A perception fusion and multi-scene optimization evaluation method for children's lunch break desk and chair

Abstract

The application discloses a kind of children's lunch desk and chair perception fusion and multi-scene optimization evaluation method, including, the evaluation system of multiple dimensions is built, the score data of student and expert in learning and lunch break etc. Multiple scene data are collected, entropy weight method is used to fuse double subject data, obtain comprehensive score, the data after fusion is constructed as "product-dimension-scene" three-order tensor and factor decomposition is carried out, scene potential factor is extracted, scene factor and other evaluation dimensions are used as input characteristics, machine learning regression model is constructed to predict comfort, uses shapley additive interpretation technique to analyze model, quantifies the contribution of each design dimension and its interaction term to comfort, and generates structure and function optimization suggestions;The application realizes the scientization and objectivity of evaluation process, can accurately locate the key design factors affecting children's experience, finally outputs optimization suggestions, provides efficient and reliable decision support for the research and development and selection of children's desk and chair.

Description

Technical Field

[0001] This invention relates to the fields of ergonomics, educational equipment and intelligent assessment technology, and in particular to a method for sensory fusion and multi-scenario optimization evaluation of children's nap desks and chairs. Background Technology

[0002] As the Ministry of Education continues to strengthen the management of lunch break environments in primary and secondary schools, desks and chairs, as essential equipment for students' daily learning and lunch breaks, are receiving increasing attention. Existing evaluation methods for desks and chairs are mainly based on adult ergonomic standards, emphasizing static performance parameters such as structure and materials. However, when applied to children, especially lower-grade students, these evaluation methods fail to fully consider children's comfort, perceptual differences, and emotional needs in different usage scenarios, resulting in an inaccurate assessment of the desks and chairs' performance in actual use.

[0003] Currently, commonly used methods for evaluating children's desks and chairs include traditional ergonomic assessments, subjective ratings, and behavioral observation. While these methods can assess the comfort and safety of desks and chairs, they are usually limited to static evaluations and lack dynamic adaptation analysis for use in multiple scenarios (such as studying, resting, and adjusting). Furthermore, there are often significant discrepancies between expert evaluations and student evaluations, and traditional methods lack effective scoring integration mechanisms, resulting in an inability to achieve a comprehensive and accurate overall evaluation.

[0004] To address these issues, artificial intelligence, particularly artificial neural networks (ANNs), has been applied to solving nonlinear problems in recent years. In children's desk and chair design, ANNs can use existing data to train and obtain a comfort assessment model for the desks and chairs, and can overcome the accuracy limitations of traditional methods by training the scoring system. However, existing ANN-based evaluation methods often lack comprehensive analysis adapting to multiple dimensions and scenarios, failing to achieve a complete closed loop from data collection to optimization feedback. Furthermore, their internal decision-making logic is often like a "black box," failing to clearly reveal the contribution of each design dimension and its interaction effects to the end-user experience, thus making it difficult to provide a clearly directional and interpretable optimization basis for the desk and chair design and iteration process. Summary of the Invention

[0005] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0006] In view of the aforementioned existing problems, this invention is proposed. Therefore, this invention provides a perceptual fusion and multi-scenario optimization evaluation method for children's nap desks and chairs to solve the problems mentioned in the background art.

[0007] To address the aforementioned technical problems, this invention provides the following technical solution: a method for sensory fusion and multi-scenario optimization evaluation of children's nap desks and chairs, comprising:

[0008] A multi-dimensional evaluation system for children's desks and chairs is established, which consists of evaluation dimensions in the structural domain, operational domain, perceptual domain, and affective domain.

[0009] In multiple preset usage scenarios, student groups and expert groups respectively collected rating data on the children's desks and chairs in various evaluation dimensions;

[0010] Based on the differences in the distribution of student rating data and expert rating data across various evaluation dimensions, the two are dynamically weighted and fused to generate a fused score that comprehensively reflects the perceptions of both parties.

[0011] Construct a third-order rating tensor consisting of desks and chairs, evaluation dimensions, and usage scenarios, and perform factorization on the tensor;

[0012] Using the comfort dimension score in the fusion score as the prediction target, and the scores of the other evaluation dimensions and their interaction items as input features, a regression model is trained to establish a mapping relationship between the input features and the prediction target;

[0013] Interpretability analysis is used to analyze the trained regression model, quantify the contribution of each input feature to the comfort dimension score, and output optimization suggestions based on the contribution analysis results.

[0014] As a preferred embodiment of the sensory fusion and multi-scenario optimization evaluation method for children's nap desks and chairs described in this invention, the evaluation dimensions include: ergonomics, safety, ease of learning, convenience, perception, aesthetics, friendliness, and comfort.

