Device and method for evaluating children's higher-order cognition and emotions based on painting pixel temporal information

Abstract

The present invention discloses a device for assessing children's higher-level cognition and emotions based on the timing information of painting pixels. The device comprises: a drawing board with a built-in electromagnetic module, which includes a main control chip and an electromagnetic induction coil for real-time acquisition of pixel timing data from an electromagnetic pen; an electromagnetic pen including a pressure sensor, an electromagnetic transmitting coil, a pen tip, and a signal amplifier for transmitting position, pressure, and timestamp signals to the electromagnetic module; a data processing module connected to the electromagnetic module for parsing the pixel timing data and calculating at least 20 painting behavior features; and a server for receiving the painting behavior features and performing feature selection, fusion, and modeling on the painting behavior features through machine learning to generate an assessment report that scores the child's originality, flexibility, fluency, and sophistication of thinking. The present invention achieves automated and comprehensive recording of a child's painting process and, based on the process data of the stylus pixel timing, enables intelligent assessment of the child's higher-level cognition and emotional state.

Description

Technical Field

[0001] The present invention relates to the technical field of automated assessment of children's (3-18 years old) cognitive ability and emotional development, and in particular to a device and method for assessing children's high-level cognition and emotions based on temporal information of painting pixels. Background Art

[0002] Children's growth and education are attracting increasing attention. Traditional methods for assessing children's cognitive abilities are limited by static testing environments and limited interaction, resulting in high labor and time costs during the assessment process, and requiring a high degree of professionalism and complexity. Therefore, it is difficult to comprehensively and accurately assess children's cognitive abilities and emotional development. Based on this, existing technologies attempt to provide more interactivity through digital means such as learning machines. For example, Patent Publication No.: CN115394395A discloses a method, system and device for neuropsychological assessment and intervention based on painting creation, which can dynamically, accurately and effectively capture and analyze the physiological changes of patients during the diagnosis, treatment and rehabilitation of mental and psychological diseases in real time, and complete disease assessment and treatment intervention through neuropsychological painting creation, realizing the integrated analysis and comprehensive evaluation of painting creation, neurophysiology and mental psychology, increasing the fun and objectivity of diagnosis and treatment, thereby improving patients' participation in diagnosis and treatment and treatment compliance, improving diagnostic accuracy and treatment effectiveness, and assisting clinical diagnosis and treatment practice.

[0003] For another example, patent publication number: CN114550918A discloses a psychological disorder assessment method and system based on painting feature data. It predicts psychological disorders by combining painting image features with psychological disorder diagnosis results, and uses machine learning methods to perform regression analysis on painting features to predict whether the user has a psychological disorder.

[0004] For another example, patent publication number CN108230427A discloses an intelligent drawing device, an image analysis system, and an image processing method. A touch operation is performed on a display screen using a stylus pen to form an image. A processing module in the intelligent picture frame records the stylus operation process and the image formed. A communication module in the intelligent picture frame sends the operation process and the image formed to a server, where the server performs a user psychological analysis based on the operation process and the image formed. Children's drawings can typically describe their psychological state. Children can draw using an intelligent drawing device, which records the child's drawing process and the image drawn, and sends it to a server. The child's psychological state is evaluated and analyzed based on the drawing process and the image drawn. Parents can use the psychological analysis results to promptly understand their child's psychological development and adjust the parent-child relationship in a timely manner.

[0005] However, the above existing solutions cannot effectively automate the assessment of children's higher-order cognitive abilities and emotional states. The main reasons are:

[0006] 1. Insufficient utilization of temporal behavioral data: Existing technologies rely on static image features (such as color and lines) or single pressure data, and fail to extract dynamic temporal features during the painting process (such as stroke intervals, speed variability, and thinking time distribution). This results in an inability to capture the dynamic changes in children's cognitive abilities, limiting the ability to perform refined modeling.

[0007] 2. Lack of assessment of higher-level cognitive abilities: Existing solutions primarily focus on detecting psychological disorders or basic emotion analysis and do not adequately cover quantitative indicators of higher-level cognitive abilities such as creativity, mental flexibility, and originality. However, characteristics of creative thinking, such as fluency, flexibility, and originality, have been studied and can be used to define these abilities. Furthermore, existing solutions have yet to establish machine learning models that directly link behavioral indicators to cognitive abilities.

