An interactive training system based on backtracking memory

Abstract

The application relates to an interactive training system based on backtracking memory, which comprises a collection module, a task scheduling module, an interaction module and a dynamic difficulty adjustment module. The collection module is used for collecting first brain electrical signals generated by a user in the process of processing a first training task in a memory encoding stage and collecting second brain electrical signals and behavior data generated by the user in the process of processing a second training task in an extraction stage. The task scheduling module is used for generating the first training task according to a current difficulty parameter in the memory encoding stage and outputting the first training task, outputting the second training task in the extraction stage, and judging whether to end the output of the second training task based on the behavior data and the second brain electrical signals collected by the collection module. The interaction module is used for triggering an attention reminding or a situation strengthening strategy based on the first brain electrical signals in the memory encoding stage, generating interactive auxiliary information in the case of outputting the second training task in the extraction stage, and outputting the interactive auxiliary information. The dynamic difficulty adjustment module is used for updating the current difficulty parameter and transmitting the current difficulty parameter to the task scheduling module. The training precision can be improved by adopting the system.

Description

Technical Field

[0001] This application relates to the field of human-computer interaction technology, and in particular to an interactive training system based on retrospective memory. Background Technology

[0002] Retrospective memory, also known as memory of past events and information, is a mainstream theory in cognitive neuroscience that suggests that during the memory encoding stage, the brain has difficulty effectively binding the core content of information with contextual elements such as "time" and "place." This results in the inability to accurately access the target memory during retrieval due to a lack of effective contextual cues.

[0003] Related studies indicate that the brain does not store information in isolation, but rather connects and integrates new information with existing, structurally similar memory networks through the principle of "similarity." This process is a key mechanism for consolidating and enhancing memory.

[0004] However, in the fields of cognitive training and human-computer interaction, especially in applications involving information memory and retrieval, existing systems mainly focus on attention training and behavioral data correction, which have significant limitations. They lack targeted training programs that can integrate the theory of "contextual cue binding" and the principle of "similarity" memory. Traditional memory training is also mostly based on mechanical repetition, which cannot dynamically evaluate and assist the user's brain's encoding and retrieval process. In addition, most existing cognitive training systems rely solely on the user's behavioral data to adjust the difficulty, without being able to understand the underlying neurocognitive state, such as encoding strength, retrieval effort, and cognitive load when dealing with similarity interference, resulting in insufficient training accuracy.

[0005] Therefore, there is an urgent need for an interactive training system based on retrospective memory to help users establish a "similarity" link between new information and a stable, orderly contextual framework, so as to meet the needs of professional training or auxiliary scenarios that have higher requirements for memory ability. Summary of the Invention

[0006] Therefore, it is necessary to provide an interactive training system based on backtracking memory that can improve training accuracy in response to the above-mentioned technical problems.

[0007] In a first aspect, this application provides an interactive training system based on retrospective memory, the system comprising: a data acquisition module, a task scheduling module, an interaction module, and a dynamic difficulty adjustment module;

[0008] The acquisition module is used to acquire the first EEG signal generated by the user during the first training task in the memory encoding stage, and to acquire the second EEG signal and behavioral data generated by the user during the second training task in the extraction stage.

[0009] The task scheduling module is used to generate a first training task based on the current difficulty parameter during the memory encoding stage, and output the first training task. The first training task requires the user to bind the memory item with the corresponding scene for memorization. During the extraction stage, a second training task determined based on the first training task and the first EEG signal is output, and the second training task is determined based on the behavioral data collected by the acquisition module and the second EEG signal to determine whether to end the output of the second training task.

[0010] The interaction module is used to trigger attention reminders or contextual reinforcement strategies based on the first EEG signal during the memory encoding stage, and to generate interactive assistance information based on the behavioral data and the second EEG signal during the retrieval stage when the second training task is output, and to output the interactive assistance information.

[0011] The dynamic difficulty adjustment module is used to update the current difficulty parameter based on the behavioral data collected by the acquisition module in the extraction stage and the second EEG signal, and transmit the updated current difficulty parameter to the task scheduling module to adjust the difficulty parameter in the memory encoding stage.

[0012] In one embodiment, the task scheduling module is further configured to output each calibration task in a preset order before the memory encoding stage begins;

[0013] The acquisition module is also used to acquire third EEG signals of the user when performing each of the calibration tasks;

[0014] The system also includes:

[0015] The baseline calibration module is used to generate preset neural baseline thresholds based on the third EEG signal, and to generate an initial difficulty coefficient based on the third EEG signal and the neural baseline thresholds, and to send the initial difficulty coefficient to the task scheduling module.

[0016] The task scheduling module is also used to generate an initial training task for the memory encoding stage based on the initial difficulty coefficient, and output the initial training task.

[0017] In one embodiment, the task scheduling module is further configured to sequentially output resting state task, attention task, memory encoding task and memory retrieval task before the memory encoding stage begins;

[0018] The baseline calibration module is also used for:

[0019] The third EEG signal of the user during the resting state task was processed to obtain the median absolute power of the Theta band in the prefrontal lobe and the Alpha band in the parietal lobe as the resting state threshold.

[0020] The third EEG signal of the user during the attention task is processed to obtain the prefrontal Theta power and the parietal Alpha desynchronization ratio. The average value of the prefrontal Theta power and the median of the parietal Alpha desynchronization ratio are used as the attention threshold, wherein the parietal Alpha desynchronization ratio is the rate of decrease of Alpha power in the parietal region relative to the resting state threshold.

[0021] The third EEG signal of the user during the memory encoding task is processed to obtain the P300 waveform appearing at the Pz electrode point, and the average amplitude of the P300 waveform is calculated as the memory encoding threshold.

[0022] The third EEG signal of the user during the memory retrieval task is processed to obtain the power change of the Gamma band in the temporal lobe region within the target time window after stimulus presentation, and the memory retrieval threshold is obtained based on the power change of the Gamma band.

[0023] In one embodiment, the current difficulty parameter includes target scene similarity and target item overlap; the task scheduling module includes:

[0024] The parameter receiving submodule is used to receive the initial difficulty parameter sent by the baseline calibration module and the updated current difficulty parameter sent by the dynamic difficulty adjustment module.

[0025] The training task generation submodule is used to select a first scene and a second scene from the scene library. The scene similarity of the first scene and the second scene matches the similarity of the target scene. It generates memory item lists for the first scene and the second scene respectively, and the item overlap of the memory item lists corresponding to the first scene and the second scene matches the overlap of the target items. Based on the first scene, the second scene and the memory item lists corresponding to the first scene and the second scene, a first training task is generated.

