A motor imagery training system based on body affiliation

Abstract

This invention provides a motor imagery training system based on body ownership, comprising: a parameter generation module that extracts baseline individual alpha frequencies and generates temporal regulation parameters; a body ownership construction module that, based on the temporal regulation parameters, induces the construction of body ownership by temporally pairing virtual limb movements with multimodal feedback; a reference template construction module that extracts feedback-locked EEG features to establish a stable reference template for ownership; a multimodal feedback module that decodes motor imagery intentions to drive virtual limbs and couples out multimodal feedback, while simultaneously collecting the individual alpha frequencies and feedback-locked EEG features in the current task state; and an adaptive regulation module that calculates the deviation between the above two features and updates the temporal regulation parameters or reconstructs body ownership and the reference template based on the deviation. This invention achieves stable binding and individualized closed-loop intervention between motor intentions and the representation of the affected limb, improving training stability.

Description

Technical Field

[0001] This application relates to the field of intelligent control technology, and in particular to a motor imagery training system based on body affiliation. Background Technology

[0002] Traditional motor imagery brain-computer interface systems achieve external stimulus training through fixed-sequence feedback. While they have a basic training function, they suffer from problems such as weak body belonging perception, poor multi-sensory matching, discrete training effects, and significant individual differences, resulting in limited training effectiveness.

[0003] Although existing technologies have made improvements in feedback modes, due to the inherent limitations of their temporal regulation modes, they have not yet provided a solution that can match individual neural characteristics and achieve dynamic adaptive adjustment of temporal parameters. Summary of the Invention

[0004] The purpose of this application is to provide a motor imagery training system based on body ownership. In innovative motor imagery (MI) training, it establishes and maintains the target user's body ownership of virtual limbs, and performs individualized and dynamic control of the multimodal feedback timing based on the target user's individual alpha frequency in the parietal lobe, thereby enhancing the binding relationship between motor intention and the representation of the affected limb and improving training effectiveness.

[0005] In some embodiments, this application provides a motor imagery training system based on body ownership, the system comprising: a regulation parameter generation module, used to collect resting-state EEG signals of a target user, extract baseline individual alpha frequencies, calculate an individualized time-binding window based on a preset log-linear model, and generate corresponding temporal regulation parameters; a body ownership construction module, used to temporally pair virtual limb movements with multimodal feedback according to the temporal regulation parameters, perform virtual limb body ownership induction, and complete the construction of body ownership; and a reference template construction module, used to, after the body ownership is constructed, intercept EEG signals and extract feedback-locked EEG features to establish a stable reference template for the stable state of body ownership, wherein the feedback-locked EEG features are based on the actual effective time of the multimodal feedback pairing event as the time lock point, and the pairing event is intercepted. The system includes: a brainwave signal segment from before to after the event, extracting temporal and frequency domain features from the brainwave signal segment; a multimodal feedback module, used to present a motor imagery task, decode the target user's motor imagery intention, drive virtual limb movements and couple multimodal feedback output after the ownership and stable reference template is constructed; and an adaptive control module, used to calculate the first deviation of the task-state individual's alpha frequency relative to the baseline individual's alpha frequency and the second deviation of the feedback-locked brainwave features relative to the ownership and stable reference template after each round of multimodal feedback module execution, updating the timing control parameters for the next round of training based on the first deviation or reconstructing the body ownership and ownership and stable reference template based on the second deviation.

[0006] In some embodiments, the body ownership construction module includes: a temporal pairing unit, configured to constrain the time offset of the paired event relative to the dominant event in the multimodal feedback within an individualized time binding window, and execute multimodal stimulus sequence output; and a body ownership construction unit, configured to perform weighted calculation of the change amplitude of preset indicators and physiological indicators of the stimulated state relative to physiological indicators of the unstimulated state, determine an ownership index, and complete the construction of body ownership in response to the ownership index meeting preset conditions; the preset indicators include the ratio of the number of times the target limb is subjectively judged to be stimulated by the target user to the number of times the virtual limb is actually stimulated.

[0007] In some embodiments, the reference template construction module includes: an EEG segment extraction unit, used to extract valid EEG signal segments within a preset time period from before to after the occurrence of a multimodal feedback pairing event, using the actual effective time of the multimodal feedback pairing event as the time lock point after the body ownership is constructed; and a reference template construction unit, used to extract feedback locking EEG features from the valid EEG signal segments, and to calculate the average of the feedback locking EEG features of the valid EEG signal segments from multiple rounds to form the ownership stable reference template.

[0008] In some embodiments, the multimodal feedback module includes: a motor imagery decoding unit, used to present a motor imagery task of a virtual limb, collect training EEG signals and complete motor imagery intention recognition and decoding to obtain the real motor intention; and a multimodal feedback coupling unit, used to drive virtual limb movements according to time-series control parameters and the real motor intention, synchronously schedule multimodal feedback output, and collect the individual alpha frequency and feedback-locked EEG features of the current task state.

[0009] In some embodiments, the adaptive control module includes: an individual alpha frequency offset monitoring unit, used to calculate a first deviation of the task-state individual alpha frequency relative to the benchmark individual alpha frequency, and determine whether it exceeds a first preset threshold; a body ownership deviation monitoring unit, used to calculate a second deviation of the feedback-locked EEG features relative to the belonging stable reference template, and determine whether it exceeds a second preset threshold; and an adaptive control execution unit, which, in response to the first deviation exceeding the first preset threshold, updates the temporal control parameters based on the real-time collected individual alpha frequency and applies them to the next round of training; and, in response to the second deviation exceeding the second preset threshold, triggers the re-execution of the body ownership construction module and the reference template construction module to reconstruct body ownership based on the latest temporal control parameters.

[0010] In some embodiments, in the adaptive control execution unit, the first preset threshold is set based on the statistical distribution of the offset of individual alpha frequency during training, and the second preset threshold is set based on the statistical distribution of the second deviation.

[0011] In some embodiments, the adaptive control module further includes: a training closed-loop protection unit, configured to re-execute the body ownership construction module and the reference template construction module when the first deviation is detected to exceed the first preset threshold multiple times consecutively; and to perform cyclic training of the multimodal feedback module and the adaptive control module after the body ownership construction module and the reference template construction module are completed.

[0012] In some embodiments, the regulation parameter generation module includes: an EEG acquisition and evaluation unit, used to acquire resting-state EEG signals of the target user, perform preprocessing and power spectrum analysis, and extract the individual alpha frequency within a preset frequency band as the baseline individual alpha frequency; and an individual time-binding window calculation unit, used to call a pre-calibrated offline log-linear model, calculate the center value and range of the individualized time-binding window based on the baseline individual alpha frequency, and generate timing regulation parameters including feedback delay, trigger time, and synchronization error.