[0015] As a preferred embodiment of the sensory fusion and multi-scenario optimization evaluation method for children's nap desks and chairs described in this invention, the preset multiple usage scenarios include at least: a learning scenario for evaluating daily reading and writing functions, an adjustment scenario for evaluating structural adjustment experience, and a nap scenario for evaluating the effect of lying down for relaxation.

[0016] As a preferred embodiment of the sensory fusion and multi-scenario optimization evaluation method for children's nap desks and chairs described in this invention, the dynamic weighted fusion step includes:

[0017] Calculate the Jensen-Shannon divergence between the student rating distribution and the expert rating distribution to quantify the perceived discrepancy between the two across various evaluation dimensions;

[0018] The perceived divergence is taken as input and converted into a dynamic weight value through a preset nonlinear mapping function;

[0019] The student scores and expert scores are weighted and summed using the dynamic weight values ​​to obtain the fusion score.

[0020] As a preferred embodiment of the perception fusion and multi-scenario optimization evaluation method for children's nap desks and chairs described in this invention, the adjustment mechanism of the dynamic weight value is as follows: the greater the perception discrepancy, the lower the proportion of the expert score in the weighted summation; the smaller the perception discrepancy, the higher the proportion of the expert score in the weighted summation.

[0021] As a preferred embodiment of the perceptual fusion and multi-scenario optimization evaluation method for children's nap desks and chairs described in this invention, the tensor is factored by:

[0022] The third-order rating tensor was decomposed using a parallel factor analysis method to obtain factor loading matrices associated with desks and chairs, evaluation dimensions, and usage scenarios, respectively, and sensitive evaluation factors were identified based on the factor loading matrices.

[0023] As a preferred embodiment of the perception fusion and multi-scenario optimization evaluation method for children's nap desks and chairs described in this invention, wherein: when training the regression model, the scores of the dimensional interaction items in the input features are calculated based on a preset dimensional combination.

[0024] As a preferred embodiment of the perception fusion and multi-scenario optimization evaluation method for children's nap desks and chairs described in this invention, the training regression model adopts the gradient boosting decision tree algorithm to model and fit the nonlinear relationship between the input features and the prediction target.

[0025] As a preferred embodiment of the perception fusion and multi-scenario optimization evaluation method for children's nap desks and chairs described in this invention, the dimensional interaction items include at least one of the following three:

[0026] Perceptual-ergonomic interaction terms used to characterize the synergistic effect of material tactile sensation and structure;

[0027] Safety-convenience interaction term used to characterize the combined effects of structural reliability and ease of operation;

[0028] Aesthetic-affinity interaction items used to reveal the combined effect of appearance design and emotional connection.

[0029] As a preferred embodiment of the sensory fusion and multi-scenario optimization evaluation method for children's nap desks and chairs described in this invention, the interpretability analysis technique employs the Shapley additive interpretation method, including:

[0030] Calculate the SHAP value of each input feature, evaluate the positive and negative impacts and marginal contributions of each evaluation dimension and its interaction items on user comfort, and based on the ranking of the SHAP values, identify the most critical design factors of desks and chairs and generate suggestions for improving the structure and function of desks and chairs.

[0031] Compared with existing technologies, the beneficial effects of this invention are as follows: Based on children's multidimensional needs, an evaluation system for school desks and chairs with eight dimensions across four categories—structure, operation, perception, and emotion—is constructed. It also integrates expert and student ratings, resolving the problem of perceptual conflict. Furthermore, tensor decomposition is used to analyze the adaptability of school desks and chairs in three scenarios: learning, adjustment, and lunch break. Simultaneously, the SHAP method is used to explain factors influencing comfort, thereby outputting optimization suggestions for the structure and function of school desks and chairs. This invention achieves a closed-loop process from multi-source rating fusion to intelligent feedback, possessing the advantages of comprehensive evaluation, reasonable integration, interpretable analysis, and deployable application. It is suitable for the design, development, and selection decisions of school desks and chairs for children's lunch break. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the 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. Wherein:

[0033] Figure 1 This is a flowchart illustrating the overall process of the sensory fusion and multi-scenario optimization evaluation method for children's nap desks and chairs according to an embodiment of the present invention.

[0034] Figure 2 This is a dimensional structure diagram of the evaluation system (PLEASURC model) for children's nap desks and chairs, which is based on the perception fusion and multi-scenario optimization evaluation method of children's nap desks and chairs according to an embodiment of the present invention.

[0035] Figure 3 This is a scene sensitivity map of tensor decomposition for the perception fusion and multi-scene optimization evaluation method of children's nap desks and chairs according to an embodiment of the present invention.

[0036] Figure 4 This is a scene adaptation matrix diagram of tensor decomposition for the perception fusion and multi-scene optimization evaluation method of children's nap desks and chairs according to an embodiment of the present invention.

[0037] Figure 5The learning curve of the XGBoost model for the perception fusion and multi-scenario optimization evaluation method of children's nap desks and chairs according to an embodiment of the present invention is shown.