[0026] In one embodiment, the task scheduling module further includes:

[0027] The training mode generation submodule is used to determine the task mode based on the similarity of the target scene;

[0028] The interaction module includes:

[0029] The attention detection submodule is used to generate the current prefrontal Theta power and the current parietal Alpha desynchronization ratio based on the first EEG signal, and to determine whether the user's attention meets the standard based on the current prefrontal Theta power, the current parietal Alpha desynchronization ratio and the attention threshold.

[0030] The first interaction submodule is used to trigger an attention reminder when it is determined that the user's attention is insufficient;

[0031] The memory coding detection submodule is used to generate the average amplitude of the P300 waveform at the current Pz electrode based on the first EEG signal, and to determine the user's memory coding level based on the average amplitude of the P300 waveform at the current Pz electrode and the memory coding threshold.

[0032] The second interaction submodule is used to trigger encoding enhancement interaction according to the task mode if it is determined that the user's memory encoding level is not up to standard.

[0033] In one embodiment, the second interaction submodule includes:

[0034] The first interaction unit is used to trigger a sensory enhancement mode when the task mode is the first similarity mode, and the sensory enhancement mode is used to identify the item in the corresponding scene.

[0035] The second interaction unit is used to trigger a differentiation compensation mechanism when the task mode is the second similarity mode; wherein the target scene similarity corresponding to the first similarity mode is lower than the target scene similarity corresponding to the second similarity mode, and the differentiation compensation mechanism is used to add a new identifier in the scene.

[0036] In one embodiment, the task scheduling module includes:

[0037] The judgment submodule is used to determine, during the extraction phase, whether the user can correctly recall based on the behavioral data.

[0038] The loop control submodule is used to end the output of the second training task when the user can successfully recall the training or the number of times the interactive assistance information is output reaches a preset value.

[0039] The interaction module is further configured to determine the corresponding neurocognitive state based on the second EEG signal and each of the neural baseline thresholds when the judgment submodule determines that the user cannot recall or answers incorrectly, determine interactive assistance information based on the neurocognitive state, and output the interactive assistance information.

[0040] In one embodiment, the dynamic difficulty adjustment module includes:

[0041] The evaluation submodule is used to obtain an evaluation score by weighting the behavioral data, the second EEG signal, and the interactive assistance information.

[0042] The difficulty adjustment submodule is used to determine the difficulty parameter adjustment method based on the evaluation score, the current improvement threshold, and the current consolidation threshold, and to adjust the current difficulty parameter based on the determined difficulty parameter adjustment method.

[0043] In one embodiment, the dynamic difficulty adjustment module further includes:

[0044] The threshold determination submodule is used to determine an initial boost threshold and an initial consolidation threshold based on the obtained N evaluation scores in the initial stage of the second training task output; it is also used to update the initial boost threshold and initial consolidation threshold according to the evaluation scores obtained in each training and the first N-1 evaluation scores to obtain new boost threshold and consolidation threshold.

[0045] In one embodiment, the system further includes:

[0046] The incentive module is used to output incentive information when all users are able to recall the success.

[0047] The aforementioned interactive training system based on retrospective memory includes a data acquisition module that collects first EEG signals generated by the user during the first training task in the memory encoding phase, and second EEG signals and behavioral data generated by the user during the second training task in the retrieval phase. A task scheduling module generates and outputs a first training task based on the current difficulty parameters during the memory encoding phase. The first training task requires the user to associate memory items with corresponding scenarios. During the retrieval phase, a second training task determined based on the first training task and the first EEG signals is output, and the system determines whether to end the second training task based on the behavioral data and second EEG signals collected by the acquisition module. An interaction module is used to collect the first EEG signals generated by the user during the memory encoding phase, and to collect the second EEG signals and behavioral data generated by the user during the second training task. An EEG signal triggers an attention reminder or contextual reinforcement strategy. During the extraction phase, while outputting the second training task, interactive auxiliary information is generated based on the behavioral data and the second EEG signal, and the interactive auxiliary information is output. A dynamic difficulty adjustment module is used to update the current difficulty parameter based on the behavioral data and the second EEG signal collected by the acquisition module during the extraction phase, and transmit the updated current difficulty parameter to the task scheduling module to adjust the difficulty parameter during the memory encoding phase. In this way, by dynamically quantifying the difficulty parameter and automatically adjusting it by integrating behavioral data and neural data, it is ensured that the training task always matches the user's optimal ability challenge zone, avoiding a one-size-fits-all training mode, greatly improving the user-specificity and long-term effectiveness of training, and improving training accuracy. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a block diagram of an interactive training system based on retrospective memory in one embodiment;

[0050] Figure 2 This is the hardware of an interactive training system based on retrospective memory in one embodiment;

[0051] Figure 3 This is a flowchart illustrating the initial difficulty coefficient determination step in one embodiment;

[0052] Figure 4 This is a flowchart illustrating the interactive steps of the memory encoding stage in one embodiment;

[0053] Figure 5 This is a flowchart illustrating the interactive steps of the extraction phase in one embodiment.

[0054] Figure 6 This is a flowchart illustrating the steps for adjusting the difficulty level in one embodiment;

[0055] Figure 7 This is a flowchart illustrating the threshold adjustment steps in one embodiment. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0057] In one exemplary embodiment, such as Figure 1 As shown, an interactive training system based on retrospective memory is provided. The system includes: a data acquisition module, a task scheduling module, an interaction module, and a dynamic difficulty adjustment module.

[0058] The acquisition module is used to acquire the first EEG signal generated by the user during the first training task in the memory encoding stage, and to acquire the second EEG signal and behavioral data generated by the user during the second training task in the extraction stage.

[0059] The task scheduling module is used to generate and output the first training task based on the current difficulty parameters during the memory encoding stage. The first training task requires the user to bind the memory items with the corresponding scenarios. During the retrieval stage, the module outputs the second training task determined based on the first training task and the first EEG signal. The module also determines whether to end the second training task based on the behavioral data collected by the acquisition module and the second EEG signal.

[0060] The interaction module is used to trigger attention reminders or contextual reinforcement strategies based on the first EEG signal during the memory encoding stage. During the retrieval stage, when outputting the second training task, it generates interactive auxiliary information based on behavioral data and the second EEG signal, and outputs the interactive auxiliary information.

[0061] The dynamic difficulty adjustment module is used to update the current difficulty parameters based on the behavioral data collected by the extraction phase acquisition module and the second EEG signal, and transmit the updated current difficulty parameters to the task scheduling module to adjust the difficulty parameters of the memory encoding phase.

[0062] This application mainly includes three stages: baseline calibration stage, memory encoding stage, and extraction stage.

[0063] In the baseline calibration phase, the initial difficulty coefficient of the task is determined based on the third EEG signal when the user performs each calibration task, so as to serve as the initial difficulty coefficient of the task in the memory encoding phase.

