Acupuncture rehabilitation training data processing method based on combination of VR and digital therapy

Abstract

The invention relates to an acupuncture rehabilitation training data processing method based on the combination of VR and digital therapy, and the method comprises the steps: obtaining an acupuncture stimulation signal and a VR training interaction signal, carrying out the synchronous association of the two signals based on a unified time reference, and dynamically determining a signal processing priority according to the state of a patient when the acupuncture operation and a virtual training task conflict, carrying out backtracking calibration on the offset signal; the task action rhythm, force and completion condition of a patient are monitored in real time in VR rehabilitation training, data of abnormal behaviors are collected and recorded, whether the abnormal behaviors are caused by acupuncture stimulation or not is analyzed by combining acupuncture stimulation parameters and time sequence information, and data confirmed to belong to acupuncture effect interference are marked as independent samples; real-time subjective feedback information of a patient is collected in a VR scene, subjective indexes of comfort, aching feeling and fatigue are associated with physiological and behavior data, analysis weights of corresponding data segments are adjusted according to feedback, and an individualized rehabilitation data processing strategy of the patient is formed based on historical feedback data.

Description

Technical Field

[0001] This invention relates to acupuncture rehabilitation combining VR and digital therapy, specifically a method for processing acupuncture rehabilitation training data based on the combination of VR and digital therapy. Background Technology

[0002] Existing technologies, such as Chinese patent CN116313029A, a method, system, and device for dynamic control optimization of digital acupuncture, although introducing digital control and multi-dimensional feedback monitoring mechanisms on the basis of traditional acupuncture, improving the intelligence level and dynamic optimization capability of the acupuncture process, still have several significant shortcomings and drawbacks, as follows: First, in terms of interaction, the solution described in this patent mainly focuses on the automatic control and feedback optimization of the acupuncture process itself. Its interaction form is still limited to the traditional physiological detection and parameter adjustment of the acupuncture environment, lacking digital therapy scenarios related to the patient's active participation, such as task-driven interaction and behavioral response training in virtual reality (VR).

[0003] Secondly, regarding the granularity and processing strategies of feedback analysis, while this patent achieves real-time acquisition of multi-body acupoint recognition and physiological and psychological feedback, it lacks a refined subjective feedback modeling mechanism. In actual rehabilitation, the patient's subjective feelings have a significant impact on the sustainability of training effects, but this comparative technology does not systematically construct a mapping relationship between subjective feedback and objective physiological data. Furthermore, regarding conflict handling and temporal mapping mechanisms, although the patent's control logic encompasses dynamic planning and strategy generation based on feedback data, it lacks a systematic processing logic for temporal conflicts between acupuncture operations and other synchronous behaviors. For example, in actual rehabilitation training, acupuncture stimulation may cause neural overload during high-load phases of VR tasks, thereby masking or distorting feedback signals.

[0004] Furthermore, the patent does not provide a mechanism for evaluating the consistency trend of training data and conflict propagation. Its optimization path is mainly adjusted based on the current feedback state, lacking the ability to evaluate the propagation of data windows before and after in time series. Finally, from the perspective of individualized modeling and long-term training adaptation, although the patent covers "dynamic programming and multi-dimensional strategy generation," its core is still based on the local optimal adjustment of static control parameters.

[0005] In summary, although existing technologies have certain technological advantages in optimizing digital acupuncture, they also have significant shortcomings in terms of rehabilitation perspective, neural response linkage, subjective participation mechanisms, conflict regulation strategies, and individualized model construction. Summary of the Invention

[0006] The purpose of this invention is to provide a method for processing acupuncture rehabilitation training data based on the combination of VR and digital therapy, thereby solving some of the drawbacks and shortcomings pointed out in the background art.

[0007] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows: A method for processing acupuncture rehabilitation training data based on the combination of VR and digital therapy, including: acquiring acupuncture stimulation signals and VR training interaction signals, and synchronizing the two based on a unified time reference; when there is a conflict between acupuncture operation and virtual training task, dynamically determining the signal processing priority according to the patient's status, and backtracking and calibrating the signals with deviation after the acquisition is completed.

[0008] Multidimensional physiological data of patients were collected during acupuncture stimulation. Low-amplitude, short-term local physiological signals were enhanced. Delayed neural or muscular responses after acupuncture stimulation were detected by continuous sampling. Irrelevant noise was eliminated by cross-validation with VR training performance to obtain physiological feedback data induced by acupuncture stimulation. In VR rehabilitation training, the rhythm, intensity and completion of patients' task movements were monitored in real time. Data on abnormal behaviors were collected and recorded. The abnormal behaviors were analyzed by combining acupuncture stimulation parameters and time series information to determine whether they were caused by acupuncture stimulation. Data confirmed to be interference from acupuncture effects were marked as independent samples.