[0038] Figure 6 The graph shows the convergence curve of the XGBoost model relative to the L2 error of the perception fusion and multi-scene optimization evaluation method for children's nap desks and chairs according to an embodiment of the present invention.

[0039] Figure 7 This is a SHAP analysis result of the sensory fusion and multi-scenario optimization evaluation method for children's nap desks and chairs according to an embodiment of the present invention.

[0040] Figure 8 This diagram illustrates the impact of the SHAP value on the XGBoost model output of the perception fusion and multi-scene optimization evaluation method for children's nap desks and chairs according to an embodiment of the present invention. Detailed Implementation

[0041] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0042] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0043] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0044] This invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of this invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not adhering to the usual scale. Furthermore, the schematic diagrams are merely examples and should not be construed as limiting the scope of protection of this invention. In actual fabrication, the three-dimensional spatial dimensions of length, width, and depth should be included.

[0045] Furthermore, in the description of this invention, it should be noted that the terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used solely for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. In addition, the terms "first," "second," or "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0046] Unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" in this invention should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; similarly, they can refer to mechanical connections, electrical connections, or direct connections, or indirect connections through an intermediate medium, or internal connections between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0047] Example 1

[0048] Reference Figures 1 to 8 This is the first embodiment of the present invention, which provides a method for perceptual fusion and multi-scenario optimization evaluation of children's nap desks and chairs, including:

[0049] S1. Establish a multi-dimensional evaluation system for children's desks and chairs, which consists of evaluation dimensions in the structural domain, operational domain, perceptual domain, and affective domain;

[0050] Specifically, refer to Figure 2 This invention first constructs a multi-dimensional evaluation system for children's desks and chairs called PLEASURC. This evaluation system aims to systematically divide the key factors affecting children's user experience into four domains, which are further subdivided into eight specific evaluation dimensions. The evaluation system consists of evaluation dimensions from the structural domain, operational domain, perceptual domain, and affective domain, as detailed below:

[0051] Structural Domain: This domain focuses on the fundamental physical properties and safety features of desks and chairs, ensuring they meet the physiological developmental needs of children and provide a safe physical environment. This domain includes two evaluation dimensions:

[0052] Ergonomics: Defined as the adaptability and support of desks and chairs for children's sitting posture, spinal health and body support. Its evaluation criteria include, but are not limited to: whether the height, depth, width and back curvature of the seat match the anthropometric data of children in the target age group; whether the height of the desktop is conducive to maintaining the correct reading and writing posture; whether the desks and chairs can effectively distribute body pressure in different scenarios and reduce muscle fatigue or bone development risks caused by prolonged use.

[0053] Safety: Defined as the comprehensive performance of desks and chairs in terms of structural stability, environmental friendliness of materials, and protection against physical injury. The evaluation criteria include: whether the overall structure of the chair is stable and there is no risk of tipping over or collapsing under normal use and foreseeable misuse; whether all materials meet the national standards for the limit of toxic and harmful substances in children's furniture; whether the edges and corners of the desks and chairs are rounded and whether there are sharp edges, corners, or dangerous gaps that may pinch or scratch children.

[0054] Operational Domain: This domain focuses on the interactive experience of adjusting the functions of desks and chairs, measuring the ease and effectiveness of children's independent operation. The operational domain includes two evaluation dimensions:

[0055] Ease of learning: Defined as the ease with which children understand the various adjustment functions of desks and chairs when they first come into contact with them. The assessment criteria include: whether the labels on the adjustment devices are clear and intuitive; whether the operating logic conforms to children's cognitive habits; and whether children can quickly master the adjustment methods of functions such as height and angle with little or no guidance.

[0056] Convenience: Defined as the efficiency, accuracy, and fault tolerance of children in operating various functions during daily use; its evaluation criteria include: whether the adjustment process is effortless and smooth; whether the locking and unlocking mechanism is reliable; and whether the switching operation between different modes such as learning and nap is simple and quick, and not prone to misoperation.

[0057] The sensory domain focuses on the direct sensory input that desks and chairs provide to children through physical media (materials, shapes, colors, etc.). This sensory domain includes two evaluation dimensions:

[0058] Sensory experience: defined as the tactile and thermal experience that the material surface of desks and chairs brings to children. Its evaluation criteria include: whether the material of the main contact surfaces such as seats, backrests, and armrests is skin-friendly and soft; whether the material has good breathability to avoid stuffiness from sitting for a long time; and whether the material surface is too cold or too hot under different ambient temperatures, affecting physical comfort.

[0059] Aesthetics: Defined as the visual appeal of the overall shape, color scheme, and design details of the desks and chairs to children. The evaluation criteria include: whether the color scheme of the desks and chairs is lively and harmonious, and in line with children's aesthetic preferences; whether the shape design is interesting and modern; and whether the proportions and lines of each component are harmonious and unified.