[0064] The memory encoding stage is mainly used to automatically generate and execute a specific training task, and to ensure the quality of the encoding through real-time EEG monitoring during the encoding process.

[0065] The retrieval phase begins after memory encoding is complete and a retention interval (ranging from a few minutes to a few days, which is extended as training progresses). Its core is to use situational cues to trigger retrieval and form a closed loop of "EEG interpretation - strategic cues".

[0066] In each stage, the acquisition module can collect EEG signals and behavioral data. Specifically, the acquisition module includes an EEG acquisition device and a behavioral data acquisition device, wherein the EEG acquisition device is used to collect the user's EEG signals, and the behavioral data acquisition device is used to collect the user's behavioral data. Specifically, in the memory encoding stage, the first EEG signal generated by the user during the processing of the first training task is collected; in the retrieval stage, the second EEG signal and behavioral data generated by the user during the processing of the second training task are collected. The baseline calibration module collects the third EEG signal of the user when performing each calibration task.

[0067] The task scheduling module maintains a structured scene library and item library, and automatically generates training tasks that meet the current difficulty requirements based on the two quantitative parameters of scene similarity (SSS) and item overlap (IOR) fed back by module four. It transforms abstract difficulty instructions into concrete training experiences.

[0068] The task scheduling module is used to generate the first training task based on the current difficulty coefficient during the memory encoding stage, and output the first training task, for example, by sending it to the user through the display unit / playback unit.

[0069] The interactive module ensures attention: triggering reminders when attention is lost. Encoding quality enhancement: triggering contextual reinforcement (low SSS) or differential compensation (high SSS) when P300 amplitude is insufficient. Extraction strategy assistance: providing differential cues or encouragement based on the Theta+Gamma model.

[0070] Furthermore, during the memory encoding phase, while the user performs the training task, the EEG acquisition device synchronously performs real-time EEG monitoring. The interaction module can assess the user's attention index and P300 amplitude based on the real-time monitored first EEG signal, and trigger different response mechanisms based on the assessment results, such as triggering attention reminders or situational reinforcement strategies. Specifically, triggering attention reminders is used to alert the user (through display units / playback units / vibration units, etc.), while situational reinforcement strategies are used to reinforce and adjust the first training task, presenting the corresponding target to be reinforced to the user through at least one of the display unit, playback unit, vibration unit, etc., thereby completing the user's training process and outputting the training process data (including first EEG data and behavioral data). The training process data is stored in the storage unit.

[0071] At a preset time interval, when the user forgets, the retrieval phase begins. The task scheduling module retrieves historical training process data and personalized baseline thresholds from the storage unit. Based on the retrieved historical training process data and personalized baseline thresholds, it analyzes the data to recreate the encoding scenario and determine the second training task. This task is then sent to the user through at least one of the display unit or playback unit, allowing the user to respond to the specified content of the second training task. During this period, the EEG acquisition device simultaneously performs real-time EEG monitoring to obtain the second EEG signal and acquires the user's response as behavioral data. The interaction module triggers different subsequent analysis processes based on the behavioral data and the second EEG signal. If the user provides correct behavioral feedback, the training data is recorded to the storage unit. If the user cannot provide correct behavioral feedback, the system analyzes the behavioral data and the second EEG signal, and outputs interactive assistance information. Based on the interactive assistance information, the system guides the user through at least one of the display unit and playback unit, and records the second EEG signal and behavioral data to feed back to the storage unit. After completing the real-time EEG closed loop of the current round, the dynamic difficulty adjustment module updates the current difficulty parameters based on the behavioral data and the second EEG signal collected by the extraction phase acquisition module. Specifically, it obtains a performance evaluation score based on the behavioral data and the second EEG signal collected by the extraction phase acquisition module. Then, the performance evaluation score is input to the task scheduling module, which outputs an adjustment strategy and adjusts the training task according to the adjustment strategy.

[0072] The dynamic difficulty adjustment module employs a dynamic dual-threshold (boost threshold / consolidation threshold) decision-making mechanism to automatically determine whether to increase, maintain, or decrease the difficulty parameter (SSS / IOR). This forms a crucial closed loop: it feeds back the determined new difficulty parameter to the task scheduling module, driving the generation of the next training task, thereby achieving continuous adaptation.

[0073] This application is a complete closed-loop system that starts with an individual's neural baseline, takes real-time EEG decoding and interaction as its core, uses dynamic difficulty parameters as an adaptive link, and uses gamified incentives as a sustainable driving force.

[0074] Among them, the combination Figure 2 As shown, the hardware components involved in this application include a display unit (e.g., a display screen), a playback unit (e.g., a speaker), an EEG acquisition device (e.g., an EEG cap), a computing unit, a control unit, and a storage unit. The computing unit includes a baseline calibration module, a task scheduling module, an interaction module, and a dynamic difficulty adjustment module. The acquisition module includes the EEG acquisition device. The display unit and playback unit are used to output training tasks, the storage unit is used to store data involved in each stage, and the control unit is used to control the display unit, playback unit, and storage unit.

[0075] In some optional embodiments, the task scheduling module is further configured to output each calibration task in a preset order before the start of the memory encoding stage; the acquisition module is further configured to acquire the third EEG signal of the user when performing each calibration task; the system also includes: a baseline calibration module, configured to generate preset neural baseline thresholds based on the third EEG signal, and generate an initial difficulty coefficient based on the third EEG signal and the neural baseline thresholds, and send the initial difficulty coefficient to the task scheduling module; the task scheduling module is further configured to generate the initial training task of the memory encoding stage based on the initial difficulty coefficient, and output the initial training task.

[0076] When a user wears an EEG acquisition device to perform various baseline calibration tasks, the display unit presents visual stimuli to the user, and the playback unit plays specified audio (such as white noise) to the user. The EEG acquisition device simultaneously acquires the third EEG signal generated by the user when performing each baseline calibration task and transmits it to the storage unit so that the storage unit can store the acquired third EEG signal. Afterward, the baseline calibration module obtains the third EEG signal from the storage unit, performs calculations on the third EEG signal, obtains the user's neural baseline threshold, and stores the neural baseline threshold in the storage unit.

[0077] The baseline calibration module establishes a unique neural function baseline for each user, including resting-state power threshold, attention threshold (prefrontal Theta / parietal Alpha), memory encoding threshold (Pz point P300 amplitude), and memory retrieval threshold (temporal lobe Gamma power). Its output personalized thresholds serve as the "benchmark" for all real-time judgments made by the interaction module, and are fundamental to the system's personalization.

[0078] After the user completes the calibration task in step one, the system will calculate the initial difficulty coefficient, namely the SSS and IOR values, based on the user's performance in the attention task and memory encoding task.

[0079] Combination Figure 3 As shown, the calculation of the initial scene similarity (SSS_initial) includes:

[0080] Data source: Behavioral data from the attention baseline calibration task.