[0013] In some embodiments, the log-linear model is a log-linear model pre-fitted based on sample data, and the individual alpha frequency in the log-linear model is negatively correlated with the individualized time binding window.

[0014] In some embodiments, the regulation parameter generation module, the body ownership construction module, and the reference template construction module are executed sequentially; wherein, after the reference template construction module is completed, the multimodal feedback module and the adaptive regulation module are executed cyclically.

[0015] Compared with the prior art, the present invention has the following technical effects:

[0016] 1. This invention extracts the baseline individual alpha frequency by collecting the resting-state EEG signal of the target user, and calculates the individualized time binding window based on a preset log-linear model to generate temporal regulation parameters. It utilizes the systematic correlation between the individual alpha frequency and the multi-sensory temporal integration window, so that the temporal structure of multimodal feedback can accurately match the target user's current neural timescale and integrable time range. This overcomes the cross-modal time difference tolerance mismatch problem caused by the fixed feedback timing in the prior art, enhances the temporal integration efficiency of multi-source somatic related signals, and provides a neural temporal basis for the establishment of body ownership.

[0017] 2. This invention induces body ownership by temporally pairing virtual limb movements with multimodal feedback according to temporal control parameters before formal training, and extracts feedback to lock EEG features to establish a stable reference template for ownership after ownership is established. This provides a pre-calibration mechanism and neural response benchmark for establishing virtual limb ownership, so that feedback in subsequent training can be stably attributed to the affected limb itself. This solves the problem that feedback is difficult to incorporate into one's own body representation in the prior art, and significantly improves the immersion and compliance of training.

[0018] 3. This invention continuously collects individual task-state alpha frequencies and feedback-locked EEG features during training, calculates their deviation from the baseline, and adaptively updates temporal regulation parameters or triggers the reconstruction of body ownership and reference templates based on the deviation. This constructs a closed-loop intervention logic of first establishing ownership, then strengthening training, and reconstructing after mismatch. When the target user experiences neural timescale drift or ownership decline due to fatigue or attention fluctuations, the system can dynamically correct the feedback timing or force a return to the calibration phase, ensuring a continuous match between feedback and the target user's current multisensory integration ability. This avoids a re-mismatch between body representation and feedback during training and improves training stability. Attached Figure Description

[0019] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, wherein:

[0020] Figure 1 This is a schematic diagram of the overall system flow provided in one embodiment of this application;

[0021] Figure 2 This is a flowchart illustrating the adaptive control module provided in one embodiment of this application;

[0022] Figure 3 This is a schematic diagram of the system execution process provided in one embodiment of this application. Detailed Implementation

[0023] 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.

[0024] The technical solutions of the various embodiments of this application can be combined with each other, but only if they are based on the ability of a person skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by this application.

[0025] This solution does not aim to obtain disease diagnosis results or health status. It is simply a system that processes the user's brainwaves to achieve motor imagery training. All steps are performed by information processing methods implemented by devices such as computers.

[0026] It should be fully understood that the user information involved in this application (including but not limited to EEG signals) is information and data authorized by the user or fully authorized by all parties. The use of user information shall comply with the privacy policies and practices of the industry that are generally considered to meet or exceed the 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.

[0027] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, specific embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0028] Example 1:

[0029] like Figures 1-2As shown, this embodiment provides a motor imagery training system based on body ownership. The system constructs a closed-loop logical framework that first establishes ownership, then strengthens training, and reconstructs after mismatch. Through the interaction and data flow of system-level modules, it transforms abstract neural temporal calibration and dynamic adjustment mechanisms into an executable intervention loop. The system includes: a regulation parameter generation module, a body ownership construction module, a reference template construction module, a multimodal feedback module, and an adaptive regulation module.

[0030] like Figure 1 As shown, in some embodiments, the regulation parameter generation module is used to collect the resting-state EEG signal of the target user, extract the baseline individual alpha frequency, calculate the individualized time binding window based on a preset log-linear model, and generate the corresponding temporal regulation parameters. Specifically, the regulation parameter generation module is the starting point of the entire closed-loop system and the source of the neural temporal benchmark. In motor imagery training, different target users have significant individual differences in their tolerance range for cross-modal time differences, which are manifested in the difference of individual alpha frequencies in neurophysiology. If the system directly adopts a fixed feedback timing or empirical parameters, it is very easy to cause the feedback presentation timing to be mismatched with the target user's own integrable window, so that the subsequent attribution establishment and training lose the support basis of neural rhythm. Therefore, the present invention must first extract the baseline individual alpha frequency that can characterize its current neural time scale and multi-sensory temporal integration ability from the target user's resting-state EEG signal through the regulation parameter generation module, and use a preset log-linear model to transform it into an individualized time binding window, thereby generating temporal regulation parameters that include details such as feedback delay, trigger time, and synchronization error. These timing modulation parameters are output to downstream modules as core scheduling instructions, ensuring that all subsequent multimodal stimulus sequences can be accurately matched within the target user's current available timeframe. It should be understood that although the modulation parameter generation module in this embodiment extracts parameters based on resting-state EEG signals, in other embodiments, it can also extract parameters based on short-term EEG fragments from the initial stage of the task, as long as it satisfies the function of providing a stable, individualized timing baseline for the system.

[0031] like Figure 1As shown, in some embodiments, the body ownership construction module is used to temporally pair virtual limb movements with multimodal feedback according to the temporal control parameters, perform virtual limb body ownership induction, and complete the construction of body ownership. Specifically, the body ownership construction module, which receives the temporal control parameters output by the control parameter generation module, is key to solving the problem in the prior art: feedback is difficult to attribute to the affected limb itself. In existing training systems, there is often a lack of a pre-calibration mechanism for establishing virtual limb ownership, and target users are prone to regard the movement of virtual limbs as external events, resulting in limited immersion and compliance in training. The reason why this invention must enforce the body ownership construction module before formal training is that the formation of body ownership depends on the temporal integration of multi-source somatic signals such as vision, touch, and proprioception. Only when the temporal offset between virtual limb movements (such as visual touching or deformation) and multimodal feedback (such as tactile vibration) is strictly constrained within the individualized temporal binding window can the target user integrate these two originally independent cross-modal events into the same event related to their own target limb, thereby generating a sense of belonging to the virtual limb. The body ownership construction module enables target users to gradually attribute currently presented body-related events to their own target affected side by performing this personalized temporal induced pairing, thus laying the representational foundation for subsequent motor imagery control learning.