[0060] The emotional domain focuses on the psychological and emotional responses of children during the use of desks and chairs, representing a higher level of user experience. The emotional domain includes two evaluation dimensions:

[0061] Affinity: Defined as whether the design of desks and chairs can evoke emotional resonance, a sense of identity and belonging in children. The evaluation criteria include: whether the design elements incorporate themes that children like, such as cartoons and nature; whether the desks and chairs create a friendly and warm atmosphere that makes children willing to approach and use them; and whether the design of the desks and chairs reflects care for children's psychological needs.

[0062] Comfort: Defined as the comprehensive subjective feeling of children throughout the use of desks and chairs, it is the sum of physical relaxation and psychological pleasure. As the final core evaluation indicator, its evaluation criteria rely on the user's subjective reports and comprehensively reflect the final experience effect under the combined effect of the aforementioned seven dimensions, such as the degree of fatigue after long-term study, the degree of relaxation during lunch break, and overall user satisfaction.

[0063] S2. In multiple preset usage scenarios, collect rating data from student groups and expert groups on the children's desks and chairs in various evaluation dimensions;

[0064] Specifically, to ensure the comprehensiveness and authenticity of the aforementioned evaluation system, this invention simulates typical activities of children during a school day, setting at least three core usage scenarios to cover the main functional states of children's nap desks and chairs. These three scenarios are as follows:

[0065] Learning Scenario: This scenario is used to evaluate the functional performance of desks and chairs under standard learning conditions such as daily reading, writing, and listening. In this scenario, participants are required to adjust the desks and chairs to the learning mode and perform tasks such as reading, writing, or drawing for a continuous period of time. The aim is to focus on examining the ergonomic support for maintaining correct posture, the comfort performance under long-term concentration, and safety factors such as the stability of the desks and chairs.

[0066] Adjustment Scenario: This scenario is used to evaluate the user experience when switching desks and chairs from one state to another (e.g., from study mode to lunch break mode). Participants need to manually operate the various adjustment mechanisms of the desks and chairs, such as raising and lowering the desktop height, tilting the backrest angle, and extending and retracting the footrest. This scenario aims to evaluate the ease of learning and convenience of the adjustment process, including whether the operation is intuitive and effortless, and whether there are any safety hazards.

[0067] Lunch break scenario: This scenario is used to evaluate the relaxation effect and comfort of desks and chairs when fully extended into a reclining position. Participants need to adjust the desks and chairs to the lunch break mode and lie down to rest. This scenario aims to evaluate its functionality as a "bed", paying particular attention to the ergonomics in the reclining position, the perception of materials (such as breathability and skin-friendly feel), the safety of the overall structure (such as the risk of tipping over), and the final comfort and friendliness (sense of security).

[0068] Furthermore, in the aforementioned core usage scenarios, student groups and expert groups respectively collected rating data on the children's desks and chairs in each evaluation dimension;

[0069] It should be noted that the data collection work was carried out in parallel by student groups and expert groups in order to obtain complementary data from users' subjective perceptions and professional objective judgments;

[0070] Specifically, for the collection of student data, a number of primary school students who fit the target age group for desks and chairs were selected as subjects;

[0071] In addition, in order to eliminate the interference of the order effect on the scoring results, the present invention also adopts a Latin square experimental design to ensure that the order in which each student experiences all the desks and chairs to be evaluated is balanced. That is, in each preset scenario, after completing the specified task, students need to independently rate the current desk and chair on the eight dimensions defined by the PLEASURC model (for example, using a 10-point scale) based on their own true feelings. This process collects the most direct user experience data.

[0072] Specifically, for the collection of expert group data, an evaluation team was formed by inviting experts with professional knowledge backgrounds in fields such as human factors engineering, industrial design, and child safety. Each expert scored the performance of each desk and chair in various scenarios according to a standardized functional observation and testing process. Moreover, the experts' evaluation was not only based on subjective observation, but also combined with the measurement and analysis of objective parameters such as the size, structure, and materials of the desks and chairs. Their scores focused more on technical specifications, design principles, and potential risks.

[0073] It should be noted that the data collected above can be processed through classic standardization operations to form a multi-dimensional original scoring dataset;

[0074] S3. Based on the distribution differences between student rating data and expert rating data across various evaluation dimensions, the two are dynamically weighted and fused to generate a fused score that comprehensively reflects the perceptions of both parties.

[0075] It should be noted that after obtaining the original rating data from students and experts, the core purpose of this step is to resolve the "perceptual conflict" caused by the different perspectives of the two groups, and to generate a single, reliable integrated rating that can reflect the children's real experience and incorporate the experts' professional judgment. Since traditional static weighting cannot adapt to the changes in perceptual differences on different evaluation dimensions, this invention adopts a dynamic weighted integration mechanism, the weight allocation of which is directly related to the degree of perceptual disagreement between the two parties on that dimension.