[0008] 3. Data Dimension and Algorithm Limitations: Existing technologies only process pressure, coordinates, and other data at the mean calculation level, without exploring variability and distribution characteristics (such as speed standard deviation and quadrant proportion differences), resulting in a single evaluation dimension. Furthermore, machine learning models only correlate painting results with psychological states, and fail to achieve dynamic modeling of cognitive abilities through temporal behavioral characteristics (such as brushstroke distance variability and starting angle).

[0009] 4. Manual reliance and lack of timeliness: Traditional methods rely on manual scoring by experts, which is time-consuming and highly subjective. In addition, existing solutions lack real-time data processing capabilities and cannot generate evaluation reports simultaneously during the painting process.

[0010] Therefore, it is necessary to propose improvement schemes to achieve intelligent assessment of children's higher-order cognitive and emotional states. Summary of the Invention

[0011] The purpose of the present invention is to provide a device and method for evaluating children's higher-order cognition and emotions based on the temporal information of painting pixels.

[0012] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0013] A device for assessing children's higher-order cognition and emotions based on temporal information of painting pixels, including:

[0014] A drawing board with a built-in electromagnetic module, which includes a main control chip and an electromagnetic induction coil, and is used to collect pixel timing data of the electromagnetic pen in real time;

[0015] The electromagnetic pen includes a pressure sensor, an electromagnetic transmitting coil, a pen tip and a signal amplifier, which are used to transmit position, pressure and time stamp signals to the electromagnetic module;

[0016] a data processing module, connected to the electromagnetic module, configured to analyze the pixel timing data and calculate at least 20 painting behavior features;

[0017] The server is used to receive the drawing behavior characteristics, and perform feature selection, fusion and modeling on the drawing behavior characteristics through machine learning to generate an evaluation report that scores the originality, flexibility, fluency and sophistication of children's thinking.

[0018] Furthermore, the pixel timing data is collected at a frequency of 240 times per second, and the collected data includes two-dimensional coordinates, pen-down status, pressure value and timestamp; the pen-down status uses 0 or 1 to indicate whether the pen is down.

[0019] Furthermore, the electromagnetic induction coils are evenly arranged horizontally and vertically, and are used to divide the drawing board into multiple sensing areas to capture the precise position of the electromagnetic pen.

[0020] The present invention also provides a method for evaluating children's higher-order cognition and emotions based on painting pixel temporal information, comprising the following steps:

[0021] (1) Collect pixel temporal data on the child’s drawing board, including position, pressure, timestamp, and pen-down status;

[0022] (2) parsing the data, restoring the painting pattern, and calculating at least 20 painting behavior features;

[0023] (3) Using a machine learning model to select, fuse, and model the characteristics of drawing behavior, and then generate automated evaluation results of children's originality, flexibility, fluency, and sophistication based on the selected drawing behavior characteristics; the calculation formulas for originality, flexibility, fluency, and sophistication are as follows:

[0024] Originality = -0.15*total handwriting distance + 0.14*total drawing time + 0.18*speed variability + 0.21*handwriting distribution breadth -0.11*quadrant proportion difference + 0.09*pen pressure variability;

[0025] Flexibility = 0.18*time to start writing - 0.07*total number of strokes - 0.1*quadrant proportion difference + 0.14*thinking time variability + 0.11*distance from the starting point to the origin;

[0026] Fluency = 0.28*total number of strokes + 0.19*breadth of handwriting distribution - 0.18*thinking time variability + 0.29*average stroke distance;

[0027] Fineness = 0.17*total number of strokes + 0.22*stroke density + 0.19*average stroke distance + 0.15*speed variability;

[0028] (4) Output an evaluation report containing the scores.

[0029] Furthermore, the machine learning model screens painting behavior features through statistical analysis and optimizes weights based on 10-fold cross validation.

[0030] Specifically, the specific process of the machine learning model to screen painting behavior characteristics is: through statistical analysis, screen the behavior characteristics that are significantly correlated with each dimension, and the screening criteria are: the absolute value threshold of the Pearson correlation coefficient r>0.3 or the significance level p<0.05; then use the screened behavior characteristics as input variables to train the multiple linear regression model, and optimize the parameters of the multiple linear regression model through cross-validation.