[0032] like Figure 1 As shown, in some embodiments, the reference template construction module is used to intercept EEG signals and extract feedback-locking EEG features after the body ownership is established, to establish a stable reference template for the stable state of body ownership. The feedback-locking EEG features are extracted from EEG signal segments within a preset time period from before to after the occurrence of the multimodal feedback pairing event, using the actual effective time of the multimodal feedback pairing event as the time lock point. Specifically, the reference template construction module follows the body ownership construction module in logical sequence, and its purpose is to provide a neural fingerprint comparison benchmark for ownership state monitoring in subsequent training. Judging whether ownership is maintained solely based on subjective feelings or behavioral responses is not objective and is easily interfered with; it is necessary to extract the EEG signal pattern in the successfully established ownership state as a reference. By intercepting EEG signals and extracting feedback-locking EEG features, the reference template construction module quantifies and solidifies the target user's brain response pattern to multimodal events related to the current target affected side, forming a stable reference template for ownership. This template characterizes the distribution of neural response features when body ownership is stably established. This allows the system to objectively capture minor declines in ownership status during subsequent training by calculating the second deviation between real-time features and the template, rather than intervening only after the target user has completely lost their sense of belonging.

[0033] like Figure 1 As shown, the multimodal feedback module is used to present the motor imagery task, decode the target user's motor imagery intention, drive virtual limb movements, and couple multimodal feedback output after the stable reference template is constructed. Simultaneously, it collects the individual's alpha frequency and feedback-locked EEG features in the current task state. Specifically, the multimodal feedback module is the core execution unit for the system's formal training phase. The multimodal feedback module is only allowed to start after the stable reference template is constructed, ensuring that training is a reinforcement learning process conducted under the premise that the target user already has stable body ownership. During execution, the multimodal feedback module is not only responsible for presenting the task, decoding the intention, and driving virtual limb movements, but more importantly, it must, based on the previously generated temporal control parameters, perform closed-loop coupling output of the virtual limb movements driven by the motor imagery decoding results and the multimodal feedback, forming a unified mapping relationship between motor intention, visual events, and affected-side sensory events. Simultaneously, to support subsequent adaptive control, the multimodal feedback module must synchronously collect the individual's alpha frequency and feedback-locked EEG features in each round of feedback output, transmitting real-time neural state data during training to the downstream adaptive control module. It should be understood that the specific form of multimodal feedback can be a combination of at least one or more types of feedback, such as visual feedback, tactile feedback, and electrical stimulation feedback. The system can be flexibly scheduled according to the specific condition and tolerance of the target user, as long as it can couple the intended result with the sensory event on the affected side.

[0034] like Figure 1As shown, the adaptive control module is used to calculate, after each round of multimodal feedback module execution, the first deviation of the task-state individual's Alpha frequency relative to the baseline individual's Alpha frequency and the second deviation of the feedback-locked EEG features relative to the stable reference template of ownership. Based on the first deviation, the temporal control parameters for the next round of training are updated, or based on the second deviation, the stable reference template of body ownership and ownership is reconstructed. Specifically, the adaptive control module is a dynamic protection mechanism to maintain the stable operation of the entire training loop. During long-term training, the target user will inevitably experience neural timescale drift or ownership decline due to fatigue, attentional fluctuations, or changes in task load. If the system only maintains the initial parameters, the feedback and body representation will become mismatched again. The reason this invention must use dual deviation determination instead of single determination is that the deviation of the task-state individual's Alpha frequency reflects the drift of the target user's underlying neural timescale, which is a macroscopic parameter change in temporal integration ability; while the deviation of the feedback-locked EEG features reflects the microscopic degradation of the neural response of body ownership in the current round. These two deviations have different physical meanings and mechanisms, thus requiring different control strategies. Specifically, when the underlying neural timescale drifts, the temporal control parameters for the next round of training only need to be updated based on the real-time collected individual alpha frequency, allowing the feedback timing to rematch the target user's current integration ability. However, when ownership neural responses degenerate, simply adjusting the temporal parameters is insufficient to salvage the collapse of the representation; it is necessary to trigger the re-execution of the body ownership construction module and the reference template construction module to reconstruct body ownership based on the latest temporal control parameters, forcing a regression to the calibration phase. Through this dual deviation determination and dual-path control strategy, the adaptive control module achieves a closed-loop intervention logic for reconstruction after mismatch, preventing the training system from falling into an ineffective infinite loop when parameters are mismatched or ownership is lost, thus ensuring the stability of training.

[0035] like Figure 1As shown, in some embodiments, the regulation parameter generation module includes: an EEG acquisition and evaluation unit, used to acquire the resting-state EEG signal of the target user, complete preprocessing and power spectrum analysis, and extract the individual alpha frequency within a preset frequency band as the baseline individual alpha frequency; and an individual time-binding window calculation unit, used to call a pre-calibrated offline log-linear model, calculate the center value and range of the individualized time-binding window based on the baseline individual alpha frequency, and generate timing regulation parameters including feedback delay, trigger time, and synchronization error. Specifically, after acquiring the resting-state EEG signal of the target user, the EEG acquisition and evaluation unit first performs preprocessing operations such as bandpass filtering, artifact removal, abnormal segment removal, and rereference to eliminate interference from electrooculography (EOG) and electromyography (EMG); then, it performs power spectral density estimation on the preprocessed signal and identifies the frequency point with the highest power within the 8 to 13 Hz alpha band as the baseline individual alpha frequency. It should be understood that although this embodiment uses the maximum power frequency point extraction method, in other embodiments, algorithms such as FOOOF can be used to separate the non-periodic background components in the power spectrum before extracting the periodic alpha peak frequency to improve the stability and accuracy of peak frequency identification, as long as the reference frequency characterizing the individual neural time scale can be extracted. After receiving the reference individual alpha frequency, the individual time binding window calculation unit calls a pre-calibrated offline log-linear model to calculate the center value and range of the individualized time binding window, and generates timing control parameters including feedback delay, trigger time, and synchronization error. These timing control parameters are the core instructions for subsequent precise scheduling of multimodal feedback. Their generation process transforms abstract neural rhythm characteristics into executable hardware control timing, ensuring accurate matching between the feedback presentation and the target user's own integrable window.

