Cognitive training scene dynamic generation method and system

Abstract

The invention discloses a cognitive training scene dynamic generation method and system, and belongs to the technical field of cognitive training, and the method comprises the steps: obtaining a current scene parameter and a user response index; according to the current scene parameters and the user response indexes, determining the response time standard deviation, the error rate change gradient and the thermodynamic diagram dispersion of the user and the weights of the response time standard deviation, the error rate change gradient and the thermodynamic diagram dispersion; performing weighted fusion on the response time standard deviation, the error rate change gradient and the thermodynamic diagram dispersion through the weights of the response time standard deviation, the error rate change gradient and the thermodynamic diagram dispersion, and determining a difficulty coefficient of the scene; and determining a next training scene according to the difficulty coefficient of the scene. Accurate and dynamic adjustment of the next training scene is realized according to the current scene parameter and the user response index, so that personalized cognitive ability training is provided. The problem that a current training scene cannot be dynamically adjusted according to user behaviors is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cognitive training, and particularly relates to a cognitive training scene dynamic generation method and system. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute the prior art.

[0003] For people with cognitive impairment, cognitive training is currently a relatively effective treatment method.

[0004] In related methods, a fixed cognitive training scene is provided for cognitive training of a user, and the fixed cognitive training scene is prone to cause a problem of weakened user response during long-term cognitive training, resulting in limited cognitive training effect on the user. SUMMARY

[0005] To solve the above problems, the present application provides a cognitive training scene dynamic generation method and system, which realizes accurate dynamic adjustment of the next training scene according to the current scene parameters and the user response index, thereby providing personalized cognitive ability training and effectively improving the cognitive training effect of the user.

[0006] To achieve the above purpose, the present application adopts the following technical solutions: In a first aspect, a cognitive training scene dynamic generation method is provided, comprising: obtaining current scene parameters and user response indexes; determining the response time standard deviation, error rate change gradient, heat map dispersion, and the weights of the response time standard deviation, error rate change gradient, and heat map dispersion of the user according to the current scene parameters and user response indexes; weighting and fusing the response time standard deviation, error rate change gradient, and heat map dispersion by the weights of the response time standard deviation, error rate change gradient, and heat map dispersion to determine the difficulty coefficient of the scene; determining the next training scene according to the difficulty coefficient of the scene.

[0007] Further, the current scene parameters include the number of interference items, time limit, layout complexity, and effective target density of the current scene; The user response indexes include the number of user reflections, success hit rate, and response time of each reflection; The response time standard deviation is obtained according to the number of user reflections and the response time of each reflection; The error rate change gradient is calculated and determined according to the success hit rate, success hit rate, effective target density, and time limit; According to the number of interference terms, the layout complexity and the effective target density, the dispersion of the heat map is calculated and determined.

[0008] Further, according to the success hit rate, the success hit rate, the effective target density and the time limit, the error rate of the current scene is calculated. According to the error rate of the current scene and the error rate of the last scene, the error rate change gradient is calculated and determined.

[0009] Further, the success hit rate is the success hit rate of the two consecutive training scenes.

[0010] Further, according to the current scene parameters, the user response indicators and the trained weight prediction model, the weights of the response time standard deviation, the error rate change gradient and the heat map dispersion are obtained; wherein the weight prediction model takes the scene parameters and the user response indicators as inputs, takes the weights of the response time standard deviation, the error rate change gradient and the heat map dispersion as outputs, and is constructed by using an LSTM model.

[0011] Further, according to the difficulty coefficient of the scene, the difficulty level of the scene is determined. According to the difficulty level of the scene, the next training scene is determined.

[0012] In a second aspect, a cognitive training scene dynamic generation system is provided, comprising: A data acquisition unit is configured to acquire current scene parameters and user response indicators. An indicator and indicator weight determination unit is configured to determine the response time standard deviation, the error rate change gradient, the heat map dispersion and the weights of the response time standard deviation, the error rate change gradient and the heat map dispersion of the user according to the current scene parameters and the user response indicators. A difficulty coefficient determination unit is configured to determine the difficulty coefficient of the scene by weighting and fusing the response time standard deviation, the error rate change gradient and the heat map dispersion according to the weights of the response time standard deviation, the error rate change gradient and the heat map dispersion. A next training scene determination unit is configured to determine the next training scene according to the difficulty coefficient of the scene.