[0081] Key metrics: Stability and control of attention. The standard deviation of reaction time (RT_Std) and the omission error rate are used as surrogate metrics. Large fluctuations in reaction time and numerous omissions indicate unstable attentional control.

[0082] Gear initialization (implementation example):

[0083] Based on a large amount of normative data, the system presets several SSS levels:

[0084] Level 1 (Low SSS): If the user's RT_Std and Omission Errors are both higher than the preset threshold, it is determined that their attention control ability is weak, and the initial SSS is set to the lowest level (e.g., 0.2, representing a huge difference in the scene).

[0085] Level 2 (Medium SSS): If one or all of a user's metrics are at a medium level, the initial SSS is set to a medium level (e.g., 0.4).

[0086] Tier 3 (Higher SSS): If the user's RT_Std and Omission Errors are both excellent (below a certain threshold), the initial SSS can be set to a relatively challenging tier (e.g., 0.55).

[0087] The calculation of the initial item overlap (IOR_initial) includes:

[0088] Data source: Behavioral data from the memory-encoded baseline calibration task.

[0089] Core metric: Basic efficiency of memory encoding and retrieval. Recognition accuracy is used as a surrogate metric.

[0090] Gear initialization (implementation example):

[0091] Based on a large amount of normative data, the system presets several IOR levels, including:

[0092] Level 1 (Zero Overlap): If the accuracy rate of recognition is low (e.g., <70%), the initial IOR is set to 0, meaning that the list of items to be memorized is completely different in different scenarios, avoiding any interference.

[0093] Level 2 (Low Overlap): If the recognition accuracy is moderate (e.g., 70%-85%), then the initial IOR is set to a lower value (e.g., 0.2).

[0094] Level 3 (Medium Overlap): If the recognition accuracy is very high (e.g., >85%), the initial IOR can be set to a medium level (e.g., 0.4) to introduce preliminary similarity interference.

[0095] In the above embodiments, through this initialization strategy, the system achieves a certain degree of personalized difficulty preset before the user's first formal training. For users with weaker abilities, the system starts with low SSS and zero IOR, providing a very friendly and supportive starting point, focusing on building confidence and a successful experience; for users with stronger abilities, the system starts with medium SSS and low to medium IOR, quickly entering challenging training and improving training efficiency. This method ensures that the system is immediately usable and effective for users with different skill levels. After the first training, the system will immediately enter the previously detailed adaptive loop based on the comprehensive performance evaluation score and dynamic threshold, quickly optimizing the values ​​of SSS and IOR to rapidly approach the user's true ability level.

[0096] In some optional embodiments, the task scheduling module is also used to sequentially output the resting state task, attention task, memory encoding task, and memory retrieval task before the memory encoding phase begins.

[0097] The baseline calibration module is also used to: process the user's third EEG signal during a resting-state task, obtaining the median absolute power of the Theta band in the prefrontal cortex and the Alpha band in the parietal cortex as the resting-state threshold; process the user's third EEG signal during an attention task, obtaining the Theta power in the prefrontal cortex and the Alpha desynchronization ratio in the parietal cortex, and using the average value of the Theta power in the prefrontal cortex and the median of the Alpha desynchronization ratio in the parietal cortex as the attention threshold, where the Alpha desynchronization ratio in the parietal cortex is the rate of decrease of the Alpha power in the parietal cortex relative to the resting-state threshold; process the user's third EEG signal during a memory encoding task, obtaining the P300 waveform appearing at the Pz electrode point, and calculating the average amplitude of the P300 waveform as the memory encoding threshold; and process the user's third EEG signal during a memory retrieval task, obtaining the power change of the Gamma band in the temporal lobe region within the target time window after stimulus presentation, and obtaining the memory retrieval threshold based on the power change of the Gamma band.

[0098] Specifically, calibration tasks include resting state tasks, attention tasks, memory encoding tasks, and memory retrieval tasks, which will be described in detail below.

[0099] The resting-state task requires the user to relax and watch a neutral, dynamic visual stimulus (such as slowly floating clouds or a smoothly rotating geometric object) for a predetermined duration. Simultaneously, soothing white noise is played through headphones.

[0100] The third EEG signal collected for resting-state tasks is the user's raw EEG signal. This third EEG signal is used to calculate the median absolute power in the Theta (4-8 Hz) band of the prefrontal cortex and the Alpha (8-12 Hz) band of the parietal cortex. The data obtained at this stage serves as the resting-state power baseline. In subsequent training tasks, the brain's activity power will be compared to this baseline. For example, task-related desynchronization of the parietal alpha wave is a key indicator, calculated as (resting alpha power - task-based alpha power) / resting alpha power. A higher ratio indicates a higher level of cognitive engagement.

[0101] Attention tasks present users with a simplified, sequential task. For example, different colored spaceships may appear randomly and rapidly on the screen. Users only need to press a button when a green spaceship (the target stimulus) appears, and do not need to react when other colored spaceships appear.

[0102] The third EEG signal required for attention tasks is acquired by analyzing the EEG signal within a 500-800 millisecond time window following the appearance of the target stimulus. The acquired behavioral data is used to calculate the average reaction time and accuracy of correct responses. This third EEG signal is used to calculate: Prefrontal Theta Power: The average power in the Theta band of the prefrontal cortex electrodes is calculated. This metric is highly correlated with maintaining attention and working memory load. Parietal Alpha Desynchronization: The rate of decrease in alpha power relative to the resting-state baseline during the task is calculated at the parietal cortex electrodes.

[0103] The method for establishing this threshold includes: selecting all trials with correct and rapid responses (e.g., the top 60% of fast responses); calculating the average prefrontal Theta power and the median parietal Alpha desynchronization rate in these "high-attention performance" trials. These two values ​​will be set as the user's initial personal threshold for "focused attention," i.e., the attention threshold. Furthermore, during formal memory encoding training, the system will monitor these two indicators in real time. If either consistently falls below the attention threshold, the system determines the user is in a "distracted" state and triggers intervention (e.g., gentle reminders).

[0104] The memory encoding task combines the "Oddball" paradigm with image memory. Users are presented with a series of images, mostly ordinary objects (such as chairs and books, accounting for 70%, "standard stimuli"), but some novel and interesting images (such as dinosaurs and castles, accounting for 30%, "alternative stimuli") are randomly interspersed. Users are asked to memorize these images and are informed that a memory test will follow.

[0105] The third type of EEG signal includes event-related potentials (ERPs) triggered by biased stimuli. Special attention is paid to the P300 component appearing at the Pz electrode within a time window of 250-500 milliseconds after stimulus presentation. The amplitude of P300 (the difference between peak voltage and baseline, in μV) is widely considered an effective electrophysiological indicator of working memory updates and the strength of episodic memory encoding.