[0036] like Figure 1As shown, in some embodiments, the body ownership construction module includes: a temporal pairing unit, used to constrain the time offset of the paired event relative to the dominant event in multimodal feedback within an individualized time binding window, and execute multimodal stimulus sequence output; a body ownership construction unit, used to perform weighted calculation of the change amplitude of preset indicators and physiological indicators of the stimulated state relative to physiological indicators of the unstimulated state, determine the ownership index, and complete the construction of body ownership in response to the ownership index meeting preset conditions; wherein, the preset indicators include the ratio of the number of times the target limb is subjectively stimulated by the target user to the number of times the virtual limb is actually stimulated. Specifically, when the temporal pairing unit performs pairing, it involves two types of key events: one type is the dominant event that is more easily perceived by the target user first, such as visual changes such as the corresponding part of the virtual limb being touched, undergoing local deformation, changing posture, or the local area being highlighted; the other type is the pairing event that is time-coordinated with the dominant event, such as tactile vibration, electrical stimulation, mechanical traction, temperature change, or sound cues, etc., and the pairing event plays the role of attributing the sensation to one's own limb. The temporal pairing unit strictly constrains the time offset of the paired event relative to the dominant event within a personalized time binding window, ensuring that these two cross-modal events are integrated into a single body-related event in the target user's perception. The body ownership construction unit calculates an ownership index using multiple equivalent indicators to objectively assess the degree of belonging established. For example, the system provides stimulation to a virtual hand (i.e., a virtual limb) or the affected hand (i.e., the target limb) (the stimulation includes touching, pulling, etc.). The target user observes and judges whether the affected hand is stimulated. The system counts the number of times the virtual hand is stimulated but the target user judges that the affected hand is stimulated, to the number of times the virtual hand is stimulated, denoted as S_beh. The system collects peripheral physiological indicators such as skin conductance and electromyography (EMG) signals of the target user in the unstimulated state, denoted as the baseline state. The system collects peripheral physiological indicators such as skin conductance and EMG signals of the target user in the stimulated state (e.g., when the virtual hand presents body-related visual events such as needle pricks, touches, and pressure), judges whether the corresponding affected hand shows matching skin conductance enhancement, local EMG response, or other physiological responses, and calculates the change amplitude relative to the baseline state. This change amplitude is normalized and denoted as the physiological index S_phys. Finally, S_beh and S_phys are weighted and fused according to preset weights to obtain the ownership index own_index, for example:

[0037] own_index=w1*S_beh+w2*S_phys,

[0038] In the formula, w1 represents the weight of S_beh, w2 represents the weight of S_phys, and w1 + w2 = 1. When the ownership index (own_index) reaches the preset threshold, it is determined that the establishment of body ownership is successful. In another embodiment, the ownership index can also be calculated through the target user's subjective rating (such as the rating of the virtual hand's ownership in a questionnaire). When the ownership index meets the preset conditions, such as a subjective rating of 4 points or above (out of 5), the system determines that the current body ownership has been effectively established, thus completing the construction. The significance of the preset conditions is to provide the system with an objective and quantitative threshold for exiting the induction phase, avoiding infinite induction loops caused by vague subjective feelings, and also preventing the risk of entering the training phase too early before ownership is stable. It should be understood that the calculation method of the ownership index is highly flexible and replaceable. The most suitable combination of indicators can be selected according to the target user's specific condition and cognitive state, as long as it can quantify the degree of belonging.

[0039] like Figure 1As shown, in some embodiments, the reference template construction module includes: an EEG segment extraction unit, used to extract valid EEG signal segments within a preset time period from before to after the occurrence of a multimodal feedback pairing event, using the actual effective time of the multimodal feedback pairing event as the time lock point after the body ownership is constructed; and a reference template construction unit, used to extract feedback-locked EEG features from the valid EEG signal segments, and to average the feedback-locked EEG features of multiple rounds of valid EEG signal segments to form the ownership stable reference template. Specifically, the extraction logic of the EEG segment extraction unit has specific neurophysiological considerations. Feedback-locked EEG features refer to the time-domain and frequency-domain features extracted from the EEG signal segments extracted and extracted by the system using the actual effective time of the multimodal feedback pairing event as the time lock point. Specifically, time-domain features include signal average amplitude, peak-to-peak value, root mean square value, and variance. Frequency-domain features can be obtained through Fast Fourier Transform, power spectral density estimation, or time-frequency analysis. These features include absolute power, relative power, and band power ratios for the delta band (0.5-3Hz), theta band (4-7Hz), alpha band (8-13Hz), and beta band (14-30Hz). The actual effective time, rather than the dominant event time, is chosen because the actual effectiveness of the paired event (such as tactile vibration) is the key trigger point for the completion of multisensory integration and the generation of a sense of belonging neural response. The feedback-locked EEG features typically capture the time period from 200 milliseconds before to 800 milliseconds after the paired event, covering the preparation period for integration and the latency and peak period of the neural response after integration. After extracting feedback-locked EEG features (such as time-domain average amplitude, amplitude standard deviation, and power spectral density of each frequency band), the reference template construction unit averages the features of effective EEG signal segments from multiple rounds to form a stable reference template. For example, in the j-th effective round of the reference template construction stage, the original EEG signal is denoted as X_j(t,c), where t is the time point and c is the channel number; the actual effective time of the paired event is denoted as T_pair(j). The system uses T_pair(j) as the time lock point to extract the event-locked segment from the original EEG:

[0040] E_j(τ,c)=X_j(T_pair(j)+τ,c),τ∈[−200ms,800ms]

[0041] This involves taking an EEG segment from 200 ms before the paired event to 800 ms after the event for each round, used to describe the target user's brain response pattern to multimodal events related to the affected side of the target brain. The temporal features f_j (e.g., average amplitude, amplitude standard deviation) and frequency features p_j (e.g., power spectral density of each frequency band) of the EEG segment from 200 ms before the paired event to 800 ms after the event are extracted to obtain the feature vector F_ref_j for the j-th reference round. The mean of the feature vectors F_ref_j for the N valid rounds of the reference phase is calculated to obtain the assigned stable reference template F_ref and the standard deviation of each dimension of the features. For example, when calculating the average signal amplitude, the absolute value of the signal is first taken, and then the arithmetic mean of all absolute values ​​of the signal within that time period (i.e., from 200 ms before the paired event to 800 ms after the event) is calculated; this arithmetic mean is the average signal amplitude. The purpose of this multi-round feature averaging noise reduction is to eliminate random artifacts and transient fluctuations in a single EEG signal acquisition, and to extract the most representative and robust neural response patterns under stable ownership conditions, thereby providing a high signal-to-noise ratio comparison benchmark for deviation calculation in subsequent training.