[0013] In a third aspect, a computer device is provided, comprising: A processor adapted to execute a computer program; A computer readable storage medium having a computer program stored therein, wherein the computer program is executed by the processor to implement the cognitive training scene dynamic generation method of the first aspect.

[0014] In a fourth aspect, a computer readable storage medium is provided, which stores a computer program, and the computer program is adapted to be loaded by a processor and execute the cognitive training scene dynamic generation method of the first aspect.

[0015] In a fifth aspect, a computer program product is provided, which comprises a computer program, and the computer program is executed by a processor to implement the cognitive training scene dynamic generation method of the first aspect.

[0016] Compared with the prior art, the present application has the following beneficial effects: The cognitive training scene dynamic generation method and system provided by the present application, when performing cognitive training on a user, obtains current scene parameters and user response indicators, and determines the response time standard deviation, error rate change gradient, heat map dispersion, and the weights of the response time standard deviation, error rate change gradient, and heat map dispersion according to the current scene parameters and user response indicators. The response time standard deviation, error rate change gradient, and heat map dispersion are weighted and fused by the weights of the response time standard deviation, error rate change gradient, and heat map dispersion to determine the difficulty coefficient of the scene. By fusing the response time standard deviation, error rate change gradient, and heat map dispersion and updating the weights of the three indicators in real time, the difficulty coefficient of the scene is accurately determined. Then, the next training scene is accurately determined by using the difficulty coefficient, the dynamic generation of the cognitive training scene is realized, and personalized cognitive ability training is provided according to the real-time performance of the user, thereby improving the effect of the user's cognitive training.

[0017] The advantages of the additional aspects of the present application will be partially given in the following description, partially become obvious from the following description, or be understood by the practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0018] The accompanying drawings, which form a part of the present application, are included to provide further understanding of the present application, and serve to explain the illustrative embodiments of the present application and to enable an artisan to make and use the same. The drawings illustrate embodiments of the present application and, together with the description, serve to explain the principles of the present application.

[0019] Figure 1 A cognitive training scene dynamic generation method flowchart disclosed by the embodiments. DETAILED DESCRIPTION

[0020] The present application will be further described below in conjunction with the drawings and embodiments.

[0021] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise indicated, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0022] It is to be noted that the terms used herein are only intended to describe specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise, and it is further understood that the terms "comprising" and / or "including" when used in this specification, specify the presence of stated features, steps, operations, devices, components and / or combinations thereof.

[0023] Embodiment 1 For people with cognitive impairment, cognitive training is currently the most effective treatment method.

[0024] The mainstream technical solutions of current cognitive training mainly include three categories: one is the template scene generation technology, which generates training scenes by pre-setting a fixed scene library and using a random extraction mechanism, but the scene element quantity, layout rule and other parameters are fixed, and cannot be dynamically adjusted according to individual ability; the second is the rule type difficulty adjustment technology, which linearly calculates the difficulty based on experience formula according to single indicators such as reaction time and error times, resulting in inaccurate difficulty adjustment; the third is the artificial intervention system, which relies on therapists to manually set training parameters and phase evaluation, resulting in significant lag in adjustment (usually delayed for 1-2 training periods).

[0025] The disadvantages of existing cognitive training technologies include: 1. Medical standard disconnection problem: the existing system does not establish a mapping relationship between scene parameters and DSM-5 clinical indicators. For example, "persistent inattention" (Criterion A1) in the ADHD diagnosis standard is not quantified as a dynamic generation rule for interference items in the scene, resulting in a training target deviation rate of more than 60%.

[0026] 2. Data closed loop missing: traditional solutions only realize one-way scene generation, lacking dynamic adjustment mechanism based on biological feature feedback.

[0027] 3. Discrete evaluation mechanism: difficulty adjustment relies on regular manual evaluation (usually once a week), which is spatially and temporally disconnected from the continuous training process. Clinical data shows that this method causes "difficulty cliff" phenomenon (difficulty and ability mismatch > 40%) in 56% of ADHD children during the evaluation interval.