[0076] Specifically, this invention introduces Jensen-Shannon Divergence (JSD) from information theory to quantify the differences in the rating distributions between the student group and the expert group on each evaluation dimension. JSD is a symmetric and smooth method for measuring the similarity between two probability distributions, and its mathematical formula is as follows:

[0077]

[0078] Where, α i To evaluate the expert weights for dimension i, JSD i , where represents the Jensen-Shannon divergence between experts and students on evaluation dimension i, and K is a sensitivity control parameter, typically ranging from 5 to 10.

[0079] Furthermore, the set of all student ratings under a specific dimension is regarded as one probability distribution, and the set of all expert ratings is regarded as another probability distribution. The JSD value between the two is calculated. The larger the value, the greater the discrepancy between the students' and experts' perceptions of that dimension, and the more different the evaluation conclusions are. Conversely, the smaller the JSD value, the more consistent the perceptions of both parties are, and the more similar the evaluation conclusions are.

[0080] Specifically, for each dimension i, an expert score P is calculated. i Student rating Q i Distributed JSD:

[0081]

[0082] in, Represented as a dynamic weight adjustment term, D KL Represented as KL divergence;

[0083] Furthermore, this invention also constructs a dynamic weight adjustment mechanism, the allocation logic of which follows the following principles:

[0084] When the discrepancy between students' and experts' perceptions is small, it means that the experts' professional judgments are basically consistent with the students' actual experiences. At this point, the experts' scores are given higher weight because their evaluations not only align with user feelings but also include deeper technical insights and security considerations.

[0085] When there is a significant discrepancy between the perceptions of students and experts, it indicates that the experts' judgments may be out of touch with the actual user experience of children (for example, a design that is technically perfect but that children find unfun or unfriendly). In this case, the weight of the expert rating should be reduced, and the student rating should be given higher weight to ensure that the final evaluation results can truly reflect the core needs of the target users.

[0086] Furthermore, to achieve the aforementioned dynamic weight adjustment mechanism, this invention also takes the perceived divergence (JSD value) calculated in the previous step as input and converts it into a dynamic expert weight value through a preset nonlinear mapping function. This function is designed to be a monotonically decreasing function, that is, the larger the JSD value, the smaller the output weight value, and vice versa. At the same time, the weight value of this function is constrained within the interval [0,1].

[0087] Specifically, after obtaining the dynamic expert weight value, the student weight value is 1 minus the remaining value of the dynamic expert weight value. Then, the average student score and the average expert score are weighted and summed using the obtained dynamic weight values ​​to calculate the final integrated score.

[0088] It should be noted that the above nonlinear mapping function uses the Sigmoid function to ensure that the fused score reflects both the expert's professional judgment and the student's actual user experience. The fused score is calculated using the following formula:

[0089] S i =α i ·P i +(1-α i )·Q i

[0090] Among them, S i Let α represent the fusion score for evaluation dimension i. i Represented as dynamic expert weight values;

[0091] S4. Construct a third-order rating tensor consisting of desks and chairs, evaluation dimensions, and usage scenarios, and perform factorization on the tensor;

[0092] It should be noted that after obtaining the integrated score that can comprehensively reflect the perceptions of students and experts, this step aims to reveal from a high dimension which key evaluation dimensions dominate the user experience of different desks and chairs in different scenarios. Furthermore, since traditional two-dimensional matrix analysis (such as product-dimension) cannot simultaneously handle the variable of "scenario", this invention introduces tensor analysis, which can model and analyze the complex interaction relationship between "product-dimension-scenario" within a unified mathematical framework.

[0093] Furthermore, the generated fusion score is organized into a third-order tensor, denoted as X, where the three orders of this tensor are:

[0094] The first level is the evaluation dimension, which has a length of 8, corresponding to the eight evaluation dimensions in the PLEASURC evaluation system.

[0095] The second tier consists of school desks and chairs. The length of this tier depends on the number of school desks and chairs participating in the evaluation (set to 5 in this embodiment).

[0096] The third stage is the usage scenario, which has a length of 3 and corresponds to the three preset scenarios of learning, adjustment, and lunch break;

[0097] Specifically, this third-order tensor can be represented as:

[0098]

[0099] Where I is the number of evaluation dimensions, J is the number of types of desks and chairs, and K is the number of usage scenarios. This tensor is used to describe the rating data of each product in each dimension and in different scenarios. Therefore, the size of this tensor is 8×5×3. The value of each element in the tensor represents the fusion score obtained by the j-th desk and chair in the k-th usage scenario on the i-th evaluation dimension.

