Intelligent toy with child cognitive ability evaluation capability and evaluation method thereof

Abstract

The application relates to an intelligent toy with a child cognitive ability evaluation capability and an evaluation method thereof, and belongs to the technical field of intelligent toys, which comprises a toy body, a function module and a multi-modal sensor array, the toy body is a cuboid base suitable for being gripped by children, and the function module and the multi-modal sensor array are integrated in the toy body. In the application, a standardized cognitive evaluation task is ingeniously embedded in a children's game, rich data are collected in a natural situation by using a highly integrated multi-modal sensor, and the data are converted into scientific evaluation results by a cloud intelligent algorithm, the application solves the problems of a depressing environment, single data and non-sustainable problems in a traditional evaluation mode, realizes evaluation in the form of play, and provides an objective, convenient and professional children's cognitive development tracking and support tool for families and early education institutions.

Description

Technical Field

[0001] This invention relates to the field of intelligent toy technology, and in particular to an intelligent toy with the ability to assess children's cognitive abilities and its assessment method. Background Technology

[0002] Early assessment and intervention of children's cognitive abilities are crucial for their learning and development. Traditional standardized assessment tools (such as the Wechsler Intelligence Scale) must be administered by psychologists in professional settings, which presents problems such as environmental stress, high costs, and difficulty in frequent follow-up.

[0003] To improve the accessibility and objectivity of assessments, existing technologies have proposed various solutions, mainly falling into the following two categories:

[0004] The first category is automated assessment tools based on physical interaction. For example, by using motion sensors built into building blocks, motion data (such as acceleration and completion time) can be automatically collected when children complete building tasks, thereby achieving automated assessment of cognitive and fine motor skills. However, the data dimensions of such solutions are relatively singular, mainly relying on the final operation results. They lack the ability to capture the deep psychological processes of children during the task process, such as cognitive load, emotional state, and attention allocation. The assessment dimensions are not comprehensive enough. At the same time, their functions are mainly focused on assessment and screening, and they fail to form an effective closed loop with daily personalized intervention training.

[0005] The second category is highly specialized digital assessment systems. For example, some systems integrate multimodal biometric devices such as electroencephalography (EEG) and eye trackers, and use machine learning algorithms for analysis to achieve high-precision assessment of attention or cognitive abilities. However, these systems are usually expensive, complex to operate, and require use in a controlled environment. They are difficult to integrate into children's natural life and play scenarios, and the "ecological validity" of the assessment is still limited, making it impossible to apply them on a large scale to daily home monitoring.

[0006] In addition, there are some smart toys on the market with simple interactive or memory training functions, but their original design purpose is mostly entertainment or single ability training. They lack a systematic cognitive assessment theoretical framework and scientific quantitative analysis model as support, and cannot provide professional and reliable assessment results.

[0007] Therefore, the following technical problems still need to be solved in the existing technology: how to create an integrated intelligent assessment solution that can simultaneously collect multi-dimensional behavioral and physiological data in a stress-free natural play environment for children, conduct in-depth quantitative analysis based on cognitive science theory, and finally automatically provide personalized development suggestions. To this end, an intelligent toy with the ability to assess children's cognitive abilities and its assessment method are proposed. Summary of the Invention

[0008] This invention provides an intelligent toy with the ability to assess children's cognitive abilities and its assessment method, in order to solve the problems of traditional assessment methods such as oppressive environment, limited data, and unsustainability, and to achieve a children's cognitive development support program that integrates assessment with entertainment, scientific quantification, and closed-loop feedback.

[0009] The solution of the present invention to the above-mentioned technical problems is as follows:

[0010] In a first aspect, there is an intelligent toy with the ability to assess children's cognitive abilities, comprising a toy body, functional modules and a multimodal sensor array, wherein the toy body is a cubic base suitable for children to grasp, and the functional modules and the multimodal sensor array are integrated into the toy body.

[0011] The functional modules include at least a sequence memory light array module for evaluating working memory and executive functions, an intelligent interlocking panel module for evaluating fine motor skills and spatial planning, and a voice interaction and classification module for evaluating speech comprehension.