[0106] Determining the memory encoding threshold involves: averaging the EEG signals triggered by all biased stimuli to extract a clear P300 waveform. The average amplitude of this P300 waveform is calculated, and this value is set as the user's personal reference threshold for "effective memory encoding." During the contextual encoding phase of formal training, when the system presents a key item to the user, it analyzes the P300 amplitude induced by the Pz point in real time. If this amplitude is significantly lower than (e.g., below 80%) the memory encoding threshold, the system determines that the item is "weakly encoded" and immediately triggers contextual reinforcement mechanisms (such as highlighting, amplifying, or adding special sound effects to the item in a virtual scene) to perform neural-level "reinforcement" encoding.

[0107] The memory retrieval task is combined with "memory encoding baseline calibration." After the user views and memorizes a series of images, a brief fill-in-the-blank task (such as a simple math problem to eliminate the influence of short-term memory) is performed, followed by a recognition memory test. In the test, "old" images (just seen) and "new" images (unseen) are presented in a mixed format, and the user is asked to judge whether the images have appeared before.

[0108] The extraction methods for the third EEG signal include: timing: recording the third EEG signal begins when the system presents an "old" image (i.e., the target to be successfully extracted) to the user; analysis of brain regions and indicators: focusing on the power changes in the Gamma band (30-45Hz) of the temporal lobe region (such as T5 and T6 electrode points) within a 300-500 ms time window after stimulus presentation. Scientific basis: when an old memory is successfully identified and retrieved, the brain (especially the temporal lobe) exhibits significant Gamma oscillation activity, which is considered an electrophysiological marker of successful activation of the memory trace.

[0109] The memory retrieval threshold is determined by filtering out the EEG segments corresponding to all correctly recognized "old" images. The average or median temporal lobe Gamma power is calculated for these segments. This value is set as the user's personal reference threshold for "successful activation of memory traces" and serves as the memory retrieval threshold. During the formal training memory retrieval phase, when the system presents only situational cues, it monitors the user's temporal lobe Gamma power in real time. If the user exhibits "high frontal lobe Theta" (active searching) but "low temporal lobe Gamma" (Gamma power below this personal baseline), the system can clearly determine that they are in a state of "insufficient activation of memory traces," thereby accurately triggering "differential cues."

[0110] Through the above four progressively ordered calibration tasks, which are preferably completed in one go, the system creates a unique "neurocognitive profile" for each user, making subsequent accurate and adaptive interaction possible. This is one of the core advantages of this invention compared to the traditional "one-size-fits-all" training scheme.

[0111] In some optional embodiments, the current difficulty parameters include target scene similarity and target item overlap; the task scheduling module includes:

[0112] The parameter receiving submodule is used to receive the initial difficulty parameters sent by the baseline calibration module and the updated current difficulty parameters sent by the dynamic difficulty adjustment module.

[0113] The training task generation submodule is used to select a first scene and a second scene from the scene library. The scene similarity of the first scene and the second scene matches the similarity of the target scene. Memory item lists are generated for the first scene and the second scene respectively. The item overlap of the memory item lists corresponding to the first scene and the second scene matches the overlap of the target items. The first training task is generated based on the first scene, the second scene and the memory item lists corresponding to the first scene and the second scene.

[0114] The task scheduling module automatically generates and executes a specific training task based on the latest difficulty parameters output by the baseline calibration module or the dynamic difficulty adjustment module, namely Scene Similarity Score (SSS) and Item Overlap Ratio (IOR) (these two difficulty parameters belong to two different dimensions and are not completely related). During the encoding process, real-time EEG monitoring is used to ensure the encoding quality.

[0115] Specifically, the task scheduling module intelligently matches tasks from the structured scene library and the memory item library based on the SSS and IOR values.

[0116] Specifically, the parameter receiving submodule obtains the preset difficulty parameters for the current training cycle from the baseline calibration module or the dynamic difficulty adjustment module, including: target scene similarity (SSS): a specific value between 0 and 1; target scene overlap (IOR): a specific value between 0 and 1.

[0117] The training task generation submodule performs the following operations based on the received SSS and IOR values:

[0118] Scene matching: Automatically select a pair of scenes (including the first scene and the second scene) from the scene library, and match the similarity (SSS) of the scene pair with the target SSS.

[0119] Item list generation: Generate memory item lists for the two paired scenarios, and ensure that the actual overlap ratio of items between the two lists is consistent with the target IOR.

[0120] Training mode definition: Based on SSS, the system internally defines the task mode and further adjusts the task difficulty under the task mode based on IOR. For example: SSS ≤ 0.5, the system is in "low similarity training" mode, which aims to establish a solid memory foundation; SSS > 0.5, the system is in "high similarity training" mode, which aims to challenge the memory's discrimination and anti-interference ability.

[0121] In some optional embodiments, the task scheduling module further includes a training mode generation submodule, used to determine the task mode based on the similarity of the target scene.

[0122] The interactive modules include:

[0123] The attention detection submodule is used to generate the current prefrontal cortex Theta power and the current parietal cortex Alpha desynchronization ratio based on the first EEG signal, and to determine whether the user's attention meets the standard based on the current prefrontal cortex Theta power, the current parietal cortex Alpha desynchronization ratio and the attention threshold.

[0124] The first interaction submodule is used to trigger an attention reminder when the user's attention level is deemed insufficient.

[0125] In this application, all trials with correct and rapid responses (e.g., the top 60% of fast responses) were selected. The average prefrontal Theta power and the median parietal Alpha desynchronization rate were calculated for these "high-attention performance" trials. These two values ​​will be set as the user's initial personal threshold for "focused attention," i.e., the attention threshold. Furthermore, during formal memory encoding training, the system will monitor these two indicators in real time. If either consistently falls below the attention threshold, the system determines the user is in a "distracted" state and triggers intervention (e.g., gentle reminders).

[0126] The memory coding detection submodule is used to generate the average amplitude of the P300 waveform at the current Pz electrode based on the first EEG signal, and to determine the user's memory coding level based on the average amplitude of the P300 waveform at the current Pz electrode and the memory coding threshold.

[0127] The second interaction submodule is used to trigger coding reinforcement interaction based on the task mode when it is determined that the user's memory coding level is not up to standard.

[0128] During the contextual encoding phase of formal training, when the system presents a user with an item that needs to be memorized, it analyzes the P300 amplitude induced by the Pz point in real time. If the amplitude is significantly lower than (for example, below 80%) the memory encoding threshold, the system determines that the item is "weakly encoded" and immediately triggers a contextual reinforcement mechanism (such as highlighting, enlarging, or adding special sound effects to the item in a virtual scene) to perform neural-level "reinforcement" encoding, that is, encoding reinforcement interaction.