[0100] Furthermore, in order to extract meaningful latent structures from this complex multidimensional data volume, this invention employs a parallel factor analysis method to decompose the third-order tensor. PARAFAC is a method that decomposes a tensor into the sum of three factor loading matrices. Its core idea is to find a set of R latent factors to best reconstruct the original tensor. Its decomposition form can be expressed as:

[0101]

[0102] Where R is the rank of the decomposition, representing the number of latent factors extracted. In this invention, the value of R is set to 2 to ensure a balance between low reconstruction error and high model interpretability; a r b r c r These are scenario factors, product factors, and dimensional factors. Represents the outer product of vectors.

[0103] Furthermore, through decomposition, we can obtain three factor loading matrices:

[0104] The scene factor loading matrix has a size of 3×R. Each column of the matrix corresponds to a latent factor, and the element value in the column represents the correlation strength between the three scenes "learning", "adjustment" and "lunch break" and the latent factor.

[0105] Product factor loading matrix: Its size is 5×R, and each column of the matrix represents the "score" or performance of five products on the corresponding potential factor;

[0106] Dimension factor loading matrix: Its size is 8×R. Each column of the matrix shows the contribution of the eight evaluation dimensions to the latent factor.

[0107] It should be noted that the ultimate goal of tensor decomposition is to interpret these latent factors and identify scene-sensitive factors. By analyzing the three loading matrices, patterns hidden behind the data can be revealed, as shown in the appendix of this invention. Figure 3 and attached Figure 4 As shown, by analyzing the decomposition results of rank 2, the following conclusions can be drawn:

[0108] Potential Factor 1, in the scenario factor matrix, is highly correlated with "writing scenario" and "adjustment scenario". At the same time, in the dimension factor matrix, the dimensions strongly correlated with this factor are mainly ergonomics, safety, ease of learning and convenience. This indicates that this factor can be interpreted as a task-oriented functional factor. When children are actively learning or operating desks and chairs, their experience is mainly determined by these basic functional attributes of the structural domain and the operational domain.

[0109] Potential Factor 2: In the scenario factor matrix, this factor has the highest correlation with the "naptime scenario". In the dimension factor matrix, the dimensions strongly correlated with it are perception, aesthetics, affinity and comfort. Therefore, this factor can be interpreted as the emotional aspect of the rest experience. When children enter a relaxed and restful naptime state, their overall feeling is more influenced by experiential factors such as the perceptual and emotional domains.

[0110] S5. Using the comfort dimension score in the fusion score as the prediction target, and the scores of the other evaluation dimensions and their interaction items as input features, train a regression model to establish a mapping relationship between the input features and the prediction target.

[0111] It should be noted that after identifying the dominant evaluation factors in different scenarios through tensor decomposition, this step aims to analyze the impact of specific evaluation dimensions and their interactions on the user's final core experience—"comfort"; and, in order to predict this comfort score, this invention constructs a multi-level input feature set (i.e., independent variables), which not only includes basic dimensions but also introduces the interaction effects between dimensions.

[0112] Specifically, the fusion scores of the other seven dimensions besides comfort in the evaluation system are used as the basic input features. Furthermore, since a single dimension's score cannot completely determine the final level of comfort, the synergistic or restrictive effects between different dimensions are equally crucial. For example, excellent material feel (high perception) must be combined with an ergonomic structure (high ergonomics) to maximize the comfort experience. Therefore, this invention also pre-defines and calculates scores for multiple dimensional interaction items with clear physical or emotional meanings, and incorporates them into the input feature set. In this embodiment, at least the following three items are included:

[0113] The Perceptual-Ergonomic Interaction Item quantifies the synergy between material tactile qualities (such as softness, hardness, skin-friendliness, and breathability) and structural design (such as curved surfaces and support). Its score is obtained by multiplying the original scores of ergonomics and perceptuality.

[0114] The safety-convenience interaction item is used to characterize the combined impact of structural reliability and ease of operation. A structure that is easy to adjust but has safety hazards will inevitably have a poor overall experience. Similarly, its score is obtained by multiplying the original scores of safety and convenience.

[0115] The Aesthetics-Friendliness Interaction Item reveals the combined effect of appearance design (visual appeal) and emotional connection (psychological identification), measuring whether a product, in addition to being "good-looking," has further achieved "friendliness." Similarly, its score is obtained by multiplying the original scores of aesthetics and friendliness.