[0028] In order to solve the above problems, in the present embodiment, a cognitive training scene dynamic generation method is proposed, which constructs a four-level improvement system: (1) To address the issue of reduced cognitive response in children during long-term training with fixed scene templates, a natural language-driven dynamic parameterization generation technique is proposed. This technique converts natural language descriptions (e.g., "generate a visual search scene with progressive interference") into dynamically adjustable scene topology parameters (element density δ ∈ [0.2, 0.8], layout complexity χ 2 ≥1.35) through an LLM instruction analysis engine, achieving a 2.7-fold improvement in scene template variability (SDI index from 0.32 to 0.86). (2) To overcome the industry bottleneck of disconnection between clinical indicators and training parameters, a conditional library constrained by DSM-5 standards is constructed, encoding the 9 core diagnostic criteria of ADHD into a 21-dimensional constraint vector. A multi-layer perceptron (MLP) is used to establish a dynamic mapping relationship between scene parameters and clinical indicators (R 2 =0.91), reducing the training target deviation rate from 60.7% to 8.3%. (3) To address the cognitive load shock caused by passive difficulty adjustment, an active feedback mechanism based on attention heat maps is designed. Kernel density estimation (bandwidth h=0.05) is used to generate real-time heat distribution maps, and a spatiotemporal feature extraction module (including 3D convolution layers and LSTM layers) is used to predict difficulty coefficients, achieving dynamic parameter fine-tuning every 5 seconds (adjustment accuracy ±0.15 levels). (4) To overcome the low efficiency of manual configuration, a multi-modal interactive instruction analysis framework is developed, integrating speech recognition (MFCC feature extraction + CTC decoding) and text analysis (Bi-LSTM + CRF model) dual channels, supporting end-to-end conversion of unstructured instructions to structured parameters (conversion time <800ms), improving therapist operation efficiency by 6.2 times.

[0029] The cognitive training scene dynamic generation method disclosed in this embodiment dynamically generates training scenes based on current scene parameters and user response indicators, providing personalized cognitive ability training according to real-time performance. The application system architecture of this method includes four core modules, and their relationships are as follows: (1) Natural language interaction module: As the input layer of the system, a multi-modal fusion ubiquitous interaction interface is constructed, with visual motion capture as the core, integrating two perception channels: High-precision motion capture system: A depth sensor array is used to track skeletal joints, and when the amplitude of the captured motion exceeds a threshold, a standardized motion semantic code is generated.

[0030] Intelligent terminal control screen module: Develop a low-latency mirroring protocol (based on WebSocket optimization) that supports inputting complex commands on the control screen through pen gestures (such as drawing a circle to select and sliding to adjust). After the touch events are classified by the Gaussian Mixture Model (GMM), structured operation instruction sets are generated.

[0031] Finally, capture the therapist's and user's instructions, behavior data, and convert unstructured natural language or action information into structured data that the system can process.

[0032] (2) Semantic analysis engine As an intermediate processing layer, it receives the input of the natural language interaction module, analyzes the instruction semantics, and converts it into a standardized parameter set.

[0033] In this semantic analysis engine, the present embodiment proposes a medical enhanced LLM architecture, which relies on LLaMA-213B as the base model and continuously optimizes its processing ability for medical professional terms and clinical descriptions through domain adaptation training. First, the base model is pre-trained and fine-tuned on large-scale medical text data to ensure its stability and accuracy in recognizing and understanding complex semantics. Second, to fully integrate authoritative medical knowledge, the system constructs approximately 153,000 triples from the DSM-5 diagnostic criteria, ADHD treatment guidelines, and other content, with each triple consisting of (subject-relation-object). These triples include: 1. Diagnostic criteria triples: For example, (attention deficit symptoms - manifested as - inability to focus on details). 2. Symptom and training mapping triples: For example, (persistent attention deficit - requires training - selective attention). 3. Training parameter association triples: For example, (selective attention training - parameter setting - increase in the number of interference items). 4. Difficulty parameter constraint triples: For example, (primary difficulty - interference item number range - 3 to 5). These triples are stored in RDF format and constitute the core knowledge base of the system, enabling the semantic analysis engine to accurately map professional medical knowledge to training scenario parameters and fine-tune model parameters using the LoRA method to achieve knowledge injection and enhance the model's sensitivity to medical backgrounds. The model processed in the above manner can perform deep semantic analysis on the input natural language instructions and output structured parameter sets in JSON format, containing 6 core dimensions: N i : Interference items refer to non-target but similar characteristic task elements that appear as camouflage in color, shape, position, language, or action in different task scenarios, mainly used to increase attention suppression difficulty and improve selective processing ability.