[0012] The multimodal sensor array is used to collect operational data, physiological data and image data generated during children's interaction in real time. The multimodal sensor array specifically includes an operational sensor group, a physiological sensor, an image acquisition unit, a data processing and communication module, and a power supply module.

[0013] Based on the above technical solution, the present invention can be further improved as follows.

[0014] Furthermore, the sequence memory light array module is composed of programmable RGB LEDs, the intelligent interlocking panel module is embedded with an RFID reader and configured with interlocking building block components adapted to the RFID reader, and the bottom of each interlocking building block component is embedded with a uniquely identified RFID tag. The voice interaction and classification module integrates a speaker. The programmable RGB LEDs not only provide rich visual cues to attract children's attention, but can also present complex colors and flashing sequences through programming, providing a precise and controllable stimulus source for assessing working memory. The intelligent interlocking panel integrates RFID technology, enabling the toy to automatically, accurately, and non-contactly identify each interlocked building block and its position, realizing objective and non-intrusive data collection of children's interlocking process (such as selection, sequence, and orientation), greatly improving the automation and accuracy of the assessment. The voice module integrates a speaker, enabling the toy to independently complete functions such as task guidance and voice feedback, forming a complete interactive loop without relying on external devices, enhancing the toy's independence and user experience.

[0015] Furthermore, the operational sensor group includes a capacitive touch sensor, an RFID reader, and a six-axis inertial measurement unit (IMU). The capacitive touch sensor is located below the sequence memory light array module, the RFID reader is located in the smart interlocking panel module, and the six-axis IMU is encapsulated inside the interlocking block assembly. The capacitive touch sensor, combined with the light array, can accurately record the reaction time, accuracy, and sequence of a child touching specific light positions. The RFID reader is used to capture spatial planning behavior on the interlocking panel. By embedding the six-axis IMU within the interlocking blocks, it can directly capture fine motor data (such as acceleration, angular velocity, and attitude) during the child's grasping, moving, rotating, and placing of blocks. This provides direct, high-quality time-series data that traditional external sensors cannot obtain for assessing fine motor coordination and operational strategies.

[0016] Furthermore, the physiological sensor is a photoelectric heart rate sensor module integrated inside the toy's handle, which can monitor the child's heart rate signal non-intrusively and continuously during game interaction.

[0017] Furthermore, the image acquisition unit is a miniature camera set on the top frame of the toy body, which can stably capture the child's facial expressions, gaze direction, upper body posture and overall behavior pattern when operating the toy. This visual information is an important supplement to the data from the operation sensor and physiological sensor, and can be used to analyze the child's emotional response, level of concentration, problem-solving strategies (such as hesitation, trial-and-error) and other unstructured behavioral characteristics, enriching the dimensions of the assessment.

[0018] Furthermore, the data processing and communication module has a built-in main control processor that is electrically connected to the multimodal sensor array. It is used to receive, synchronize timestamps, perform fusion preprocessing, and store data streams from each sensor. The main control processor is an embedded processor that acts as an "edge computing" node, responsible for centralized control, data acquisition, and preliminary processing of all sensors. Its key function is to assign a unified and accurate timestamp to all heterogeneous sensor data streams, which is the foundation for subsequent multimodal data fusion and analysis. At the same time, local preprocessing (such as filtering and compression) and temporary storage capabilities reduce the redundancy and noise of the original data and ensure that data is not lost when the network is unstable, thereby improving the overall robustness and data quality of the system and preparing for reliable uploading and cloud analysis.

[0019] Secondly, a method for assessing children's cognitive abilities using a smart toy includes the following steps:

[0020] S1, Task guidance and interaction, presents a structured sequence of game tasks to children through the voice interaction and classification module, the sequence memory light array module and the connected application;

[0021] S2, Multimodal data synchronous acquisition: During the child's task performance, raw data is synchronously and continuously acquired through the multimodal sensor array;

[0022] S3, data upload and cloud processing, uploads the collected, locally encrypted and packaged multimodal time series data packets to the cloud server via wireless network;

[0023] S4, Algorithm Analysis and Evaluation: The cloud server calls a pre-trained evaluation algorithm model to analyze the uploaded data;

[0024] S5, Report Generation and Feedback: Based on the analysis results of S4, it automatically generates assessment reports and provides feedback to parents or educators through the application.