[0116] Furthermore, in order to effectively capture the complex and non-linear mapping relationship between the aforementioned input features and comfort scores, this invention preferably employs the Gradient Boosting Decision Tree (GBDT) algorithm to train the regression model. In this embodiment, a highly efficient implementation of this algorithm—XGBoost (eXtremeGradient Boosting)—is specifically used. This algorithm is chosen because it can automatically handle interactions between features, has a strong ability to fit non-linear relationships in the data, exhibits high model robustness, and demonstrates superior predictive performance. The training process for this model is as follows:

[0117] First, the fused scoring dataset generated in step S3, containing all samples (all products and all scenarios), is divided into a training set and a validation set according to a certain ratio. On the training set, five-fold cross-validation is used to train the model to enhance its generalization ability and prevent overfitting. Then, through iterative training, the model continuously learns the mapping rules from input features (7 basic dimensions + 3 interaction terms) to the prediction target (comfort score). Finally, the model's performance is evaluated on the validation set, as shown in the appendix. Figure 5 and attached Figure 6 As shown, after about 1000 rounds of training, the training loss and validation loss of the model converged to a low level (about 0.15 to 0.2), and the two curves fit well, indicating that the model did not have obvious overfitting. At the same time, the relative L2 error also converged steadily to about 2.3% to 3.1%, which fully proves that the trained regression model can predict the comfort score of children's desks and chairs with high accuracy.

[0118] S6. Use interpretability analysis technology to analyze the trained regression model, quantify the contribution of each input feature to the comfort dimension score, and output optimization suggestions based on the contribution analysis results.

[0119] It should be noted that after constructing and validating an XGBoost regression model that predicts comfort scores, the model itself is a complex ensemble model, and its internal decision-making logic is opaque to the designer, i.e., it has a "black box" problem. Therefore, in order to transform the predictive power of the model into a design with "transparent" meaning, the core task of this step is to interpret and analyze the model to identify the key design factors that have the greatest impact on children's comfort experience. To solve this problem, this invention uses the Shapley Additive Interpretation (SHAP) method to analyze the trained XGBoost model. This method is derived from the Shapley value theory in game theory. Its core advantage is that it can provide a fair, consistent, and theoretically sound explanation for each prediction of the model. It can not only evaluate the global importance of each input feature (i.e., evaluation dimensions and their interaction terms) to the final prediction result, but also reveal how each feature value pushes the prediction result from the baseline value (the average prediction value of all samples) to the final value in each specific prediction. That is, it quantifies the marginal contribution of each feature, including the direction (positive or negative) and magnitude of its influence.

[0120] Furthermore, the trained XGBoost model and the complete test dataset are used as inputs, and the SHAP analysis tool is applied to calculate the SHAP value of each input feature for each sample in the dataset. These SHAP values ​​form the basis of the model interpretation.

[0121] Furthermore, to intuitively understand the analysis results, this invention uses SHAP values ​​to generate a SHAP analysis result graph. In this graph, features are sorted from top to bottom according to their average absolute SHAP values. The feature at the top has the greatest average impact on the model's predicted comfort score and is the most critical factor determining user experience. Each feature corresponds to a row of scatter points, with each point representing a sample. The position of the point on the horizontal axis represents the feature's SHAP value within that sample. A positive SHAP value indicates that the feature's value increases the predicted comfort score, while a negative value indicates a decrease. The density of the points reflects the distribution of the samples. The color of the points in the graph typically represents the original value of the feature (e.g., red represents a high score, blue represents a low score). By observing the relationship between color and SHAP value, one can gain a deeper understanding of how features affect the results. For example, if the red dots (high scores) of a feature are predominantly distributed in the region with positive SHAP values, it indicates that the higher the feature score, the more positively it improves comfort.

[0122] Specifically, in this embodiment, it can be achieved through the appendix Figure 7 and attached Figure 8 As we can see, the perceptual-ergonomic interaction item ranks first in the importance of features. This indicates that simply having excellent materials or a reasonable structure is not enough to guarantee optimal comfort; the synergistic effect of the two is the decisive factor. Furthermore, the tactile feel, softness, and breathability of the materials must be perfectly combined with the seat's curvature and support point design to maximize the child's lying and learning experience. Safety and aesthetics follow closely behind, ranking second and third respectively. This shows that after meeting basic ergonomic needs, an environment that makes children feel absolutely safe and free of potential dangers, along with a visually pleasing design that aligns with their aesthetic preferences, are important drivers for improving their psychological comfort and overall satisfaction. Other dimensions such as convenience and friendliness also have varying degrees of impact on comfort.

[0123] Furthermore, based on the contribution ranking and influence patterns of each input feature revealed by the SHAP analysis above, these data-driven insights are transformed into specific and actionable product optimization suggestions. This process follows the principle of "focusing on the big picture and letting go of the small details," prioritizing improvement directions for the key factors with the highest contribution.

[0124] Regarding perceptual and ergonomic interaction, given its paramount importance, it is recommended that designers consider material selection and structural design as a holistic system during the iteration of desks and chairs. For example, for a product with an excellent backrest curve design but user feedback indicating low comfort during lunch breaks, the optimization direction should not be to redesign the structure, but rather to focus on testing soft padding materials of different densities, thicknesses, and breathability to find the "golden combination" that best matches the existing curve, and vice versa.