[0034] 2. Time limit T lim .

[0035] 3. Layout complexity χ 2 : To measure the degree of dispersion of the spatial distribution of elements in the training scene, based on the chi-square deviation statistics to compare the deviation value of the current distribution and the theoretical grid model. The higher the layout complexity, the more significant the visual redundancy, spatial occlusion and navigation difficulty in the scene.

[0036] 4. Effective target density D t .

[0037] 5. User reaction times R n .

[0038] 6. Success hit rate H r .

[0039] These six core dimensions provide accurate data support for the subsequent dynamic generation of cognitive training scenes.

[0040] (3) Dynamic generation system As the execution layer, based on the six core parameter sets output by the semantic analysis engine, combined with real-time feedback data, generate cognitive training scenes that meet the training objectives.

[0041] The cognitive training system works collaboratively through the natural language interaction module, semantic analysis engine, and dynamic generation system, forming a complete data loop: therapist input → semantic analysis → scene generation → user interaction → data collection → parameter adjustment → dynamic scene update, thereby realizing personalized and precise cognitive training.

[0042] As shown in Figure 1 , the cognitive training scene dynamic generation method disclosed in this embodiment, as shown in Figure 1 , includes: Obtain the current scene parameters and user response indicators; According to the current scene parameters and user response indicators, determine the response time standard deviation, error rate change gradient, heat map dispersion, and the weights of the response time standard deviation, error rate change gradient, and heat map dispersion; Weight the response time standard deviation, error rate change gradient, and heat map dispersion by their weights to determine the difficulty coefficient of the scene; According to the difficulty coefficient of the scene, determine the next training scene.

[0043] In the cognitive training of the user, the current scene parameters and the user response indicators are obtained; and according to the current scene parameters and the user response indicators, the response time standard deviation, the error rate change gradient, the heat map dispersion, and the weights of the response time standard deviation, the error rate change gradient, and the heat map dispersion of the user are determined; the response time standard deviation, the error rate change gradient, and the heat map dispersion are weighted and fused through the weights of the response time standard deviation, the error rate change gradient, and the heat map dispersion, so as to determine the difficulty coefficient of the scene; the three indicators of the response time standard deviation, the error rate change gradient, and the heat map dispersion are fused, and the weights of the three indicators are updated in real time, so as to accurately determine the difficulty coefficient of the scene; and then the difficulty coefficient is used to accurately determine the next training scene, so as to realize the dynamic generation of the cognitive training scene and provide personalized cognitive ability training according to the real-time performance of the user, thereby improving the effect of the cognitive training of the user.

[0044] In the related art, a linear regression model is used to calculate the difficulty level according to a single indicator (for example, the response time or the number of errors) according to an empirical formula, and then the next training scene is determined according to the difficulty level, for example:

[0045] This method is difficult to comprehensively reflect the behavior fluctuation and attention distribution of the user, so that the determined difficulty level is not accurate.

[0046] However, the embodiment no longer relies on a single behavior indicator for difficulty evaluation, but comprehensively uses multi-dimensional data to capture the global characteristics of the user behavior, generates the difficulty coefficient of the scene, and then determines the difficulty level of the scene according to the difficulty coefficient of the scene; and determines the next training scene according to the difficulty level of the scene.

[0047]

[0048] In the formula, is the response time standard deviation, is the error rate change gradient, is the heat map dispersion; , and are the weights of , and respectively.

[0049] The current scene parameters of the embodiment include the number of interference items of the current scene N i , the time limit T lim , the layout complexity χ 2 , and the effective target density D t ; The user response indicators include the number of user reflections R n , the success hit rate H r , and the response time of each reflection ; According to the number of user reflections and the response time of each reflection, the response time standard deviation is obtained; According to the success hit rate, the success hit rate, the effective target density and the time limit, the error rate change gradient is calculated and determined; According to the number of interference terms, the layout complexity and the effective target density, the heat map dispersion is calculated and determined.