[0025] Based on the above technical solution, the present invention can be further improved as follows.

[0026] Furthermore, the analysis process in step S4 includes:

[0027] a. Extract key feature indicators from the operational data, which comes from the operational sensor group;

[0028] b. Extract indicators reflecting cognitive load from physiological data, which are derived from physiological sensors;

[0029] c. Extract behavioral pattern features from image data, wherein the image data comes from the image acquisition unit;

[0030] d. Input the extracted features into the assessment model and output the quantitative assessment results of children on multiple cognitive dimensions;

[0031] First, specialized feature extraction is performed based on the characteristics of different modalities of data: efficiency indicators reflecting accuracy, speed, and strategy are extracted from operational data; load indicators related to psychological effort are analyzed from physiological data; and emotional and behavioral pattern indicators are identified from image data. Finally, these three heterogeneous but complementary feature vectors are fused and input into the assessment model. This multimodal fusion analysis method overcomes the limitations of a single data source and can cross-validate children's cognitive performance from multiple perspectives, including external behavior, internal state, and nonverbal behavior, making the final quantitative assessment results more comprehensive, objective, and reliable.

[0032] Furthermore, the evaluation algorithm model adopts a supervised learning-trained machine learning model, whose inputs are temporal features, statistical features and image features extracted from the multimodal data, and whose output layer corresponds to the estimation of multiple latent variables of cognitive abilities.

[0033] A supervised learning machine learning model is employed, trained on a large amount of labeled paired data of "children's play data - cognitive ability standard assessment results". This enables the model to learn the mapping relationship between complex, high-dimensional multimodal features and abstract cognitive abilities. The model can simultaneously process temporal features (such as response sequences), statistical features (such as average reaction time), and image features (such as facial expression changes), demonstrating its powerful integration and modeling capabilities for multi-source heterogeneous information. The output layer directly corresponds to the "latent variable estimation" of multiple cognitive dimensions, indicating that the model can complete the quantitative scoring of multiple cognitive concepts such as working memory, executive function, and fine motor skills in one go. The assessment efficiency is high, and the results have interpretable psychological significance.

[0034] Furthermore, for the assessment of working memory, the assessment algorithm model adopts a three-parameter Logistic item response theory model, which estimates the latent trait score based on the child's response pattern on the sequential memory lamp array module task.

[0035] The three-parameter Logistic Item Response Theory (3PL-IRT) model is one of the gold standards in the fields of psychological and educational measurement. When applied to data analysis of sequential memory lamp array tasks, it can consider not only whether children answer correctly or incorrectly (parameter 1), but also the difficulty of the questions (parameter 2), the discrimination (parameter 3), and the probability of children guessing (parameter 4). This allows for a more accurate and fair estimate of children's potential trait of "working memory ability." The ability scores estimated by this model have equidistant scale characteristics, which facilitates comparisons across tasks, time periods, and even children, providing a solid quantitative foundation for personalized assessment and development tracking.

[0036] The beneficial effects of this invention are as follows: This invention provides an intelligent toy with the ability to assess children's cognitive abilities and its assessment method, which has the following advantages:

[0037] 1. High ecological validity: The assessment is conducted in a game context that children are familiar with and enjoy, which effectively avoids test anxiety and the results can better reflect their cognitive ability level in a real and natural state.

[0038] 2. The assessment dimensions are comprehensive, breaking through the limitations of a single operation result. By combining the timing of physical operation behavior, implicit physiological signals and visual behavioral information, the cognitive process is depicted in a three-dimensional and multi-faceted way, and the assessment results are more objective and in-depth.