[0031] Furthermore, during the training of the multiple linear regression model, the significance of each behavioral feature is confirmed by t-test and F-test to determine whether to include it in the final multiple linear regression model.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] (1) The present invention uses an electromagnetic module to collect time series data such as two-dimensional coordinates, pressure values, and pen-falling status at a high frequency of 240Hz, and combines at least 20 dynamic behavioral features to achieve full-dimensional dynamic modeling of the painting process. Compared with the existing technology that only records static images and simple pressure data or relies on static image features (such as color, lines) or combines physiological signals but does not analyze the painting behavior itself, the present invention combines high-frequency time series data with multi-dimensional behavioral features, filling the gap in the existing technology for dynamic cognitive process analysis. For example, by calculating indicators such as speed variability and thinking time variability, the cognitive fluctuation characteristics of children in the painting process are quantified. Compared with the static features extracted by the existing technology, the at least 20 behavioral indicators extracted by the present invention construct an evaluation system from multiple dimensions such as time, space, and pressure, effectively overcoming the limitations of a single data dimension and insufficient use of time series behavioral data, making the evaluation results more comprehensive and close to the real cognitive process.

[0034] (2) The present invention inputs at least 20 behavioral features into a pre-trained machine learning model, and then screens the behavioral features. In combination with a self-designed calculation method, the behavioral features are directly associated with higher-order cognitive abilities such as originality and flexibility, generating automated scores for thinking fluency, flexibility, originality, and emotional state, and ultimately obtaining an evaluation report. For example, the model predicts thinking flexibility through the breadth of handwriting distribution (reflecting spatial exploration ability) and quadrant proportion differences (reflecting attention allocation), and associates the degree of emotional arousal through speed variability and pen pressure variability. Compared with the existing technology that relies on expert manual scoring or simple image analysis, the present invention establishes a direct mathematical model of dynamic behavioral features and higher-order cognitive abilities, breaking through the subjectivity of traditional manual scoring, solving the problems of "lack of higher-order cognitive assessment" and "human dependence", and improving the evaluation efficiency by more than 80%. The consistency with expert scoring (r = 0.37-0.52) is significantly higher than that of the existing solution.

[0035] (3) Experiments have shown that the present invention performs indicator calculations and model predictions simultaneously after receiving data, and can generate an evaluation report within 5 seconds after the child completes the drawing. For example, when the variability of total thinking time is detected to be abnormally high, the system automatically marks "attention distraction risk"; when the variability of brushstroke distance is lower than the threshold, it prompts "thinking stereotype tendency". In this way, compared with the existing technology that relies on offline physiological signal analysis and delayed image processing, the present invention effectively overcomes the problem of "lack of timeliness", provides an immediate basis for intervention, and is particularly suitable for continuous dynamic monitoring in educational or clinical scenarios.

[0036] (4) The present invention confirms originality, flexibility, fluency and precision through indicator calculation, making the evaluation results highly interpretable. For example, speed variability positively predicts originality of thinking (weight 0.18), indicating that the greater the fluctuation in painting speed, the stronger the creativity; the difference in quadrant proportion negatively predicts flexibility (weight -0.1), reflecting that excessive concentration of attention may lead to rigid thinking. Compared with the existing technology that relies on fuzzy physiological signal associations and black box machine learning models, the technical solution of the present invention converts subjective cognitive ability into objective quantitative indicators through a linear mixed model (feature selection based on Pearson correlation coefficient and significance test), thereby significantly reducing the subjective bias caused by manual scoring, and the ICC consistency index is greatly improved from 0.7 to above 0.9, and supports the evaluation party to make improvement suggestions for specific indicators.

[0037] (5) The various links of the present invention are closely linked, complementary and related. By capturing the pixel data of children's paintings, it realizes the automatic and comprehensive recording of the children's painting process, and based on the process data of the stylus pixel time sequence, it realizes the intelligent assessment of children's high-level cognitive and emotional state. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 Schematic diagram of the device structure of an embodiment of the present invention.

[0039] Figure 2 Schematic diagram of the structure of the electromagnetic module in an embodiment of the present invention.

[0040] Figure 3 Schematic diagram of the structure of the electromagnetic pen in an embodiment of the present invention.

[0041] Figure 4 Result of correlation analysis between the automated creativity evaluation score and the manually evaluated creativity score in the embodiment of the present invention.