[0125] Regarding safety, it is recommended to conduct child-induced operation tests on the adjustment mechanisms of all school desks and chairs, optimize the reliability of locking devices, perform secondary chamfering on all edges and corners to ensure that the smoothness meets higher standards, and conduct more stringent volatile organic compound testing on new materials to eliminate health hazards from the source.

[0126] Regarding aesthetics, it is recommended to introduce more thematic and story-based elements in color design, rather than simply piecing together color blocks. In terms of shape, we can draw inspiration from the shapes of animals and plants that children like, and carry out biomimetic design to enhance the visual appeal and emotional connection of the product, which can also improve its friendliness.

[0127] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0128] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0129] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0130] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0131] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0132] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for sensory fusion and multi-scenario optimization evaluation of children's nap desks and chairs, characterized in that, include: A multi-dimensional evaluation system for children's desks and chairs is established, which consists of evaluation dimensions in the structural domain, operational domain, perceptual domain, and affective domain. In multiple preset usage scenarios, student groups and expert groups respectively collected rating data on the children's desks and chairs in various evaluation dimensions; Based on the differences in the distribution of student rating data and expert rating data across various evaluation dimensions, the two are dynamically weighted and fused to generate a fused score that comprehensively reflects the perceptions of both parties. Construct a third-order rating tensor consisting of desks and chairs, evaluation dimensions, and usage scenarios, and perform factorization on the tensor; Using the comfort dimension score in the fusion score as the prediction target, and the scores of the other evaluation dimensions and their interaction items as input features, a regression model is trained to establish a mapping relationship between the input features and the prediction target; Interpretability analysis is used to analyze the trained regression model, quantify the contribution of each input feature to the comfort dimension score, and output optimization suggestions based on the contribution analysis results.

2. The sensory fusion and multi-scenario optimization evaluation method for children's nap desks and chairs as described in claim 1, characterized in that, The evaluation dimensions include: ergonomics, safety, ease of learning, convenience, perception, aesthetics, friendliness, and comfort.

3. The sensory fusion and multi-scenario optimization evaluation method for children's nap desks and chairs as described in claim 1, characterized in that, The preset multiple usage scenarios include at least: a learning scenario for evaluating daily reading and writing functions, an adjustment scenario for evaluating the experience of structural adjustment, and a lunch break scenario for evaluating the effect of lying down for relaxation.

4. The sensory fusion and multi-scenario optimization evaluation method for children's nap desks and chairs as described in claim 1, characterized in that, The steps of the dynamic weighted fusion include: Calculate the Jensen-Shannon divergence between the student rating distribution and the expert rating distribution to quantify the perceived discrepancy between the two across various evaluation dimensions; The perceived divergence is taken as input and converted into a dynamic weight value through a preset nonlinear mapping function; The student scores and expert scores are weighted and summed using the dynamic weight values ​​to obtain the fusion score.

5. The sensory fusion and multi-scenario optimization evaluation method for children's nap desks and chairs as described in claim 4, characterized in that, The adjustment mechanism of the dynamic weight value is as follows: the greater the perceived disagreement, the lower the proportion of the expert score in the weighted summation; the smaller the perceived disagreement, the higher the proportion of the expert score in the weighted summation.

6. The method for sensory fusion and multi-scenario optimization evaluation of children's nap desks and chairs as described in claim 1, characterized in that, Factoring the tensor includes: The third-order rating tensor was decomposed using a parallel factor analysis method to obtain factor loading matrices associated with desks and chairs, evaluation dimensions, and usage scenarios, respectively, and sensitive evaluation factors were identified based on the factor loading matrices.

7. The sensory fusion and multi-scenario optimization evaluation method for children's nap desks and chairs as described in claim 1, characterized in that, When training the regression model, the scores of the dimensional interaction terms in the input features are calculated based on a preset combination of dimensions.

8. The method for sensory fusion and multi-scenario optimization evaluation of children's nap desks and chairs as described in claim 1, characterized in that, The training regression model uses the gradient boosting decision tree algorithm to model and fit the nonlinear relationship between the input features and the prediction target.

9. The sensory fusion and multi-scenario optimization evaluation method for children's nap desks and chairs as described in claim 7, characterized in that, The dimensional interaction items include at least one of the following three: Perceptual-ergonomic interaction terms used to characterize the synergistic effect of material tactile sensation and structure; Safety-convenience interaction term used to characterize the combined effects of structural reliability and ease of operation; Aesthetic-affinity interaction items used to reveal the combined effect of appearance design and emotional connection.

10. The method for sensory fusion and multi-scenario optimization evaluation of children's nap desks and chairs as described in claim 1, characterized in that, The interpretability analysis technique employs the Shapley additive interpretation method, including: Calculate the SHAP value of each input feature, evaluate the positive and negative impacts and marginal contributions of each evaluation dimension and its interaction items on user comfort, and based on the ranking of the SHAP values, identify the most critical design factors of desks and chairs and generate suggestions for improving the structure and function of desks and chairs.

Images