[0039] 3. Scientific and professional: The game is designed based on the classic cognitive psychology task paradigm, and a pre-trained machine learning model is used to transform the raw sensor data into quantitative indicators that meet psychometric standards, ensuring the scientific nature and credibility of the evaluation.

[0040] 4. User-friendly experience: for children, the process is like playing a game, which motivates them to participate; for parents, the operation is simple, the reports are intuitive and easy to understand, and they can obtain personalized educational guidance, forming a virtuous cycle of assessment-feedback-intervention.

[0041] 5. Sustainable tracking facilitates regular and frequent use in home or kindergarten environments, forming an individual child's cognitive development curve, which is conducive to long-term monitoring of developmental changes and intervention effects, and empowers continuous educational support;

[0042] 6. Through innovative integrated design of hardware task carrier, multi-cognitive dimension assessment tasks are solidified into a single entity and combined with a specific multimodal sensor array configuration scheme, realizing the seamless and synchronous collection of high ecological validity data in natural game flow.

[0043] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it according to the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Specific embodiments of the present invention are given in detail below with reference to the accompanying drawings. Attached Figure Description

[0044] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0045] Figure 1 A schematic diagram of the structure of an intelligent toy with the ability to assess children's cognitive abilities and its assessment method provided in an embodiment of the present invention;

[0046] Figure 2 A top-view structural diagram of an intelligent toy with the ability to assess children's cognitive abilities and its assessment method, provided in an embodiment of the present invention;

[0047] Figure 3 A front view of an intelligent toy with cognitive ability assessment capabilities for children and its assessment method, provided in an embodiment of the present invention;

[0048] Figure 4 A top view of an intelligent toy with the ability to assess children's cognitive abilities and its assessment method, provided in an embodiment of the present invention;

[0049] Figure 5 This is a schematic diagram of the interlocking building block component in an intelligent toy and its assessment method for assessing children's cognitive abilities, provided in an embodiment of the present invention.

[0050] Figure 6This is a flowchart illustrating a smart toy with cognitive ability assessment capabilities for children and its assessment method, as provided in an embodiment of the present invention.

[0051] The attached diagram lists the components represented by each number as follows:

[0052] 1. Toy body; 2. Functional modules; 3. Multimodal sensor array; 201. Sequence memory light array module; 202. Intelligent interlocking panel module; 2021. Interlocking building block components; 2022. RFID tag; 203. Voice interaction and classification module; 301. Operation sensor group; 3011. Capacitive touch sensor; 3012. RFID reader; 3013. Six-axis inertial measurement unit;

[0053] 302. Physiological sensor; 303. Image acquisition unit; 304. Data processing and communication module; 3041. Main control processor; 305. Power supply module. Detailed Implementation

[0054] The following is in conjunction with the appendix Figure 1-6 The principles and features of the present invention are described below. The examples given are for illustrative purposes only and are not intended to limit the scope of the invention. The invention is described more specifically in the following paragraphs by way of example with reference to the accompanying drawings. The advantages and features of the invention will become clearer from the following description. It should be noted that the drawings are in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the invention.

[0055] It should be noted that when a component is described as "fixed to" another component, it can be directly on the other component or may have a component in between. When a component is considered "connected to" another component, it can be directly connected to the other component or may have a component in between. When a component is considered "set on" another component, it can be directly set on the other component or may have a component in between. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0057] like Figure 1-5As shown, the present invention provides an intelligent toy with the ability to assess children's cognitive abilities. This toy aims to achieve a scientific assessment of children's cognitive abilities by collecting multi-dimensional data during children's natural play interactions.

[0058] The smart toy mainly includes the toy body 1, the functional modules integrated on it 2, and the multimodal sensor array 3;

[0059] The toy body 1 is designed as a cube base suitable for children to grasp. The structure is stable and easy for children to operate on a table. All functional modules and sensors are compactly integrated into the cube structure to form an integrated device.

[0060] The functional module 2 includes at least three core parts: a sequence memory light array module 201, an intelligent splicing panel module 202, and a voice interaction and classification module 203;

[0061] The sequence memory light array module 201 consists of a set of programmable RGB LEDs, used to present visual memory tasks, such as a sequence of lights flashing in a specific order and color, requiring children to observe and reproduce it in order to assess their working memory and executive function.