[0042] The component names corresponding to the reference numerals are as follows:

[0043] 1-Drawing board, 2-Electromagnetic module, 3-Function key, 4-Page turning key, 5-On / Off button, 6-OLED display, 7-Power management module, 8-Submit button, 9-Polymer shell, 10-Electromagnetic pen, 11-Power supply module, 12-Pressure sensor, 13-Electromagnetic transmitting coil, 14-Pen tip, 15-Main control chip, 16-Electromagnetic induction coil, 17-Signal amplifier. DETAILED DESCRIPTION

[0044] The present invention will be further described below with reference to the accompanying drawings and examples. The embodiments of the present invention include but are not limited to the following examples.

[0045] Example

[0046] This embodiment provides a device for evaluating children's higher-level cognition and emotions based on the time series information of painting pixels. It uses high-precision electromagnetic sensing technology to capture the dynamic time series data of children's painting process in real time, and combines multi-dimensional behavioral indicator analysis with machine learning models to automatically evaluate their cognitive ability and emotional state. Figures 1 to 3 The solution of this embodiment is introduced.

[0047] This embodiment mainly includes a drawing board 1, an electromagnetic module 2, an electromagnetic pen 10, a data processing module and a server. The connection relationship and functions of each component are as follows:

[0048] (1) Drawing board 1: It is wrapped by a polymer shell 9, with an electromagnetic induction area on the surface. It integrates an electromagnetic module 2, a power management module 7, a Bluetooth transmission module, an OLED display 6, and a submit button 8. Its function is to interact with the electromagnetic pen 10 through the electromagnetic module 2 to collect drawing time series data.

[0049] The OLED display screen 6 is used to display the device connection status (such as the Bluetooth pairing success prompt) and the corresponding page number (the page number is turned by the page turning key 4); the submit button 8 is used to submit after the drawing is completed; the power management module 7 supplies power to each component and supports continuous operation for more than 8 hours.

[0050] (2) Electromagnetic module 2 (see Figure 2 ): includes a main control chip 15 and an electromagnetic induction coil 16, which are used to collect pixel timing data of the electromagnetic pen 10 in real time. The main control chip 15 is responsible for receiving the signal of the electromagnetic pen 10, and then calculating its coordinates (x, y), pressure value and timestamp, and transmitting them to the device connected to the drawing board 1, such as a computer, mobile phone, etc. The main control chip 15 in this embodiment can adopt a processing chip such as STM, CPU, etc. according to actual conditions.

[0051] The inductive coils 16 are arranged in a grid in a uniform manner horizontally and vertically, dividing the drawing board 1 into inductive units with an accuracy of 0.1mm. When the electromagnetic pen moves above the drawing board, it cuts the criss-crossing magnetic fields generated by these coils, thereby generating an induced electromotive force in the coils and forming an electrical signal to determine the exact position of the electromagnetic pen 10.

[0052] (3) Electromagnetic pen 10 (see Figure 3 ): It includes a pressure sensor 12, an electromagnetic transmitting coil 13, a pen tip 14 and a signal amplifier 17, which are used to transmit position, pressure and timestamp signals to the electromagnetic module 2. Among them, the pen tip 14 fits tightly to the surface of the drawing board 1 and is made of nano-silicone material to ensure smooth and unobstructed sliding, perfectly simulating the texture of real brushstrokes.

[0053] The pressure sensor 12 is used to detect the pressure value in real time and convert it into an electrical signal. Specifically, when the user presses the switch button 5 to turn on the drawing board, and presses the function key 3 to start drawing after connecting the Bluetooth adapter, when the electromagnetic pen 10 is used to apply pressure on the drawing board 1, the pressure sensor 12 will sense the change in pressure and convert it into an electrical signal, which is then transmitted to the electromagnetic module 2 through the circuit inside the electromagnetic pen 10, thereby achieving control of line thickness, lightness and other effects according to the pressure size. The power is provided by the power supply module 11.

[0054] The electromagnetic transmitting coil 13 is used to generate an electromagnetic field of a specific frequency. When the pen tip moves above the drawing board, the coil generates an electromagnetic field of a specific frequency, which interacts with the electromagnetic field of the electromagnetic module 2, allowing the drawing board 1 to sense the position and movement trajectory of the electromagnetic pen 10.