[0042] like Figure 1As shown, in some embodiments, the multimodal feedback module includes: a motor imagery decoding unit, used to present a virtual limb motor imagery task, collect training EEG signals, and complete the recognition and decoding of motor imagery intentions to obtain the real motor intention; and a multimodal feedback coupling unit, used to drive virtual limb movements according to temporal control parameters and the real motor intention, synchronously schedule multimodal feedback output, and collect the individual alpha frequency and feedback-locked EEG features of the current task state. Specifically, the motor imagery decoding unit synchronously collects training EEG signals after presenting the task, and completes the recognition and decoding of motor imagery intentions through feature extraction methods such as cosmic patterns and classification algorithms such as support vector machines to obtain the real motor intention. The multimodal feedback coupling unit then performs closed-loop coupling output of the virtual limb movements driven by the real motor intention and the multimodal feedback according to the previously generated temporal control parameters. It should be noted that the coupling unit must precisely control the presentation delay of visual feedback, the triggering time of tactile feedback, the triggering time of electrical stimulation feedback, and the allowable synchronization error range between different modalities, so that the motor imagery results, visual events, and affected-side sensory events form a unified mapping relationship within the individualized time binding window. Simultaneously, to support downstream adaptive regulation, the multimodal feedback coupling unit must synchronously collect the individual's alpha frequency and feedback-locked EEG characteristics during each round of feedback output, transmitting real-time neural state data to downstream modules. It should be understood that the specific form of multimodal feedback can be a combination of at least one or more types of feedback, such as visual feedback, tactile feedback, vibration feedback, functional electrical stimulation feedback, or mechanical traction feedback. The system can flexibly schedule modal combinations according to the target user's specific condition and tolerance level, as long as it satisfies the function of temporally coupling the intended outcome with sensory events on the affected side.

[0043] like Figure 1 and Figure 2 As shown, in some embodiments, the adaptive control module includes: an individual alpha frequency offset monitoring unit, used to calculate a first deviation of the task-state individual alpha frequency relative to the benchmark individual alpha frequency, and determine whether it exceeds a first preset threshold; a body ownership deviation monitoring unit, used to calculate a second deviation of the feedback-locked EEG features relative to the belonging stable reference template, and determine whether it exceeds a second preset threshold; and an adaptive control execution unit, which, in response to the first deviation exceeding the first preset threshold, updates the temporal control parameters according to the real-time collected individual alpha frequency and applies them to the next round of training; and, in response to the second deviation exceeding the second preset threshold, triggers the re-execution of the body ownership construction module and the reference template construction module to reconstruct body ownership according to the latest temporal control parameters.

[0044] Specifically, the first deviation is usually denoted as IAF (Individual Alpha Frequency) offset E_iaf, which is calculated as the percentage deviation of the current task-state individual's Alpha Frequency relative to the baseline individual's Alpha Frequency. For example:

[0045] E_iaf=|IAF_current-IAF_baseline| / (IAF_baseline+ε),

[0046] In the formula, ε is a minimal constant to prevent the denominator from being zero, IAF_current is the individual alpha frequency in the task state, and IAF_baseline is the baseline individual alpha frequency. The physical meaning of E_iaf reflects the drift of the target user's underlying neural timescale. This drift is usually caused by fatigue, attentional fluctuations, or changes in task load, leading to a change in the target user's tolerance range for cross-modal time differences. At this time, the window of the target user's multi-sensory temporal integration ability has only undergone a width shift, but the representational foundation of body ownership has not collapsed. Therefore, the adaptive control execution unit only needs to update the temporal control parameters for the next round of training based on the real-time collected individual alpha frequencies, so that the feedback temporal sequence can be rematched with the target user's current integration ability, without the need to rebuild ownership. It should be understood that although this embodiment uses an offset percentage to calculate E_iaf, other equivalent calculation methods such as absolute frequency difference or normalized frequency difference can also be used in other embodiments, as long as the function of quantifying the degree of neural timescale drift is satisfied.

[0047] Specifically, the second deviation is usually denoted as ownership deviation D_own, and its calculation logic is the Euclidean distance or Mahalanobis distance between the current round's feedback-locked EEG features and the stable reference template of ownership. The physical meaning of D_own reflects the micro-state degradation of the body ownership neural response in the current round, meaning that the target user can no longer stably integrate the multimodal feedback of the virtual limb into the same event related to their own target limb, and the representation binding relationship is collapsing. When this state degradation occurs, simply adjusting the temporal parameters can no longer salvage the collapse of the representation, because the ownership neural response pattern has deviated from the stable benchmark, and continuing to train under the mismatched representation will only reinforce the erroneous mapping relationship. Therefore, the adaptive control execution unit must trigger the re-execution of the body ownership construction module and the reference template construction module, forcing the system to return to the calibration phase, so as to re-induce body ownership and solidify the new neural response benchmark according to the latest temporal control parameters. Through this strategy, this embodiment achieves precise and economical closed-loop intervention, avoiding both excessive intervention during drift and ineffective training during degradation.

[0048] like Figure 1 and Figure 2As shown, in some embodiments, in the adaptive control execution unit, the first preset threshold is set based on the statistical distribution of the offset of individual alpha frequencies during training, and the second preset threshold is set based on the statistical distribution of the second deviation. Specifically, the threshold setting is not a fixed value selected based on experience, but a scientific threshold dynamically set based on its own data statistical distribution. This design provides the system with an objective standard for preventing circumvention.

[0049] Specifically, the first preset threshold, usually denoted as d2, is used to limit the allowable offset range of the individual peak frequency relative to the baseline in the current task state. It is set based on the statistical distribution of E_iaf during training, typically taking the mean plus two standard deviations. This means that within the normal range of neural rhythm fluctuations, almost all random drifts will not trigger parameter updates; only significant drifts exceeding the normal fluctuation range will be considered as requiring intervention. For example, in a specific application scenario, d2 can be set to 15%, meaning that the time-series parameters are only updated when the IAF offset exceeds 15% of the baseline.

[0050] Specifically, the second preset threshold, usually denoted as d1, is used to limit the deviation range of the feedback-locked EEG features relative to the template. Its setting is based on the statistical distribution of D_own, typically taking the mean plus two standard deviations. This is because deviations in ownership neural responses have a more sensitive and direct impact on training effectiveness, with lower tolerance. For example, in a specific application scenario, d1 can be set to 0.3, meaning that ownership reconstruction is triggered when the Euclidean distance deviation exceeds 0.3.

[0051] It should be understood that although this embodiment provides specific statistical basis for the mean ± 2 standard deviations and specific numerical examples of d1=0.3 and d2=15%, in other embodiments, other statistical quantiles such as mean ± 1 standard deviations or mean ± 3 standard deviations may be used according to the specific disease stability and training stage goals of the target user, as long as they can objectively distinguish between normal fluctuations and abnormal deviations.