[0062] The intelligent interlocking panel module 202 is set on the toy body 1, and an RFID reader 3012 is embedded on it. The module is equipped with several independent interlocking building block components 2021. Each interlocking building block component 2021 has an RFID tag 2022 with a unique identifier embedded at its bottom. Children can interlock the building blocks in different positions on the panel. The RFID reader 3012 can identify which building block is placed where without contact, thereby assessing the child's fine motor and spatial planning abilities.

[0063] Voice interaction and classification module 203: It integrates a microphone and speaker, and can realize voice dialogue with children, issue commands, play sound effects and other functions to assess children's speech comprehension and expression abilities;

[0064] The multimodal sensor array 3 is the core of data acquisition, used to collect various types of data generated when children interact with toys in real time and synchronously;

[0065] Specifically, it includes:

[0066] Operation sensor group 301: for capturing specific operational behaviors of children, further comprising:

[0067] Capacitive touch sensor 3011: Located below each LED bead in the sequence memory lamp array module 201, used to accurately detect the position (X, Y coordinates), time and sequence of a child touching the corresponding lamp position;

[0068] RFID reader 3012: As mentioned above, it is integrated inside the smart interlocking panel module 202. When a block is placed or moved, it triggers a reading event and records the "time-block ID-panel position" triplet.

[0069] Six-axis inertial measurement unit 3013: such as Figure 5 As shown, the unit is encapsulated inside each interlocking block assembly 2021 to measure acceleration and angular velocity data in real time during the process of picking up, moving, rotating and placing the blocks, thereby capturing fine motion details;

[0070] Physiological sensor 302: In this embodiment, it is a photoelectric heart rate sensor module, which is integrated on the inside of the handle of the toy body 1. When the child grasps the toy, it is in close contact with the child's palm skin and collects photoplethysmography pulse wave signals at a sampling rate of 128Hz. It is used to calculate heart rate and heart rate variability, and indirectly reflects cognitive load and emotional state.

[0071] Image acquisition unit 303: In this embodiment, it is a miniature camera, which is set at the top frame of the toy body 1. Its field of view can cover the child's face and upper body when operating the toy, and is used to capture behavioral patterns such as facial expressions, gaze, and gestures.

[0072] Data processing and communication module 304: This module has a built-in main control processor 3041 (such as an ARM architecture embedded processor) and is electrically connected to the operation sensor group 301, physiological sensor 302, and image acquisition unit 303 through circuits. Its main functions include: stamping all sensor data streams with a unified high-precision timestamp; performing preliminary filtering, compression, and other fusion preprocessing on the raw data; temporarily storing the processed data; and uploading the data in packages through its integrated wireless communication unit (such as Wi-Fi / Bluetooth). Power module 305 supplies power to the entire system.

[0073] The physiological sensor 302 is a photoelectric heart rate sensor module integrated inside the handle of the toy body 1. This module uses a dedicated chip (such as MAX86141) with an anti-motion artifact algorithm and is linked with the six-axis inertial measurement unit 3013 inside the building blocks. The main control processor 3041 performs dynamic motion compensation on the pulse wave signal through the data of the inertial measurement unit to improve the signal quality in the child's mild activity state. The extracted physiological indicators mainly serve as auxiliary and trend references to reflect the relative cognitive load changes during the task, and together with the main indicators of operational behavior, they constitute a multi-dimensional evaluation system.

[0074] like Figure 6 As shown, the specific working principle and usage method of this invention are as follows:

[0075] S1, Task guidance and interaction: Parents select the assessment mode through a dedicated application on their mobile devices, and the toy uses the speaker of the classification module 203 to issue game invitations and instructions to the child in a friendly voice.

[0076] For example, the sequence memory light array module 201 starts flashing a sequence of lights, and the voice prompts the child to "please click on the lights in the order they were lit," or the application displays a pattern on the screen and the voice guides the child to use the blocks of the smart interlocking panel module 202 to assemble the same pattern on the toy.