[0055] The signal amplifier 17 is used to amplify weak electromagnetic signals and transmit them to the main control chip 15 to ensure signal stability.

[0056] (4) Data processing module and server: wherein the data processing module is connected to the electromagnetic module 2, and is used to analyze the pixel timing data and calculate at least 20 painting behavior features; the server is used to receive the painting behavior features, and perform feature selection, fusion and modeling on the painting behavior features through machine learning, and generate an evaluation report that scores the originality, flexibility, fluency and precision of children's thinking.

[0057] The drawing behavior features in this embodiment include, but are not limited to, the time when the pen starts to be put down, the total number of pen strokes, the total handwriting distance, the total drawing time, the total thinking time, the average pen stroke distance, the average pen stroke speed, the average pen stroke time, the average thinking time, the speed variability, the pen stroke time variability, the pen stroke distance variability, the thinking time variability, the breadth of handwriting distribution, the handwriting density, the difference in quadrant proportion, the starting angle of the pen, the distance from the starting point to the origin, the average pen pressure, the pen pressure variability, etc.

[0058] The following describes the process of assessing children's higher-level cognition and emotions using the above-mentioned device:

[0059] Step 1: Collecting drawing pixel timing information

[0060] When children draw on the smart drawing board, the pixel information of handwriting and non-handwriting will be continuously collected at a frequency of 240 times per second, and the position of the pixel point where the brush is located, the state of the pen, and the time of writing will be collected each time.

[0061] Position information collection: Pixel position information is recorded in the form of two-dimensional coordinates (x, y), and the position of each pixel point collected must be recorded.

[0062] Pen status collection: The pen status of each collection is represented by the numbers 0 and 1. If the pen touches the drawing board and produces handwriting, it is detected as a pen down and recorded as 1, otherwise it is recorded as 0.

[0063] Time data collection: The time data recorded here is the relative time of this collection, that is, the time relative to the start of painting.

[0064] One of the recording forms of pixel timing information is shown in Table 1:

[0065]

[0066] Table 1

[0067] Step 2: Handwriting information analysis and index calculation

[0068] By collecting the pixel position data, pen placement status, and time data, the characteristics of children's painting behavior during the painting process can be calculated.

[0069] Step 3: Analyze data and output

[0070] For the assessment of specific high-level cognitive abilities (creativity, intelligence, curiosity, etc.) or emotional development, feature selection, fusion, and predictive modeling are performed through machine learning to automatically output the ability index scores of high-level cognitive abilities and emotional development and provide an assessment report.

[0071] The following example demonstrates the effectiveness of the evaluation of this embodiment.

[0072] This case collected pixel time series data for 480 images from 30 participants. The evaluation process was as follows:

[0073] (1) Data collection process: All 30 participants drew on a drawing board. Each person was required to complete 16 drawings, and each drawing took 2 minutes. A computer connected to the drawing board collected pixel timing data for each drawing, collecting data 120 times per second. Each collection recorded the pixel position information (two-dimensional coordinates x, y), the pen status (1 indicates pen down, 0 indicates no pen down), the pen pressure, and the timestamp data.

[0074] (2) Data preprocessing: The pixel time series data collected for each painting forms the painting data time series corresponding to the painting, and the time series is preprocessed. According to the time series data, 20 handwriting behavior indicators are calculated. Including: start time of writing, total number of strokes, total handwriting distance, average stroke distance, average stroke speed, total painting time, average stroke time, total thinking time, average thinking time, speed variability, stroke time variability, stroke distance variability, thinking time variability, handwriting distribution breadth, handwriting density, quadrant proportion difference, starting angle, starting distance from the origin, average pen pressure, pen pressure variability. The behavioral indicators are shown in Table 2:

[0075]

[0076] Table 2

[0077] (3) Feature selection: Five experts scored the creativity of the paintings, evaluating each painting on four aspects: originality, flexibility, fluency, and refinement. Feature selection was performed on each score, and a linear mixed model was used to explore which handwriting behavior indicators (the above 20 types) could predict the score. These indicators were then used to predict the corresponding creativity score.

[0078] The process of feature selection for each score is as follows:

[0079] 1) Data collection and preprocessing: Using sensors or data acquisition devices, 20 painting behavior characteristics are recorded and quantified in real time;

[0080] 2) Each painting was scored by four raters (with an ICC of 0.7 or higher) across four dimensions: fluency, flexibility, originality, and refinement. These ratings served as the target variable for correlation analysis with the behavioral profile data.