[0009] In a VR environment, real-time subjective feedback from patients is collected. Subjective indicators of comfort, pain, and fatigue are correlated with physiological and behavioral data. The analysis weights of corresponding data segments are adjusted based on the feedback, and individualized rehabilitation data processing strategies for patients are formed based on historical feedback data.

[0010] Furthermore, the synchronization between the acupuncture stimulation signal and the VR interaction signal is achieved through multi-level timestamp comparison, including the original acquisition time, signal transmission time, and VR task rendering completion time; when there is a conflict between acupuncture operation and VR training task, the patient's neural load index is calculated in real time and signals affecting neural response are processed first.

[0011] Furthermore, the multidimensional physiological data includes at least three of the following: electroencephalography (EEG), electromyography (EMG), electroskin conductance (ESC), heart rate variability, respiratory rhythm, and local skin temperature. These data are collected synchronously during the VR interactive task to reflect the combined state of nerves and movement. The delayed nerve or muscle response detection after acupuncture stimulation is achieved by setting multiple continuous time windows, and direct and indirect responses are classified according to the peak position of the signal response in different windows.

[0012] Furthermore, the subjective feedback is collected through a combination of three methods in the VR scene: voice input, gesture selection, and eye tracking, and different credibility levels are assigned according to the collection method. When there is a significant conflict between the patient's subjective pain score and physiological signals or VR task performance, the system adjusts the analysis weight of the data segment.

[0013] Furthermore, the subjective feedback acquisition includes three feedback methods: voice, gesture, and eye movement signals. After each acquisition, a time correspondence matrix is ​​constructed between the acquisition and physiological and behavioral data to detect the consistency level of responses between feedback sources in real time and dynamically adjust the credibility level accordingly. When any two of the three feedback methods conflict in level, the system executes a feedback divergence diagnosis protocol. The divergence diagnosis protocol determines to retain the main channel of subjective feedback and block deviation feedback signals based on the stability weight and neural response coordination degree in past feedback performance.

[0014] Specifically, the method includes the following steps:

[0015] S1. Synchronous Acquisition of Subjective Feedback: The system collects voice feedback S while the patient is undergoing VR rehabilitation training. v Gesture feedback h With eye-tracking feedback S e The system records the acquisition times of the three data sources, along with real-time physiological data P(t) and behavioral data B(t), and constructs a unified time alignment matrix T for multi-source feedback. map ;

[0016] S2. Constructing a response consistency level mapping function: Based on each feedback acquisition, the system calculates the degree of response consistency between the three types of subjective feedback and physiological and behavioral data, and adjusts the feedback credibility level through a high-order nonlinear function:

[0017]

[0018] in:

[0019] Ψ sync (t) is the feedback consistency strength function at time point t (the higher the value, the more consistent the feedback); Δ is the half-width of the time window for feedback analysis, which determines the synchronization alignment time range. The derivative of voice feedback and the modulation coefficient of physiological rhythm β p Coupled expression; The acceleration and behavioral response modulation coefficient γ for gesture feedback b Coupling term; φ e (τ) is the nonlinear velocity expression for eye-tracking feedback (which can be the variable term of the eye-tracking trajectory); η(τ) is the instantaneous phase perturbation factor of the neurophysiological response; The generalized divergence of the feedback coupling term is calculated to reflect the distribution density of the feedback field response; ∈ represents a stabilizing term introduced to avoid division by zero.

[0020] When the credibility level of any two of the above three subjective feedbacks is L i ,L j Significant differences exist (e.g., |L) i -L j |>δconflict The system initiates a feedback divergence diagnosis protocol: the system extracts the stability weight W for each feedback channel from historical feedback. i (This represents the average consistency between the feedback from this channel and historical physiological behavior data); combined with the degree of neural response coordination C during the current training process. i (t) (refers to whether the feedback behavior is accompanied by the expected electromyographic or electroencephalographic activation response); this determines which feedback channel to retain as the main channel, and the remaining feedback signals will be temporarily blocked or downweighted.

[0021] The above Ψ sync (t) The derivation process includes:

[0022] In virtual reality rehabilitation scenarios, user subjective feedback is typically delivered via voice (S). v ), gestures (S) h ) and eye movements (S e These three forms exist, and to form a unified and reliable analytical channel, the consistency of responses among these three forms with physiological / behavioral data must be evaluated. Considering the different forms of feedback, speech signals often reflect changes in emotional rhythm, and the speed of emotional outburst can be expressed by its first derivative; therefore, its derivative is defined as... To couple this to the patient's physiological state, a circadian rhythm modulation factor β is introduced. p (τ), which is usually related to heart rate, respiratory rhythm, etc., makes the harmonic expression of the speech feedback signal as Gesture feedback often manifests as abrupt changes in rhythm during muscle movement; therefore, its second derivative (i.e., acceleration) is chosen. And multiplied by the behavior modulation factor γ b (τ), forming Eye-tracking feedback often involves complex trajectory changes, therefore a nonlinear expression φ for its velocity changes is introduced. e (τ), and further coupled with the transient perturbation factor η(τ) of the neural response to construct eye movement-related physiological behavior coupling terms;