[0077] S2, Multimodal data synchronous acquisition: During the entire process of the child performing the above game tasks, the multimodal sensor array 3 starts to work, and the capacitive touch sensor 3011 records the accuracy and response time of each touch.

[0078] The six-axis inertial measurement unit 3013 inside the interlocking building blocks records the stability of the motion trajectory during the grasping and placement process;

[0079] The RFID reader 3012 under the panel records the correctness and spatial relationship of the final splicing result; the physiological sensor 302 inside the handle continuously monitors the heart rate signal;

[0080] The top image acquisition unit 303 records the child's facial expressions and body language, and all of this data is synchronized and timestamped by the main control processor 3041;

[0081] S3, Data Upload and Cloud Processing: The data processing and communication module 304 locally encrypts the multimodal data (including operation timing, physiological waveforms, video frames, etc.) collected for a complete task cycle, packages it into a data packet, and then transmits it to the cloud server via a wireless network.

[0082] S4, Algorithm Analysis and Evaluation: After receiving the data packet, the cloud server calls the pre-trained evaluation algorithm model for analysis. The analysis process is the core and includes:

[0083] a. Feature extraction: Extract key feature indicators such as task completion time, error rate, touch sequence consistency, and motion trajectory smoothness from operational data; extract indicators reflecting cognitive load such as heart rate rise slope and power in specific frequency bands from physiological data; and extract behavioral pattern features such as fixation point change frequency, smiling frequency, and number of hesitation pauses from image data using computer vision algorithms.

[0084] b. Model evaluation: The extracted multi-dimensional feature vectors are input into the evaluation model, which is a machine learning model (such as gradient boosting decision tree or neural network) trained using supervised learning (based on a large amount of existing children's cognitive assessment standard data). Its output layer directly corresponds to the latent variable estimation scores of multiple cognitive dimensions such as "working memory", "executive function", "fine motor skills" and "verbal comprehension". In particular, for working memory tasks, the model can use the classic three-parameter Logistic Item Response Theory (3PL-IRT) model to more accurately estimate the potential working memory ability score of children based on their response patterns on sequential memory tasks (such as the accuracy and reaction time of sequences of different difficulty).

[0085] The mathematical form of the three-parameter Logistic Response Theory model is:

[0086]

[0087] in, The ability is represented as The probability that a child answers a sequence of a specific difficulty correctly; It is a discrimination parameter. It's a difficulty parameter. It's a guess of the parameters. This is a scaling constant. The system uses maximum likelihood estimation to infer the potential working memory trait score based on the child's response patterns (correct / incorrect) on light sequence tasks of varying difficulty. This fraction has the property of an equidistant measuring scale, which facilitates accurate comparison and tracking;

[0088] S5, Report Generation and Feedback: Based on the quantitative scores of each dimension calculated in step S4, the cloud server automatically generates an assessment report with pictures and text. The report not only includes scores, but may also include percentiles compared with norms, analysis of strengths and weaknesses in the development of abilities in each dimension, and personalized game or training suggestions based on the current assessment results. The report is pushed to parents or educators through the application, such as recommending training game tasks suitable for the child in the next stage (e.g., "It is recommended to strengthen working memory, and try a longer sequence next time"), thereby starting a new round of "assessment-training" cycle and completing the closed loop from assessment to feedback.

[0089] The following is a specific implementation example of the evaluation algorithm model;

[0090] 1. Data preprocessing and feature extraction: After receiving the encrypted data packet, the cloud server first decrypts and parses it to obtain a timestamp-aligned multimodal raw data stream;

[0091] The data processing for the sequence memory light array task is as follows;

[0092] Raw data: a sequence of touch events from the capacitive touch sensor 3011, each event containing (timestamp, X coordinate, Y coordinate, touch state);

[0093] Perform feature extraction;

[0094] Response accuracy: Match the child's touch sequence with the target light sequence and calculate the complete sequence accuracy rate and the partial accuracy rate (correct position but incorrect sequence).