[0052] like Figure 1 and Figure 2As shown, in some embodiments, the adaptive control module further includes: a training closed-loop protection unit, used to re-execute the body ownership construction module and the reference template construction module when the first deviation is detected to exceed the first preset threshold multiple times consecutively; and to perform cyclic training of the multimodal feedback module and the adaptive control module after completing the body ownership construction module and the reference template construction module. Specifically, the significance of the training closed-loop protection unit is to prevent the system from falling into an infinite loop of invalid parameter updates. In actual training, there may be an extreme case: the target user's neural state continues to deteriorate, causing E_iaf to exceed d2 after each round of training. The system continuously updates the temporal parameters, but the next round still exceeds the limit. At this time, relying solely on the single parameter update of the adaptive control execution unit can no longer converge the system state. To deal with this infinite loop risk, the training closed-loop protection unit introduces a forced reconstruction mechanism for multiple consecutive exceedances. A specific example of multiple consecutive exceedances here can be three consecutive exceedances, that is, when the system detects that E_iaf exceeds d2 for three consecutive training rounds, it no longer simply updates the parameters, but directly determines that the current ownership representation has completely failed, and forcibly re-executes the body ownership construction module and the reference template construction module. After forced reconstruction is completed, the system resumes cyclical training of the multimodal feedback module and adaptive control module with new temporal control parameters and a new attribution stable reference template. This pulls the system out of the infinite loop, ensuring that the training loop always operates within a healthy logic of first establishing attribution, then strengthening training, and rebuilding after mismatch. It should be understood that although this embodiment uses three consecutive times as a specific example of triggering forced reconstruction, in other embodiments, it can also be set to two or five consecutive times according to clinical requirements for training continuity, as long as it satisfies the function of timely fallback when the system fails to converge.

[0053] The above description of the deviation calculation logic, threshold statistical distribution setting and closed-loop protection mechanism is only explanatory and not restrictive. It is intended to provide sufficient neurophysiological basis and engineering implementation details for the dual deviation determination and dual-path control strategy, and to prevent the specific values ​​or formulas of the embodiments from being imposed on the higher-level generalization when determining infringement.

[0054] In some embodiments, the log-linear model is a pre-fitted log-linear model based on sample data, and the individual alpha frequency in the log-linear model is negatively correlated with the individualized time-binding window. Specifically, the formula for the log-linear model is:

[0055] ln(TBW_i)=β_0+β_1·IAF_i+ε_i,

[0056] In the formula, TBW_i is the estimated width of the time-binding window for the i-th sample, IAF_i is the individual alpha frequency of the i-th sample, β_0 and β_1 are the model parameters obtained by offline fitting based on the reference population sample data, and ε_i is the residual term. The reason this invention uses a log-linear model instead of a simple linear model is that the width change of the time-binding window exhibits an exponentially decaying nonlinear characteristic on a physical scale. Logarithmic transformation can linearize this, thereby ensuring the accuracy of the model fitting and the normal distribution characteristics of the residuals. More importantly, the model parameter β_1 < 0, and this negative correlation has a profound neurophysiological mechanism: IAF characterizes the temporal resolution of the target user's posterior parietal cortex neural rhythms. A higher IAF means a shorter neural oscillation period, and a stronger ability for the target user to perceive and distinguish cross-modal time differences. Therefore, it can integrate multimodal events within a shorter time range into the same bodily-related event, and the corresponding individualized temporal binding window is narrower. Conversely, a lower IAF means weaker temporal resolution, a wider tolerance range for time offsets, and a correspondingly wider TBW (i.e., multisensory temporal integration window). This negative correlation mechanism ensures that the system can accurately generate a temporal window that matches the target user's integration ability based on their current neural timescale.

[0057] During the online computation phase, after receiving the baseline individual alpha frequency (IAF_baseline) of the current target user, the individual time-binding window computation unit first substitutes it into the log-linear model to calculate the center value of the individualized time-binding window:

[0058] TBW_center=exp(β_0+β_1·IAF_baseline).

[0059] The central value represents the most suitable multimodal feedback synchronization timing benchmark for the target user under the current neural rhythm state.

[0060] To support window fine-tuning and multimodal feedback timing selection in subsequent adaptive regulation, the system also needs to further calculate the range of the individualized time-binding window:

[0061] TBW_low=exp(β_0+β_1·IAF_baseline−z·ε) and TBW_high=exp(β_0+β_1·IAF_baseline+z·ε),

[0062] In the formula, z is the preset standard normal distribution quantile, and ε is the residual standard deviation in the offline calibration model. For example, when z is set to 1.96, corresponding to approximately a 95% confidence interval, it means that the allowed delay range of the timing regulation parameters generated by the system can cover 95% of normal physiological fluctuations. This avoids misjudging normal synchronization as mismatch due to an overly narrow range, and also prevents loss of timing constraint accuracy due to an overly wide range. It should be understood that although this embodiment provides a specific example of z being 1.96, in other embodiments, z can be set to other quantiles such as 1 or 2.58, depending on different clinical requirements for timing accuracy, as long as it satisfies the function of providing a scientific confidence boundary for the timing regulation parameters.

[0063] like Figure 1 As shown, in some embodiments, the regulation parameter generation module, the body ownership construction module, and the reference template construction module are executed sequentially; wherein, after the reference template construction module is executed, the multimodal feedback module and the adaptive regulation module are executed cyclically. Specifically, this embodiment not only specifies the function of each module, but also anchors the entire training process into an irreversible execution logic through strict timing constraints. The technical reason for this inverse timing sequence lies in the fact that the outputs of each module logically form a tight chain of dependencies: the timing control parameters output by the parameter generation module are the sole instruction basis for the body ownership construction module to execute multimodal timing pairing. If parameters are not generated first, pairing will become a fixed-time experience output, unable to match the individual's integration ability; the belonging experience successfully induced by the body ownership construction module is a prerequisite for the reference template construction module to capture the EEG signals of a stable ownership state. If ownership is not established first, the captured EEG features are merely passive responses to external events rather than ownership representations, and subsequent comparisons will be meaningless; the stable ownership reference template output by the reference template construction module is also the benchmark for comparing the multimodal feedback module's training initiation and the adaptive control module's calculation of deviation. If the benchmark is not solidified first, ownership degradation during training will be undetectable. Therefore, the pre-calibration timing sequence of parameter generation, ownership construction, and template construction must be strictly followed.

[0064] After the reference template is constructed, the system enters the training phase of cyclically executing the multimodal feedback module and the adaptive control module. This cyclical sequence is also unbreakable: the adaptive control module can only calculate the deviation of the current round after each round of multimodal feedback output is completed, and determine the timing parameters of the next round or whether to interrupt the cycle to trigger reconstruction based on the deviation result. If this cyclical sequence is reversed, and control is performed before feedback acquisition, it will lead to a lack of real-time data support and blind intervention. Through this rigid global timing constraint, this embodiment transforms the closed-loop intervention logic of first establishing attribution, then strengthening training, and rebuilding after mismatch from an abstract concept into an unavoidable system execution flow, completely eliminating the engineering loopholes of skipping calibration and directly training or blindly controlling without feedback.

