to allow toys with slight overlap between ability training labels and non-developmental lag dimensions to remain in the candidate set; or, a manual intervention process can be initiated by sending a prompt message to the management end to request manual supplementation of recommended toys.

[0094] Furthermore, in cases where multiple developmental lag dimensions exist, if the second set contains toys that simultaneously correspond to multiple developmental lag dimensions, these toys will be marked as comprehensive recommendations and prioritized for display in subsequent recommendation output. Comprehensive recommendations can simultaneously promote the development of multiple lagging ability dimensions, exhibiting higher intervention efficiency compared to single-function toys.

[0095] and Figure 1 Corresponding to the method, this application also provides a toy recommendation device, the structural schematic diagram of which is shown below. Figure 4 As shown, it includes: The acquisition unit 401 is used to acquire behavioral video data and behavioral observation data corresponding to the target object. The behavioral video data includes the content of the target object interacting with the toy. The behavioral observation data is the behavioral performance data of the target object within a preset time period acquired through the user terminal. The behavior analysis unit 402 is used to perform behavior analysis on the behavior video data to obtain the behavior parameters of the target object in each ability dimension. The behavior parameters include operation proficiency, interest focus, interaction duration and interaction frequency. The generation unit 403 is used to generate a capability development map of the target object based on the behavioral parameters and the behavioral observation data; The first determining unit 404 is used to determine the development lag dimension of the target object in each of the capability dimensions based on the capability development map; The second determining unit 405 is used to determine candidate toys that match the development lag dimension; Output unit 406 is used to output recommendation information for the candidate toys.

[0096] This application also provides an electronic device, the structural schematic diagram of which is shown below. Figure 5 As shown, it specifically includes a memory 501 and one or more instructions 502, wherein one or more instructions 502 are stored in the memory 501 and configured to be executed by one or more processors 403 to perform the toy recommendation method described above.

[0097] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0098] Finally, it should be noted that in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0099] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0100] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0101] The solution provided in this application has been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A toy recommendation method, characterized in that, include: The system acquires behavioral video data and behavioral observation data corresponding to the target object. The behavioral video data includes the content of the target object interacting with the toy. The behavioral observation data is the behavioral performance data of the target object within a preset time period, acquired through a user terminal. Behavioral analysis is performed on the behavioral video data to obtain behavioral parameters of the target object in various capability dimensions, including operational proficiency, focus of interest, interaction duration, and interaction frequency. Generate a capability development map of the target object based on the behavioral parameters and the behavioral observation data; Based on the capability development map, the development lag dimension of the target object is determined in each of the capability dimensions; Identify candidate toys that match the aforementioned developmental lag dimension; Output recommendation information for the candidate toys.

2. The method according to claim 1, characterized in that, The step of performing behavioral analysis on the behavioral video data to obtain behavioral parameters of the target object across various capability dimensions includes: The behavioral video data is subjected to posture recognition to obtain the limb movements and hand operation trajectories of the target object; Based on the target object's limb movements and hand operation trajectories, determine the target object's operational proficiency in interacting with the toy; Eye tracking is performed on the behavioral video data to determine the target object's focus of interest during the interaction. The duration and number of interactions between the target object and the toy in the behavioral video data are statistically analyzed to determine the interaction duration and frequency.

3. The method according to claim 2, characterized in that, The determination of the target object's operational proficiency in interacting with the toy based on the target object's limb movements and hand operation trajectories includes: Determine the degree of matching between the target object's limb movements and hand operation trajectories and preset standard operation data; The matching degree is used to determine the target object's operational proficiency in interacting with the toy.

4. The method according to claim 1, characterized in that, The process of generating the capability development map of the target object based on the behavioral parameters and the behavioral observation data includes: Establish a mapping relationship between each capability dimension and the behavioral parameters and indicators in the behavioral observation data; The behavioral parameters and the behavioral observation data are input into a preset cross-validation model, which is used to identify whether there is a conflict between the evaluation results of the behavioral parameters and the behavioral observation data for the same ability dimension. In response to the identification of a conflict, the behavioral parameters and the behavioral observation data are weighted according to a preset confidence rule; Based on the weighted behavioral parameters and the behavioral observation data, a capability development map of the target object is generated.

5. The method according to claim 4, characterized in that, The process of generating a capability development map of the target object based on the weighted behavioral parameters and the behavioral observation data includes: The weighted behavioral parameters are converted into a first feature vector; The weighted behavioral observation data is converted into a second feature vector; The first feature vector and the second feature vector are concatenated to generate a fused feature vector; The fused feature vector is input into a pre-trained evaluation model, which outputs the evaluation score of the target object in each capability dimension. The evaluation model is trained based on sample data labeled with the evaluation results of each capability dimension. Based on the assessment score, a capability development map of the target object is generated.

6. The method according to claim 5, characterized in that, The process of determining the developmental lag dimension of the target object based on the capability development map in each of the capability dimensions includes: Obtain the evaluation scores of the target object on each capability dimension, as output by the evaluation model; Obtain preset standard development reference data, which is determined based on the ability development distribution of the target object's age group, and includes the standard score range for each ability dimension. The assessment scores are compared with the standard score ranges for the corresponding ability dimensions. At least one capability dimension whose assessment score is below the lower limit of the standard score range is identified as a developmental lag dimension.

7. The method according to claim 1, characterized in that, The process of identifying candidate toys that match the developmental lag dimension includes: Obtain the corresponding ability training tags for each toy from the toy resource library, and the ability training tags correspond one-to-one with the ability dimensions; Identify candidate toys whose ability training labels match the developmental lag dimension, and generate a first set; Remove toys from the first set whose ability training labels overlap with other ability dimensions besides the developmental lag dimension, and generate a second set; The toys in the second set are identified as candidate toys that match the developmental lag dimension.

8. A toy recommendation device, characterized in that, include: The acquisition unit is used to acquire behavioral video data and behavioral observation data corresponding to the target object. The behavioral video data includes the content of the target object interacting with the toy. The behavioral observation data is the behavioral performance data of the target object within a preset time period, acquired through a user terminal. The behavior analysis unit is used to perform behavior analysis on the behavior video data to obtain the behavior parameters of the target object in various ability dimensions. The behavior parameters include operation proficiency, interest focus, interaction duration and interaction frequency. A generation unit is configured to generate a capability development map of the target object based on the behavioral parameters and the behavioral observation data. The first determining unit is used to determine the development lag dimension of the target object in each of the capability dimensions based on the capability development map; The second determining unit is used to determine candidate toys that match the development lag dimension; The output unit is used to output recommendation information for the candidate toys.

9. A storage medium, characterized in that, The storage medium includes storage instructions, wherein, when the instructions are executed, the device containing the storage medium is controlled to perform the toy recommendation method as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes a memory, and one or more instructions, wherein one or more instructions are stored in the memory and configured to be executed by one or more processors as described in any one of claims 1 to 7.

Images