[0095] Reaction time: Extract the reaction time for each correct response (from stimulus presentation to touch), and calculate the mean reaction time and the standard deviation of reaction time;

[0096] Error patterns: Counts the number of missed errors, duplicate errors, and intrusion errors (touching non-target lights);

[0097] Operation path: Calculate the total length of the operation path and the unnecessary movement distance based on the touch coordinate sequence;

[0098] For the IMU data of the building blocks (as auxiliary verification);

[0099] Raw data: Time-series data of triaxial acceleration and triaxial angular velocity from a six-axis inertial measurement unit 3013;

[0100] Feature extraction: During the waiting period of performing a memory task, the IMU signal of grasping building blocks was analyzed, and the amplitude of hand tremor frequency was extracted as a proxy indicator of attentional tension.

[0101] For physiological data;

[0102] Raw data: Photoplethysmography (PPG) signal from physiological sensor 302;

[0103] Feature extraction: The average increase in heart rate (ΔHR) and the percentage decrease in low-frequency band (LF) power of heart rate variability (HRV) relative to baseline during the task period were calculated as quantitative indicators of cognitive load.

[0104] For image data;

[0105] Raw data: Video stream from image acquisition unit 303;

[0106] Feature extraction: Using open-source facial behavior analysis toolkits (such as OpenFace), extract the total duration of staring at the light array region, the number of times the gaze left the task, and the intensity changes of facial expressions of confusion / focus.

[0107] 2. Feature Fusion and Model Evaluation

[0108] The extracted 15-20 dimensional feature vectors are standardized and then input into the pre-trained evaluation model.

[0109] Model architecture: This example uses a hybrid model of a two-layer fully connected neural network and an item response theory (IRT) module;

[0110] The first layer: The neural network input layer receives multimodal feature vectors, performs nonlinear fusion and dimensionality reduction through the hidden layer, and outputs a comprehensive "task performance latent variable";

[0111] The second layer: The "task performance latent variable" is used as the input to the three-parameter Logistic (3PL-IRT) model. Here, the parameters of the IRT model (difficulty b, discrimination a) are pre-calibrated for each specific light sequence task (i.e. "question"), and these parameters are obtained by calibration on a large number of children's samples.

[0112] Model output: The IRT model calculates the latent trait score θ of the child on the "working memory" dimension based on the child's "task performance latent variable" and the parameters of the current task. This score is a continuous value that falls within the range of a standard norm distribution (e.g., mean 100, standard deviation 15).

[0113] Model Training: The hybrid model is trained using a historical dataset for supervised learning. The dataset contains multimodal features (X) collected from thousands of children as they complete the toy task of this invention, as well as their working memory index scores (Y, as the gold standard label) obtained from the standard Wechsler Intelligence Scale for Children (WISC) test during the same period. The training objective is to maximize the correlation (such as Pearson correlation coefficient) between the model's predicted θ score and the standard score Y.

[0114] The intelligent toy and assessment method provided in this embodiment cleverly embed standardized cognitive assessment tasks into children's games. It uses highly integrated multimodal sensors to collect rich data in natural settings and transforms the data into scientific assessment results through cloud-based intelligent algorithms. It solves the problems of traditional assessment methods such as oppressive environment, limited data, and unsustainability, and achieves "assessment through play". It provides families and early education institutions with an objective, convenient, and professional tool for tracking and supporting children's cognitive development.

[0115] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Content not described in detail in this specification is prior art known to those skilled in the art.

[0116] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Those skilled in the art can readily implement the present invention based on the accompanying drawings and the above description. However, any modifications, alterations, or variations made by those skilled in the art without departing from the scope of the present invention, utilizing the disclosed technical content, are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, or variations made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the present invention.