[0081] 3) Preliminary feature screening: Pearson correlation coefficient analysis is used to preliminarily screen out handwriting behavior features that have a strong correlation with the four dimensions. Among them, the correlation threshold (such as r>0.3) or significance level (such as p<0.05) is used as the threshold standard for feature selection. This process includes testing the linear or nonlinear relationship between each handwriting behavior feature and the scores of the four dimensions to confirm which behavior features have significant predictive power in the scores of different dimensions;

[0082] 4) Regression model training and verification: The selected handwriting behavior features are used as input variables and trained using a multiple linear regression model. During the training process, the prediction accuracy of the model is evaluated through methods such as cross-validation, and the feature selection is further optimized based on the results. In this process, the significance of each behavior feature is confirmed through t-test and F-test, and a decision is made whether to include it in the final regression model. After confirming the handwriting behavior features related to the dimension scoring, the prediction effect of different data sets is tested and verified to further confirm the predictive ability of the selected features. Ensure that each selected handwriting behavior feature is reliable and stable in predicting the scoring dimension.

[0083] The final results of feature selection are shown in Table 3:

[0084]

[0085]

[0086] Table 3

[0087] (4) Machine learning model training: The four creativity scores are used as dependent variables, and the painting behavior indicators that can effectively predict the creativity scores are used as independent variables. Machine learning models are trained using machine learning (general linear model). A machine learning model is generated for each painting creativity indicator. A total of four machine learning models are generated, which are used to predict the originality, flexibility, fluency and refinement of individual painting results. The details are as follows:

[0088] Originality = -0.15*total handwriting distance + 0.14*total drawing time + 0.18*speed variability + 0.21*handwriting distribution breadth -0.11*quadrant proportion difference + 0.09*pen pressure variability;

[0089] Flexibility = 0.18*time to start writing - 0.07*total number of strokes - 0.1*quadrant proportion difference + 0.14*thinking time variability + 0.11*distance from the starting point to the origin;

[0090] Fluency = 0.28*total number of strokes + 0.19*breadth of handwriting distribution - 0.18*thinking time variability + 0.29*average stroke distance;

[0091] Fineness = 0.17*total number of strokes + 0.22*stroke density + 0.19*average stroke distance + 0.15*speed variability.

[0092] The coefficient represents the contribution weight of the prediction, a positive coefficient represents a positive prediction, and a negative coefficient represents a negative prediction.

[0093] (5) Predictive validity test: The predictive validity of the four generated machine learning models was tested. The results showed that all four creativity scores could be significantly predicted by the automated scoring of the machine learning model.

[0094] The correlation analysis results between the automated scoring of the machine learning model and the actual scoring are as follows: Figure 4 As shown in the data, for originality, there is a significant positive correlation between the automated evaluation score based on the temporal information of painting pixels (hereinafter referred to as "automated scoring") and the manually evaluated creativity score (hereinafter referred to as "manual scoring") (r=0.3655, p<0.01); for flexibility, there is a significant positive correlation between the automated scoring and the manual scoring (r=0.0979, p=0.032); for fluency, there is a significant positive correlation between the automated scoring and the manual scoring (r=0.5151, p<0.01); for elegance, there is a significant positive correlation between the automated scoring and the manual scoring (r=0.4998, p<0.01).

[0095] This shows that the prediction of painting creativity based on pixel timing information is effective. Therefore, this embodiment has practical feasibility for automatically evaluating children's high-level cognitive abilities based on painting pixel timing information.

[0096] The present invention collects dynamic behavioral characteristics of children's drawing process based on high-frequency time-series data, screens key features, and then constructs a multivariate linear regression model to achieve automated assessment of higher-order cognitive abilities. This not only solves the data monotony problem of static image analysis, but also transforms subjective cognitive abilities into objective ones, replacing inefficient manual assessments. Furthermore, the assessment efficiency is significantly higher than that of existing technologies, and the consistency matches that of expert scores (r = 0.37-0.52), reaching a clinically applicable level and meeting the demand for real-time feedback. Therefore, compared with existing technologies, the present invention has outstanding substantive features and significant improvements, making it very suitable for large-scale promotion and application.