[0050] Among them, according to the success hit rate, the success hit rate, the effective target density and the time limit, the error rate of the current scene is calculated; According to the error rate of the current scene and the error rate of the last scene, the error rate change gradient is calculated and determined.

[0051] The success hit rate is the success hit rate of the last two training scenes.

[0052] The response time standard deviation Can reflect the fluctuation degree of the user in the reaction speed:

[0053] Among them, And The system training mean; , , The adjustment parameter is specifically between 0.1 and 0.2; Based on The response time of the operation record is; The average response time.

[0054] The error rate change gradient Show the dynamic trend of error behavior over time:

[0055] Among them, The error rate, Based on the success hit rate in the last two training units, The training task standard time, , The adjustment coefficient is specifically between 0.1 and 0.2; The task interval time.

[0056] The heat map dispersion Reflects the distribution of user attention in the scene:

[0057] wherein, is the original dispersion.

[0058] More importantly, the weights α , β , γ are no longer fixed empirical values, but are obtained according to the current scene parameters, user response indicators and trained weight prediction model, the weights of response time standard deviation, error rate change gradient and heat map dispersion; wherein, the weight prediction model takes scene parameters and user response indicators as input, and takes the weights of response time standard deviation, error rate change gradient and heat map dispersion as output, and is obtained by using LSTM model. The specific process is as follows: Read the user historical training scene data sequence, which contains at least 20 groups of training scene performance data; real-time acquisition of user response data, mainly visual motion capture; parameters of the current training scene, involving interference term density and time limit.

[0059] Through 3-layer bidirectional LSTM, after time series feature extraction, full connection layer mapping is performed.

[0060] Directly output the optimized values of the three weights α , β , γ , wherein, α is the weight of response time standard deviation (range 0.2-0.6), β is the weight of error rate change gradient (range 0.1-0.5), and γ is the weight of heat map dispersion (range 0.2-0.4).

[0061] Every time a training scene is completed, the system re-collects the user response data corresponding to the training scene, and the LSTM network completes the secondary evaluation based on the newly collected data.

[0062] This dynamic adjustment mechanism makes the system automatically optimize the weight distribution using historical and real-time data, so that the calculation of each indicator weight is more accurate and adaptive.

[0063] For example, the parameters of a certain cognitive training scene are as follows: 6; 30; χ 2 2.1; 0.35; 10; 0.7; For [1.2, 1.1, 1.3, 1.0, 1.4, 1.2, 1.0, 1.1, 1.3, 1.2].

[0064] System default parameter mean value as follows: For 5; For 1.5; For 30.

[0065] Experience adjustment coefficient as follows: = 0.1, = 0.15, = 0.2; = 0.2, = 0.1; = 0.15, = 0.25, = 0.2; = 30, = 0.25, = 2.5.

[0066] Calculation process as follows: Average response time:

[0067] Standard deviation:

[0068] Error rate of current scene:

[0069] Error rate change gradient:

[0070] Heat map dispersion:

[0071] Difficulty coefficient: Assume that the current dynamic weight of the system: α = 0.4, β = 0.3, γ = 0.3, so

[0072] According to the difficulty coefficient, determine the difficulty level: Set the grading rules: For low difficulty level; For medium difficulty level; For high difficulty level. Therefore For high difficulty level.

[0073] The high-difficulty cognitive training scene is provided as the next training scene to the user, and the user can perform cognitive training.

[0074] The cognitive training scene dynamic generation method disclosed by the embodiment fully considers the correlation and interaction between multi-dimensional data, significantly improves the capturing ability of the system on complex behavior patterns, and provides more scientific and effective support for real-time dynamic generation of cognitive training scenes.

[0075] Embodiment 2 In this embodiment, a cognitive training scene dynamic generation system is disclosed, comprising: A data acquisition unit is configured to acquire current scene parameters and user response indicators. An indicator and indicator weight determination unit is configured to determine the response time standard deviation, error rate change gradient, heat map dispersion, and weights of the response time standard deviation, error rate change gradient, and heat map dispersion according to the current scene parameters and user response indicators. A difficulty coefficient determination unit is configured to determine the difficulty coefficient of the scene by weighting and fusing the response time standard deviation, error rate change gradient, and heat map dispersion according to the weights of the response time standard deviation, error rate change gradient, and heat map dispersion. A next training scene determination unit is configured to determine the next training scene according to the difficulty coefficient of the scene.