Claims

1. A smart toy with the ability to assess children's cognitive abilities, comprising a toy body (1), functional modules (2), and a multimodal sensor array (3), characterized in that: The toy body (1) is a cube base suitable for children to grasp, and the functional module (2) and the multimodal sensor array (3) are integrated into the toy body (1). The functional module (2) includes at least a sequence memory lamp array module (201) for evaluating working memory and executive functions, an intelligent interlocking panel module (202) for evaluating fine motor skills and spatial planning, and a voice interaction and classification module (203) for evaluating speech comprehension. The multimodal sensor array (3) is used to collect operational data, physiological data and image data generated during children's interaction in real time. The multimodal sensor array (3) specifically includes an operational sensor group (301), a physiological sensor (302), an image acquisition unit (303), a data processing and communication module (304), and a power supply module (305).

2. The intelligent toy with cognitive ability assessment capabilities for children according to claim 1, characterized in that, The sequence memory light array module (201) is composed of programmable RGB LEDs. The intelligent interlocking panel module (202) is embedded with an RFID reader (3012) and is equipped with interlocking building block components (2021) adapted to the RFID reader (3012). The bottom of the interlocking building block components (2021) is embedded with a uniquely identified RFID tag (2022). The voice interaction and classification module (203) integrates a speaker.

3. The intelligent toy with cognitive ability assessment capabilities for children according to claim 1, characterized in that, The operation sensor group (301) includes a capacitive touch sensor (3011), an RFID reader (3012), and a six-axis inertial measurement unit (3013); the capacitive touch sensor (3011) is located below the sequence memory lamp array module (201), the RFID reader (3012) is located in the intelligent interlocking panel module (202), and the six-axis inertial measurement unit (3013) is encapsulated inside the interlocking building block assembly (2021).

4. The intelligent toy with cognitive ability assessment capabilities for children according to claim 1, characterized in that, The physiological sensor (302) is an optoelectronic heart rate sensor module integrated inside the handle of the toy body (1).

5. The intelligent toy with cognitive ability assessment capabilities for children according to claim 1, characterized in that, The image acquisition unit (303) is a miniature camera installed on the top frame of the toy body (1).

6. The intelligent toy with cognitive ability assessment capability for children according to claim 1, characterized in that, The data processing and communication module (304) has a built-in main control processor (3041) that is electrically connected to the multimodal sensor array (3) and is used to receive, synchronize timestamps, perform fusion preprocessing and store data streams from each sensor; the main control processor (3041) is an embedded processor.

7. A method for assessing children's cognitive abilities, characterized in that, The invention employs the intelligent toy as described in any one of claims 1-6, and includes the following steps: S1, Task guidance and interaction, through the voice interaction and classification module (203), the sequence memory light array module (201) and the connected application, presents a structured game task sequence to children; S2, Multimodal data synchronous acquisition: During the child's task execution, raw data is synchronously and continuously acquired through the multimodal sensor array (3); S3, data upload and cloud processing, uploads the collected, locally encrypted and packaged multimodal time series data packets to the cloud server via wireless network; S4, Algorithm Analysis and Evaluation: The cloud server calls a pre-trained evaluation algorithm model to analyze the uploaded data; S5, Report Generation and Feedback: Based on the analysis results of S4, it automatically generates assessment reports and provides feedback to parents or educators through the application.

8. The method for assessing children's cognitive abilities according to claim 7, characterized in that, The analysis process in step S4 includes: a. Extract key feature indicators from the operation data, which comes from the operation sensor group (301). b. Extract indicators reflecting cognitive load from physiological data, which are derived from physiological sensors (302). c. Extract behavioral pattern features from image data, wherein the image data comes from image acquisition unit (303). d. Input the extracted features into the evaluation model and output the quantitative evaluation results of children in multiple cognitive dimensions.

9. The method for assessing children's cognitive abilities according to claim 8, characterized in that, The evaluation algorithm model adopts a supervised learning training machine learning model. Its input consists of temporal features, statistical features, and image features extracted from the multimodal data, and its output layer corresponds to the estimation of multiple latent variables of cognitive abilities.

10. The method for assessing children's cognitive abilities according to claim 9, characterized in that, For the assessment of working memory, the assessment algorithm model adopts a three-parameter Logistic item response theory model, which estimates the latent trait score based on the child's response pattern on the sequential memory lamp array module (201) task.

Images