[0065] The above description of the log-linear model formula and temporal logic is explanatory and not restrictive. It is intended to provide specific support for the full disclosure of algorithm patents and the defense of temporal features, and to prevent the imposition of specific numerical values ​​or formulas of the embodiments on higher-level generalizations during infringement determination.

[0066] Example 2:

[0067] To more clearly illustrate the technical solution of this invention and its operational effect in a real clinical environment, a detailed description is provided below using a practical application scenario of upper limb motor imagery training for stroke target users. It should be understood that this scenario is merely illustrative and not restrictive, intended to map abstract parameters and algorithms to specific medical behaviors, demonstrate the effectiveness of closed-loop intervention, and prevent the imposition of specific numerical values ​​or scenarios from the embodiments onto higher-level generalizations during infringement determination.

[0068] In this application scenario, the specific target user is defined as a person with right-sided upper limb paralysis after a stroke, and the virtual limb is set as the right virtual hand. For example... Figure 3 As shown, the system executes the complete training process as follows:

[0069] S100, the regulation parameter generation module acquires the resting-state EEG signal of the target user, extracts the baseline individual alpha frequency, calculates the individualized time-binding window based on a preset log-linear model, and generates the corresponding temporal regulation parameters. Specifically, the target user wears the EEG acquisition device and sits in relaxation. The system acquires the resting-state EEG signal of the relevant leads in the parietal and posterior lobe. After preprocessing and power spectrum analysis, the frequency point with the highest power in the 8 to 13 Hz frequency band is identified, and the baseline individual alpha frequency IAF_baseline = 10.2 Hz is obtained. Subsequently, the system calls the pre-calibrated offline log-linear model to calculate the center value TBW_center = 120 ms of the individualized time-binding window, and generates the temporal regulation parameters including feedback delay, trigger time, and synchronization error. It should be understood that although IAF_baseline is 10.2 Hz and TBW_center is 120 ms in this scenario, in other application scenarios for target users, these parameters will dynamically change with different individual neural time scales. This is the core of the present invention to achieve individualized temporal matching.

[0070] S200, the body ownership construction module performs temporal pairing of virtual limb movements and multimodal feedback according to the aforementioned temporal control parameters, executes virtual limb body ownership induction, and completes the construction of body ownership. Specifically, in the induction phase, the system uses visual changes as the dominant event, i.e., presenting a visual animation of the right virtual hand being touched; and uses tactile vibration as the pairing event, i.e., synchronously applying vibration stimulation to the back of the target user's right hand. The temporal pairing unit strictly constrains the time offset of the pairing event relative to the dominant event within the individualized time binding window, i.e., random or fixed pairing within ±60ms, so that these two cross-modal events are integrated into the same body-related event in the target user's perception. After multiple rounds of induction, the body ownership construction unit calculates the ownership index through the target user's subjective rating. If the target user's subjective rating of the right virtual hand belonging to them reaches 4 points (out of 5), meeting the preset conditions, the system determines that the current body ownership has been effectively established, thus completing the construction. It should be understood that although subjective ratings are used as the calculation method for the ownership index in this scenario, in other application scenarios, physiological indicators such as skin conductance response can be used to replace subjective ratings for target users with severe cognitive impairment, as long as the function of objectively quantifying the degree of belonging is met.

[0071] S300, after the body ownership is established, the reference template construction module extracts EEG signals and feedback-locked EEG features to establish a stable reference template for the stable state of body ownership. Specifically, after successful ownership establishment, the system automatically collects multiple consecutive valid rounds as reference stages. The EEG segment extraction unit uses the actual effective moment of tactile vibration feedback as the time lock point to extract valid EEG signal segments from 200 milliseconds before the vibration occurs to 800 milliseconds after it occurs. The reference template construction unit extracts time-domain and frequency-domain features from the valid EEG signal segments of multiple rounds and calculates the average to form a stable ownership reference template F_ref. This template represents the neural response fingerprint of the target user when they stably possess the sense of ownership of the right virtual hand.

[0072] S400, after the attribution stable reference template is constructed, the multimodal feedback module presents a motor imagery task, decodes the target user's motor imagery intention, drives virtual limb movements, and couples out multimodal feedback. Simultaneously, it collects the individual's alpha frequency and feedback-locked EEG characteristics for the current task state. Specifically, after entering the formal training phase, the system presents the target user with a grasping motor imagery task of the right virtual hand. The target user imagines grasping with their right hand according to the prompts. The motor imagery decoding unit collects training EEG signals and completes intention recognition and decoding, obtaining the actual motor intention as grasping. The multimodal feedback coupling unit, based on timing control parameters, drives the right virtual hand to perform the grasping action and synchronously schedules tactile vibration feedback output under individualized time-binding window constraints, forming a closed-loop coupling between the motor imagery result, the visual grasping event, and the affected side's tactile event. Simultaneously, the system synchronously collects the individual's alpha frequency and feedback-locked EEG characteristics for the current task state and transmits them to downstream modules.

[0073] In S500, after each round of multimodal feedback, the adaptive control module calculates the first deviation of the task-state individual's Alpha frequency relative to the baseline individual's Alpha frequency and the second deviation of the feedback-locked EEG features relative to the stable reference template. Based on the first deviation, it updates the timing control parameters for the next round of training or reconstructs the body ownership and stable reference template based on the second deviation. Specifically, in a certain round of training, the target user experiences fatigue due to prolonged training, causing a drift in their underlying neural rhythm, and the current task-state individual's Alpha frequency drops to 9.5Hz. The individual Alpha frequency deviation monitoring unit calculates the first deviation E_iaf=|9.5-10.2| / (10.2+0.001)≈6.8%, determining that it does not exceed the first preset threshold d2=15%. However, due to fatigue simultaneously causing the target user's attention to wander, their sense of ownership over the virtual hand undergoes a microscopic degradation. The body ownership deviation monitoring unit calculates the second deviation D_own of the current round's feedback-locked EEG features relative to the stable reference template, determining that it exceeds the second preset threshold d1=0.3. At this point, although the macroscopic drift of the underlying neural timescale is still within a tolerable range, the microscopic state of ownership neural response has degenerated, and the representation binding relationship is disintegrating. The adaptive regulation execution unit responds to the second deviation exceeding the second preset threshold by triggering the re-execution of the body ownership construction module and the reference template construction module to re-induce body belonging based on the latest temporal regulation parameters and solidify the new neural response benchmark, forcing the system to return to the calibration phase.