[0076] The application also discloses a computer device, which comprises: A processor is adapted to execute a computer program. A computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the cognitive training scene dynamic generation method disclosed in embodiment 1.

[0077] The application also discloses a computer readable storage medium, which stores a computer program, and the computer program is adapted to be loaded and executed by a processor to implement the cognitive training scene dynamic generation method disclosed in embodiment 1.

[0078] The application also discloses a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the cognitive training scene dynamic generation method disclosed in embodiment 1.

[0079] The method disclosed in the embodiment 1 can be directly embodied by a hardware processor or by a combination of hardware and software modules in the processor. The software modules can be located in a storage medium in the art, such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, an electrically erasable programmable memory, a register, or the like. The storage medium is located in the memory, and the processor reads information in the memory and combines the hardware to complete the steps of the above method. To avoid repetition, no further detailed description is given herein.

[0080] Those skilled in the art can understand that the units and algorithm steps of the examples described in combination with the embodiments can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software manner depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0081] Although the specific embodiments of the present application are described above in combination with the drawings, the description is not a limitation on the scope of protection of the present application. Those skilled in the art should understand that various modifications or variations made by those skilled in the art on the basis of the technical solutions of the present application without creative labor are still within the scope of protection of the present application.

Claims

1. A method for dynamically generating cognitive training scenarios, characterized in that, include: Obtain current scene parameters and user response metrics; Based on the current scenario parameters and user response metrics, determine the standard deviation of user response time, gradient of error rate change, heatmap dispersion, and the weights of the standard deviation of response time, gradient of error rate change, and heatmap dispersion. The difficulty coefficient of the scenario is determined by weighting the standard deviation of response time, the gradient of error rate change, and the dispersion of heatmap. The next training scenario is determined based on the difficulty level of the scenario.

2. The method for dynamically generating cognitive training scenarios as described in claim 1, characterized in that, The current scene parameters include the number of interfering items, time limit, layout complexity, and effective target density. User response metrics include the number of user responses, the success rate, and the response time for each response. The standard deviation of response time is obtained based on the number of user feedbacks and the response time for each feedback. The error rate change gradient is calculated and determined based on the success rate, effective target density, and time limit. The heatmap dispersion is calculated and determined based on the number of interference items, layout complexity, and effective target density.

3. The method for dynamically generating cognitive training scenarios as described in claim 2, characterized in that, Calculate the error rate for the current scenario based on the success rate, effective target density, and time limit; The error rate change gradient is calculated and determined based on the error rate of the current scenario and the error rate of the previous scenario.

4. The method for dynamically generating cognitive training scenarios as described in claim 2, characterized in that, The success rate is the success rate of two consecutive training scenarios.

5. The method for dynamically generating cognitive training scenarios as described in claim 1, characterized in that, Based on the current scene parameters, user response metrics, and the trained weighted prediction model, the weights of the response time standard deviation, error rate change gradient, and heatmap dispersion are obtained. The weighted prediction model takes the scene parameters and user response metrics as input and the weights of the response time standard deviation, error rate change gradient, and heatmap dispersion as output, and is constructed using an LSTM model.

6. The method for dynamically generating cognitive training scenarios as described in claim 1, characterized in that, Determine the difficulty level of the scene based on its difficulty coefficient; The next training scenario is determined based on the difficulty level of the scenario.

7. A cognitive training scenario dynamic generation system, characterized in that, include: The data acquisition unit is used to acquire current scene parameters and user response metrics; The indicator and indicator weight determination unit is used to determine the standard deviation of user response time, gradient of error rate change, heatmap dispersion, and weights of the standard deviation of response time, gradient of error rate change, and heatmap dispersion based on the current scenario parameters and user response indicators. The difficulty coefficient determination unit is used to determine the difficulty coefficient of the scene by weighting and fusing the response time standard deviation, error rate change gradient, and heatmap dispersion. The next training scenario determination unit is used to determine the next training scenario based on the difficulty coefficient of the scenario.

8. An electronic device, characterized in that, The device includes: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the method for dynamically generating cognitive training scenarios according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed by a method for dynamically generating cognitive training scenarios according to any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements a method for dynamically generating cognitive training scenarios as described in any one of claims 1-6.

Images