[0129] In some optional embodiments, the second interaction submodule includes:

[0130] The first interaction unit is used to trigger the sensory enhancement mode when the task mode is the first similarity mode. The sensory enhancement mode is used to identify the item in the corresponding scene.

[0131] The second interaction unit is used to trigger a differentiation compensation mechanism when the task mode is the second similarity mode; wherein the similarity of the target scene corresponding to the first similarity mode is lower than the similarity of the target scene corresponding to the second similarity mode, and the differentiation compensation mechanism is used to add a new identifier in the scene.

[0132] During task execution, the system performs real-time EEG monitoring. Its interaction logic is directly related to the high / low similarity characteristics of the current task, combined with... Figure 4 As shown, when the system determines that a project has "weak coding", it will trigger two different reinforcement strategies based on the pattern defined by the current task's SSS value:

[0133] a) If it is a low similarity pattern:

[0134] Intervention strategy: Trigger "sensory enhancement"; Specific operations: Highlight, enlarge, flash the item in the virtual scene, or give it a unique sound effect compatible with the scene. For example, in the "underwater world", make the "treasure chest" that needs to be remembered emit a flashing golden light; accompany the appearance of the "apple" with a crisp chewing sound; or repeat the item with an exaggerated tone, such as "Remember this big red apple!"

[0135] b) If it is a highly similar pattern:

[0136] Intervention Strategy: Trigger "Differentiation Compensation." Specifically, the system automatically adds a temporary, unique marker to the current scene, a marker that doesn't exist in the paired scene. For example, in the "night kitchen," a "steaming" animation effect is added to the weakly coded "kettle," an effect absent in the similar "afternoon kitchen." This essentially dynamically and locally reduces the SSS of the current task to assist the user in completing accurate coding.

[0137] Through the above technical steps, this application ensures that every memory encoding attempt is subject to rigorous neural quality monitoring during training of any difficulty, and maximizes the strength and accuracy of memory traces through adaptive intervention, laying a solid foundation for subsequent successful retrieval.

[0138] In some optional embodiments, the task scheduling module includes:

[0139] The judgment submodule is used during the extraction phase to determine whether the user can recall correctly based on behavioral data.

[0140] The loop control submodule is used to end the second training task when the number of times the user can recall the success or the number of times the interactive auxiliary information is output reaches a preset value.

[0141] The interaction module is also used to determine the corresponding neurocognitive state based on the second EEG signal and the baseline thresholds of each nerve when the judgment submodule determines that the user cannot recall or answers incorrectly, and to determine and output the interactive assistance information based on the neurocognitive state.

[0142] This step is initiated after memory encoding is complete and a retention interval (from a few minutes to a few days, which is extended as training progresses). Its core is to use situational cues to trigger retrieval and form a closed loop of "EEG interpretation - strategic cues".

[0143] In this application, the system does not directly ask "What words do you remember?", but instead issues the second training task by reproducing the encoding context:

[0144] a) Contextual cue presentation: The system will only present the "contextual scene" from which the memory was previously bound to the user. For example, it may display a static image or a short animation of a "red kitchen" and play corresponding ambient sounds (such as the humming sound of a kitchen). No specific memory item will be presented.

[0145] b) Extracting Instructions: The system will provide explicit instructions, which closely correspond to the binding logic in the coding phase. For example:

[0146] i. "What items do we store in this kitchen? Please list them one by one."

[0147] ii. "Return to this Rainbow Park and find all the treasures we hid here yesterday."

[0148] c) Response mode: Users can recall items by speaking, tapping or dragging images of alternative items on the touchscreen.

[0149] Combination Figure 5 As shown, the judgment submodule can determine whether the user can recall correctly based on behavioral data: when the user cannot recall correctly, the system no longer passively waits or randomly gives prompts, but instead initiates a precise intervention strategy based on the user's EEG pattern.

[0150]

[0151] Specifically, when the user makes a new round of responses based on the system prompts (such as saying the answer or selecting an item), the process will return to the judgment submodule, and the above loop will be entered again based on the new behavioral response, until the user successfully recalls the information or reaches the preset prompt limit (preset value).

[0152] All interaction data (including EEG patterns, cue types, and final results) is recorded to update the user's "similarity interference" profile and dynamically adjust the difficulty of subsequent training. Through this refined technical process, this application not only tests memory but also teaches children how to think and how to overcome memory bottlenecks. It internalizes external, system-provided cues into memory retrieval strategies that users can spontaneously use in the future, thereby achieving a fundamental transfer and improvement of abilities.

[0153] In some optional embodiments, the dynamic difficulty adjustment module includes:

[0154] The assessment submodule is used to obtain an assessment score by weighting behavioral data, second EEG signals, and interactive assistance information.

[0155] The difficulty adjustment submodule is used to determine the difficulty parameter adjustment method based on the evaluation score, the current improvement threshold, and the current consolidation threshold, and to adjust the current difficulty parameter based on the determined difficulty parameter adjustment method.

[0156] The purpose of this embodiment is to dynamically adjust the parameters of subsequent training tasks based on the user's overall performance during the retrieval phase, ensuring that training remains within the "optimal challenge" range. The core dimensions of adjustment include scene similarity, overlap of memory items, and the presentation of gamified incentives, combined with... Figure 6 As shown.

[0157] 1) Definition and quantification of scene similarity

[0158] A feature-label-based similarity calculation model is employed, rather than subjective judgment. The system incorporates a "structured scene library," where each scene is defined by a series of multi-dimensional feature labels:

[0159] a) Visual features: scene category (e.g., kitchen, park), main color, iconic object (refrigerator, swing), time (day, night), weather (sunny, rainy).

[0160] b) Auditory characteristics: ambient sounds (cooking sounds, birdsong, wind sounds).

[0161] c) Semantic features: the functionality of the scene (cooking, entertainment) and the emotional tone (lively, quiet).

[0162] Similarity calculation includes:

[0163] a) Define scene pairs: For any two scenes (Scene A, Scene B).

[0164] b) Calculate feature overlap: Calculate how well they match across all feature labels. For example, "afternoon kitchen" and "night kitchen" match perfectly in "scene category" and "main object", are opposite in "time", and are unrelated in "weather".

[0165] c) Generate a similarity score: Calculate a comprehensive similarity score (Scene Similarity Score, SSS) between 0 and 1 using a preset weighted algorithm. The higher the SSS, the more similar the scenes are, and the more difficult it is to distinguish them.

[0166] 2) Setting the overlap of memory items

[0167] Item Overlap Ratio (IOR) refers to the proportion of memory items that appear together in two (or more) similar scenarios.

[0168] a) Calculation formula: IOR = (Number of items in both scenarios) / (Length of item list in a single scenario).