[0074] This scenario clearly demonstrates that this invention, through dual deviation determination and dual-path control strategies, can promptly trigger closed-loop intervention at the critical moment when the underlying neural rhythm drift has not yet severely exceeded the limit but ownership representation has already degenerated. This avoids ineffective training where simply adjusting timing parameters cannot salvage the representation collapse, thus ensuring training stability. It should be understood that although this scenario lists specific values ​​for IAF dropping to 9.5Hz, E_iaf at 6.8%, and D_own exceeding the limit due to fatigue, in actual training for other target users, the specific deviation values ​​and trigger thresholds will be dynamically set based on individual statistical distributions, as long as they can objectively distinguish between normal fluctuations and abnormal deviations.

[0075] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. The module division, unit combination, algorithm formula, threshold values, and clinical application scenarios detailed in the foregoing embodiments of the present invention are all intended to provide a concrete interpretation and defensive expansion of the abstract claim features through specific engineering and neurophysiological logic, aiming to provide sufficient disclosure support for the claims, rather than limiting the scope of protection of the present invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention, such as equivalent substitutions for multimodal feedback modes, adaptive adjustments to the deviation calculation logic, or conventional optimizations to the number of closed-loop reconstruction triggers, should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0076] 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 specification.

[0077] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. 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 patent application should be determined by the appended claims.

Claims

1. A motor imagery training system based on body affiliation, characterized in that, The system includes: The regulation parameter generation module is used to collect the resting-state EEG signal of the target user, extract the baseline individual alpha frequency, calculate the individualized time binding window based on the preset log-linear model, and generate the corresponding time-series regulation parameters. The body ownership construction module is used to perform temporal pairing of virtual limb movements and multimodal feedback according to the temporal control parameters, execute body ownership induction of virtual limbs, and complete the construction of body ownership; The reference template construction module is used to intercept EEG signals and extract feedback locking EEG features after the body ownership is constructed, and to establish a stable reference template for the stable state of body ownership. The feedback locking EEG features are based on the actual effective time of the multimodal feedback pairing event as the time lock point, and the EEG signal segments within a preset time period from before to after the pairing event are intercepted. The time domain features and frequency domain features are extracted from the EEG signal segments. The multimodal feedback module is used to present the motor imagery task, decode the target user's motor imagery intention, drive virtual limb movements and couple out multimodal feedback after the attribution stable reference template is constructed; at the same time, it collects the individual alpha frequency and feedback-locked EEG features in the current task state. The adaptive control module is used to calculate the first deviation of the task-state individual's Alpha frequency relative to the baseline individual's Alpha frequency and the second deviation of the feedback-locked EEG features relative to the stable ownership template after each round of multimodal feedback module execution. Based on the first deviation, the temporal control parameters for the next round of training are updated, or based on the second deviation, the stable reference template for body ownership and attribution is reconstructed.

2. The system according to claim 1, characterized in that, The body ownership construction module includes: The temporal pairing unit is used to constrain the time offset of the paired event relative to the dominant event in the multimodal feedback within the range of the individualized time binding window, and to execute the multimodal stimulus sequence output. The body ownership construction unit is used to perform weighted calculations on the changes in preset indicators and physiological indicators under stimulated conditions relative to physiological indicators under unstimulated conditions to determine an ownership index. In response to the ownership index meeting preset conditions, the body ownership construction is completed. The preset indicators include the ratio of the number of times the target limb is subjectively perceived to be stimulated by the target user to the number of times the virtual limb is actually stimulated.

3. The system according to claim 1, characterized in that, The reference template construction module includes: The EEG segment extraction unit is used to extract valid EEG signal segments within a preset time period from before to after the occurrence of the multimodal feedback pairing event, using the actual effective time of the pairing event as the time lock point after the body ownership is established. The reference template construction unit is used to extract feedback-locked EEG features from the effective EEG signal segments, and to calculate the average of the feedback-locked EEG features from multiple rounds of effective EEG signal segments to form the attribution stable reference template.

4. The system according to claim 1, characterized in that, The multimodal feedback module includes: The motor imagery decoding unit is used to present virtual limb motor imagery tasks, collect training EEG signals, and complete the recognition and decoding of motor imagery intentions to obtain the real motor intentions. The multimodal feedback coupling unit is used to drive virtual limb movements based on time-series control parameters and real movement intentions, synchronously schedule multimodal feedback outputs, and collect individual alpha frequencies and feedback-locked EEG characteristics in the current task state.

5. The system according to claim 1, characterized in that, The adaptive control module includes: The individual alpha frequency offset monitoring unit is used to calculate the first deviation of the individual alpha frequency in the task state relative to the reference individual alpha frequency, and to determine whether it exceeds the first preset threshold. The body ownership deviation monitoring unit is used to calculate the second deviation of the feedback-locked EEG features relative to the stable reference template of ownership, and to determine whether it exceeds the second preset threshold. The adaptive control execution unit, in response to the first deviation exceeding the first preset threshold, updates the temporal control parameters according to the real-time collected individual alpha frequency and applies them to the next round of training; in response to the second deviation exceeding the second preset threshold, it triggers the re-execution of the body ownership construction module and the reference template construction module to reconstruct body ownership according to the latest temporal control parameters.

6. The system according to claim 5, characterized in that, In the adaptive control execution unit, the first preset threshold is set according to the statistical distribution of the offset of individual Alpha frequency during training, and the second preset threshold is set according to the statistical distribution of the second deviation.

7. The system according to claim 5, characterized in that, The adaptive control module also includes: The training closed-loop protection unit is used to re-execute the body ownership construction module and the reference template construction module when the first deviation is detected to exceed the first preset threshold multiple times in a row; after completing the body ownership construction module and the reference template construction module, it performs cyclic training of the multimodal feedback module and the adaptive control module.

8. The system according to claim 1, characterized in that, The regulation parameter generation module includes: The EEG acquisition and assessment unit is used to acquire the resting-state EEG signals of the target user, perform preprocessing and power spectrum analysis, and extract the individual alpha frequency within a preset frequency band as the baseline individual alpha frequency. The individual time-binding window calculation unit is used to call a pre-calibrated offline log-linear model to calculate the center value and range of the individual time-binding window based on the baseline individual alpha frequency, and generate timing control parameters including feedback delay, trigger time, and synchronization error.

9. The system according to claim 8, characterized in that, The log-linear model is a pre-fitted log-linear model based on sample data, and the individual alpha frequency in the log-linear model is negatively correlated with the individualized time binding window.

10. The system according to any one of claims 1 to 9, characterized in that, The control parameter generation module, the body ownership construction module, and the reference template construction module are executed sequentially; wherein, after the reference template construction module is completed, the multimodal feedback module and the adaptive control module are executed cyclically.

Images