[0097] The above embodiments are only preferred implementation modes of the present invention and should not be used to limit the scope of protection of the present invention. Any changes or modifications that are made to the main design concept and spirit of the present invention and have no substantive significance, as long as the technical problems they solve are still consistent with the present invention, should be included in the scope of protection of the present invention.

Claims

1. A method for evaluating children's higher-order cognition and emotions based on the temporal information of painting pixels, characterized by: A device for evaluating children's higher-order cognition and emotions based on painting pixel temporal information, the device for evaluating children's higher-order cognition and emotions based on painting pixel temporal information includes: A drawing board (1) has a built-in electromagnetic module (2), wherein the electromagnetic module (2) includes a main control chip (15) and an electromagnetic induction coil (16) for real-time acquisition of pixel timing data of an electromagnetic pen (10); The electromagnetic pen (10) comprises a pressure sensor (12), an electromagnetic transmitting coil (13), a pen tip (14) and a signal amplifier (17), and is used to transmit position, pressure and time stamp signals to the electromagnetic module (2); A data processing module, connected to the electromagnetic module (2), for analyzing the pixel time series data and calculating at least 20 painting behavior features; The server is configured to receive the drawing behavior characteristics, and perform feature selection, fusion, and modeling on the drawing behavior characteristics through machine learning to generate an evaluation report that scores the originality, flexibility, fluency, and sophistication of the child's thinking; The method for assessing children's higher-order cognition and emotions includes the following steps: (1) Collect pixel temporal data on the child’s drawing board, including position, pressure, timestamp, and pen-down status; (2) parsing the data, restoring the painting pattern, and calculating at least 20 painting behavior features; (3) Using a machine learning model to select, fuse, and model the characteristics of drawing behavior, and then generate automated evaluation results of children's originality, flexibility, fluency, and sophistication based on the selected drawing behavior characteristics; the calculation formulas for originality, flexibility, fluency, and sophistication are as follows: Originality = -0.15*total handwriting distance + 0.14*total drawing time + 0.18*speed variability + 0.21*handwriting distribution breadth -0.11*quadrant proportion difference + 0.09*pen pressure variability; Flexibility = 0.18*time to start writing - 0.07*total number of strokes - 0.1*quadrant proportion difference + 0.14*thinking time variability + 0.11*distance from the starting point to the origin; Fluency = 0.28*total number of strokes + 0.19*breadth of handwriting distribution - 0.18*thinking time variability + 0.29*average stroke distance; Fineness = 0.17*total number of strokes + 0.22*stroke density + 0.19*average stroke distance + 0.15*speed variability; (4) Output an evaluation report containing the scores.

2. The method for evaluating children's higher-order cognition and emotions based on painting pixel temporal information according to claim 1, characterized in that: The pixel timing data is collected at a frequency of 240 times per second, and the collected data includes two-dimensional coordinates, pen-down status, pressure value and timestamp; the pen-down status uses 0 or 1 to indicate whether the pen is down.

3. The method for evaluating children's higher-order cognition and emotions based on painting pixel temporal information according to claim 1 or 2, characterized in that: The electromagnetic induction coils (16) are arranged in a balanced manner horizontally and vertically, and are used to divide the drawing board (1) into a plurality of induction areas to capture the precise position of the electromagnetic pen (10).

4. The method for evaluating children's higher-order cognition and emotions based on painting pixel temporal information according to claim 1, characterized in that: The machine learning model screens painting behavior features through statistical analysis and optimizes weights based on 10-fold cross-validation.

5. The method for evaluating children's higher-order cognition and emotions based on painting pixel temporal information according to claim 4, characterized in that: The specific process of the machine learning model for screening painting behavior characteristics is as follows: through statistical analysis, the behavioral characteristics that are significantly correlated with each dimension are screened, and the screening criteria are: the absolute value threshold of the Pearson correlation coefficient r>0.3 or the significance level p<0.05; then the screened behavioral characteristics are used as input variables to train a multiple linear regression model, and the parameters of the multiple linear regression model are optimized through cross-validation.

6. The method for evaluating children's higher-order cognition and emotions based on painting pixel temporal information according to claim 5, characterized in that: During the training of the multiple linear regression model, the significance of each behavioral feature was confirmed by t-test and F-test to determine whether it should be included in the final multiple linear regression model.

Images