[0169] The system generates a comprehensive evaluation score based on the user's performance during the extraction phase, and uses this score to determine the direction of subsequent training difficulty. For example, the difficulty adjustment submodule determines the adjustment method for difficulty parameters based on the evaluation score, the current improvement threshold, and the current consolidation threshold, and adjusts the current difficulty parameters based on the determined adjustment method.

[0170] The performance evaluation score is calculated by weighting behavioral data (such as accuracy and reaction time), EEG efficiency indicators (determined based on the second EEG signal, such as the peak Gamma power at successful retrieval and the required retrieval effort Theta level), and the required prompt level (determined based on interactive assistance information; high score for no prompt, medium score for requiring differentiated prompts, and low score for requiring direct answer). This avoids misjudgments caused by relying solely on accuracy.

[0171] For example, the assessment score (S) = W1 × behavioral score + W2 × EEG efficiency score + W3 × cue level score.

[0172] The scoring system includes: Behavioral Score: Calculated based on accuracy and reaction time, e.g., accuracy × (1 - standardized reaction time). EEG Efficiency Score: Quantifies the efficiency of neural resource utilization. For example, in successful retrieval trials, it's the ratio of the average temporal lobe Gamma power (reflecting activation efficiency) to the average frontal lobe Theta power (reflecting effort). A higher ratio indicates greater efficiency and less effort. Cue Level Score: Cues are categorized (e.g., no cue required = 1 point, need for differentiated cue = 0.6 points, need for direct instruction = 0.3 points), and the average score is calculated. W1, W2, and W3 are preset weighting coefficients used to balance the importance of different indicators.

[0173] Difficulty adjustments include: when performance evaluation scores remain low in high similarity mode, the system decides to reduce the difficulty, and its final strategy may be to switch back to "low similarity training" mode. Conversely, when performance evaluation scores remain excellent in low similarity mode, the system decides to increase the difficulty, the core of which is to put it into "high similarity training" mode.

[0174] In some optional embodiments, the dynamic difficulty adjustment module further includes: a threshold determination submodule, configured to determine an initial boost threshold and an initial consolidation threshold based on the obtained N evaluation scores in the initial stage of the second training task output; and to update the initial boost threshold and initial consolidation threshold according to the evaluation scores obtained in each training session and the first N-1 evaluation scores to obtain new boost threshold and consolidation threshold.

[0175] The threshold determination submodule uses a "rolling time window" statistical method to establish and update two key thresholds: the "raising threshold" and the "consolidation threshold".

[0176] The initial phase includes: after the user completes the first 5-8 training sessions, the system calculates the average (M0) and standard deviation (SD0) of the overall performance evaluation scores from these training sessions. The initial improvement threshold (T_up) is set to M0 + K * SD0 (e.g., K=0.5). This means that the difficulty needs to be increased only if the user performs significantly better than their initial average level. The initial consolidation threshold (T_consolidate) is set to M0 - K * SD0. This means that if the user's performance falls below their initial average level, the difficulty needs to be reduced.

[0177] Combination Figure 7 As shown, during the runtime phase (dynamic update): the threshold determination submodule always maintains a window, such as the user's most recent N training data (e.g., N=10). After each training session, the system adds the current comprehensive score S_current to the window, removes the earliest data, and recalculates the average (M_rolling) and standard deviation (SD_rolling) of all comprehensive scores within this window. The dynamic increase threshold update is: T_up = M_rolling + K * SD_rolling. The dynamic consolidation threshold update is: T_consolidate = M_rolling - K * SD_rolling.

[0178] The advantages of this threshold determination method are: Complete personalization: The threshold is based on the user's historical performance, a comparison with themselves. Dynamic adaptation: As the user's ability improves, their average level M_rolling gradually increases, and the threshold rises accordingly, continuously providing just the right challenge. Strong robustness: Using statistical standard deviation, the threshold can adapt to normal fluctuations in the user's state, avoiding frequent difficulty adjustments due to accidental fluctuations in single performances. Alignment with system goals: It ensures that difficulty adjustments are always based on the user's recent, stable performance trends, making the "dynamic difficulty adjustment engine" a truly intelligent and adaptive core module.

[0179] In some optional embodiments, the system further includes an incentive module for outputting incentive information when all users are able to recall the success.

[0180] The system outputs excitation information to the user through at least one of the following: display unit, playback unit, vibration unit, etc.

[0181] The incentive module is the system's "power system." It transforms abstract cognitive progress (such as efficient neural activity) into concrete, immediate game rewards (such as the purity and quantity of "memory crystals"), and embeds a long-term growth narrative (such as building a memory planet). It runs throughout the process, providing continuous motivation for the training execution of task scheduling, and links with positive evaluations from the interaction module to positively enhance the user experience.

[0182] In this application, the incentive information is delivered through gamification, which aims to transform extrinsic motivation into intrinsic motivation. The core design is a long-term narrative of a "memory explorer." The following is merely an example:

[0183] a) Core asset – “Memory Crystal”:

[0184] i. Each time a user successfully extracts a memory crystal (especially when no prompt or only a slight prompt is required), the user will receive a "memory crystal".

[0185] ii. The quantity and purity of the crystals (assessed based on EEG efficiency indicators and behavioral performance at the time of extraction) vary dynamically. More efficient extraction produces "purer" crystals.

[0186] b) Growth System – “Building a Memory Planet”

[0187] i. Users own their own planet. The "memory crystals" they acquire are the resources for building their planet.

[0188] ii. Crystals obtained from low-similarity tasks are used to build the basic topography of a planet (such as plains and mountains).

[0189] iii. High-purity crystals obtained from high-similarity missions can be used to unlock and build exquisite and unique buildings or decorations on the planet (such as crystal towers and rainbow bridges).

[0190] c) Instant Feedback and Achievements

[0191] i. Process Rewards: During the retrieval process, a creative "memory wizard" will provide encouragement (such as "You are very focused!") or strategic suggestions (such as "Don't rush, think slowly!") based on the user's brainwave state.

[0192] ii. Achievement Badges: Unlock milestone achievements such as "Similarity Breaker" (first time completing a high-overlap task) and "Focus Master" (meeting attention metrics N times consecutively).

[0193] In the above embodiments, the quantification and automation of difficulty adjustment are based on scientific decisions made using multi-dimensional data. Personalization of the training process ensures that each user progresses on a challenge path suited to their individual needs. Systematic motivation maintenance, achieved through gamification, transforms tedious training into an engaging exploration, thereby guaranteeing long-term adherence and ultimate effectiveness of the intervention.

[0194] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0195] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0196] It should be fully understood that the user information involved in this application (including but not limited to user physiological information, user personal information, etc.) is information and data authorized by the user or fully authorized by all parties. The use of user information shall comply with privacy policies and practices that are often considered to meet or exceed industry or government requirements for maintaining user privacy. The collection, use and processing of related data shall comply with relevant laws, regulations and standards, and provide corresponding operation access points for users to choose to authorize or refuse.

[0197] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0198] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An interactive training system based on retrospective memory, characterized in that, The system includes: a data acquisition module, a task scheduling module, an interaction module, and a dynamic difficulty adjustment module; The acquisition module is used to acquire the first EEG signal generated by the user during the first training task in the memory encoding stage, and to acquire the second EEG signal and behavioral data generated by the user during the second training task in the extraction stage. The task scheduling module is used to generate a first training task based on the current difficulty parameter during the memory encoding stage, and output the first training task. The first training task requires the user to bind the memory item with the corresponding scene for memorization. During the extraction stage, a second training task determined based on the first training task and the first EEG signal is output, and the second training task is determined based on the behavioral data collected by the acquisition module and the second EEG signal to determine whether to end the output of the second training task. The interaction module is used to trigger attention reminders or contextual reinforcement strategies based on the first EEG signal during the memory encoding stage, and to generate interactive assistance information based on the behavioral data and the second EEG signal during the retrieval stage when the second training task is output, and to output the interactive assistance information. The dynamic difficulty adjustment module is used to update the current difficulty parameter based on the behavioral data collected by the acquisition module in the extraction stage and the second EEG signal, and transmit the updated current difficulty parameter to the task scheduling module to adjust the difficulty parameter in the memory encoding stage. The task scheduling module is also used to output each calibration task in a preset order before the memory encoding stage begins. The acquisition module is also used to acquire third EEG signals of the user when performing each of the calibration tasks; The system also includes: The baseline calibration module is used to generate preset neural baseline thresholds based on the third EEG signal, and to generate an initial difficulty coefficient based on the third EEG signal and the neural baseline thresholds, and to send the initial difficulty coefficient to the task scheduling module. The task scheduling module is also used to generate an initial training task for the memory encoding stage based on the initial difficulty coefficient, and output the initial training task. The current difficulty parameters include target scene similarity and target item overlap; the task scheduling module includes: The parameter receiving submodule is used to receive the initial difficulty parameter sent by the baseline calibration module and the updated current difficulty parameter sent by the dynamic difficulty adjustment module. The training task generation submodule is used to select a first scene and a second scene from the scene library. The scene similarity of the first scene and the second scene matches the similarity of the target scene. It generates memory item lists for the first scene and the second scene respectively, and the item overlap of the memory item lists corresponding to the first scene and the second scene matches the overlap of the target items. Based on the first scene, the second scene and the memory item lists corresponding to the first scene and the second scene, a first training task is generated.

2. The system according to claim 1, characterized in that, The task scheduling module is also used to output resting state task, attention task, memory encoding task and memory retrieval task in sequence before the memory encoding stage begins. The baseline calibration module is also used for: The third EEG signal of the user during the resting state task was processed to obtain the median absolute power of the Theta band in the prefrontal lobe and the Alpha band in the parietal lobe as the resting state threshold. The third EEG signal of the user during the attention task is processed to obtain the prefrontal Theta power and the parietal Alpha desynchronization ratio. The average value of the prefrontal Theta power and the median of the parietal Alpha desynchronization ratio are used as the attention threshold, wherein the parietal Alpha desynchronization ratio is the rate of decrease of Alpha power in the parietal region relative to the resting state threshold. The third EEG signal of the user during the memory encoding task is processed to obtain the P300 waveform appearing at the Pz electrode point, and the average amplitude of the P300 waveform is calculated as the memory encoding threshold. The third EEG signal of the user during the memory retrieval task is processed to obtain the power change of the Gamma band in the temporal lobe region within the target time window after stimulus presentation, and the memory retrieval threshold is obtained based on the power change of the Gamma band.

3. The system according to claim 2, characterized in that, The task scheduling module also includes: The training mode generation submodule is used to determine the task mode based on the similarity of the target scene; The interaction module includes: The attention detection submodule is used to generate the current prefrontal Theta power and the current parietal Alpha desynchronization ratio based on the first EEG signal, and to determine whether the user's attention meets the standard based on the current prefrontal Theta power, the current parietal Alpha desynchronization ratio and the attention threshold. The first interaction submodule is used to trigger an attention reminder when it is determined that the user's attention is insufficient; The memory coding detection submodule is used to generate the average amplitude of the P300 waveform at the current Pz electrode based on the first EEG signal, and to determine the user's memory coding level based on the average amplitude of the P300 waveform at the current Pz electrode and the memory coding threshold. The second interaction submodule is used to trigger encoding enhancement interaction according to the task mode if it is determined that the user's memory encoding level is not up to standard.

4. The system according to claim 3, characterized in that, The second interactive submodule includes: The first interaction unit is used to trigger a sensory enhancement mode when the task mode is the first similarity mode, and the sensory enhancement mode is used to identify the item in the corresponding scene. The second interaction unit is used to trigger a differentiation compensation mechanism when the task mode is the second similarity mode; wherein the target scene similarity corresponding to the first similarity mode is lower than the target scene similarity corresponding to the second similarity mode, and the differentiation compensation mechanism is used to add a new identifier in the scene.

5. The system according to any one of claims 1 to 4, characterized in that, The task scheduling module includes: The judgment submodule is used to determine, during the extraction phase, whether the user can correctly recall based on the behavioral data. The loop control submodule is used to end the output of the second training task when the user can successfully recall the training or the number of times the interactive assistance information is output reaches a preset value. The interaction module is also used to determine the corresponding neurocognitive state based on the second EEG signal and each of the neural baseline thresholds when the judgment submodule determines that the user cannot recall or answers incorrectly, determine interactive assistance information based on the neurocognitive state, and output the interactive assistance information.

6. The system according to claim 5, characterized in that, The dynamic difficulty adjustment module includes: The evaluation submodule is used to obtain an evaluation score by weighting the behavioral data, the second EEG signal, and the interactive assistance information. The difficulty adjustment submodule is used to determine the difficulty parameter adjustment method based on the evaluation score, the current improvement threshold, and the current consolidation threshold, and to adjust the current difficulty parameter based on the determined difficulty parameter adjustment method.

7. The system according to claim 6, characterized in that, The dynamic difficulty adjustment module also includes: The threshold determination submodule is used to determine an initial boost threshold and an initial consolidation threshold based on the obtained N evaluation scores in the initial stage of the second training task output; it is also used to update the initial boost threshold and initial consolidation threshold according to the evaluation scores obtained in each training and the first N-1 evaluation scores to obtain new boost threshold and consolidation threshold.

8. The system according to claim 5, characterized in that, The system also includes: The incentive module is used to output incentive information when all users are able to recall the success.

Images