[0023] To describe the overall synchronicity of the three components under multi-channel interaction, a total analysis of all feedback coupling terms is performed within the time window [t-Δ, t+Δ], and a generalized divergence operator is introduced. To measure the response density flow of the three types of feedback in the local time field, i.e., the generation Ψ sync (t), the physical and physiological significance of this formula is that the stronger the synchronicity between the three types of feedback channels and the physiological / behavioral data, the smoother the local "density change" of the feedback channel, the smaller its generalized divergence value, and consequently the lower the integral result, ultimately leading to Ψ syncA higher value for Ψ indicates stronger consistency; conversely, if the three signals diverge in direction, are not rhythmically coordinated, or have delayed feedback, the generalized divergence fluctuates drastically within the time window, increasing the integral result and thus Ψ sync (t) decreases, and the system can adjust the confidence level of the three feedback channels based on this function value; overall, this derivation nonlinearly couples the rate of change of semantics, motion and gaze, while incorporating physiological / behavioral modulation factors and neural perturbation weights, and constructs a consistency evaluation function for multimodal feedback synchronization with the divergence operator on physical quantities. It also has continuous differentiability and integrability, and is suitable for dynamic assessment and subjective feedback discrepancy screening in rehabilitation training.

[0024] Furthermore, if the deviation between the results given by the three feedback methods exceeds a preset threshold, heterogeneous distribution determination is enabled, and specific matching analysis is performed on the physiological response channels corresponding to each type of feedback signal to verify whether the feedback distribution is reasonable.

[0025] Furthermore, when the feedback signal conflicts with physiological or VR behavioral data, the analysis weight of the current data segment is adjusted, and the conflict propagation assessment is performed on adjacent feedback windows. If a trend deviation is found, the basic analysis strategy of adjacent windows is adjusted synchronously.

[0026] Furthermore, before conducting the conflict propagation assessment of the adjacent feedback windows, cross-consistency verification is performed using data from two or more non-adjacent windows that are separated from the current window. If it is found that the current conflict is highly consistent with the trend of moving away from the window, the influence coefficient of the adjacent window propagation adjustment is increased.

[0027] Furthermore, in the conflict propagation assessment, the affected adjacent window range is expanded or reduced according to the severity of the conflict. Mild conflicts only adjust a single window, moderate conflicts adjust both windows simultaneously, and severe conflicts are expanded to the window level across task cycles.

[0028] The beneficial effects of this invention are as follows: By constructing a unified time reference, the synchronous management of acupuncture operation signals and VR interaction signals is achieved, solving the signal confusion problem caused by the interleaving of acupuncture and training movements in traditional rehabilitation. This method monitors the patient's neural load index in real time, dynamically determines processing priorities when signals conflict, ensures that key neural stimulation information is not obscured by behavioral data, and performs retrospective calibration after data acquisition, effectively improving the accuracy and controllability of the coordination between acupuncture and training tasks.

[0029] A joint verification mechanism combining multidimensional physiological responses and subjective feedback is introduced, cross-validating signals such as EEG, EMG, and TENS with VR task performance. Combining three types of subjective feedback—voice, gesture, and eye movement—a time-correspondence matrix and a discrepancy diagnosis protocol are constructed to dynamically assess the consistency and credibility of feedback. When a conflict is detected between feedback and physiological behavior data, window-level weight adjustment and trend propagation control are performed, improving the system's ability to identify abnormal signals and generate individualized strategies. This enables more accurate and intelligent tracking of rehabilitation effects and intervention feedback. Attached Figure Description

[0030] Figure 1 This is a diagram showing the data processing function relationship between VR and acupuncture signals in this invention.

[0031] Figure 2 This is a simplified flowchart of the acupuncture-VR signal collaborative processing of the present invention.

[0032] Figure 3 This is a simplified functional relationship diagram of the integration of subjective feedback and objective data and conflict handling in this invention.

[0033] Figure 4 This is a simplified schematic diagram of the collaborative fusion and feedback processing of acupuncture and VR training signals in Embodiment 1 of the present invention.

[0034] Figure 5 This is a schematic diagram of multimodal feedback collaborative analysis and conflict handling for acupuncture + VR rehabilitation in Embodiment 2 of the present invention. Detailed Implementation

[0035] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0036] Combined with appendix Figure 1This invention integrates virtual reality (VR) interactive tasks with actual acupuncture treatment during acupuncture rehabilitation training. To achieve high synergy between the two at the data acquisition and response analysis levels, the system sets a unified time reference, using the system's internal high-precision timestamps as a basis to synchronously correlate the acquired acupuncture stimulation signals with VR training interaction signals. The acupuncture stimulation signals include, but are not limited to, parameters such as acupuncture depth, stimulation frequency, and technique variations, while the VR interaction signals include the task command trigger time, visual / auditory feedback timing, and patient action response information. During the acquisition process, all signals are accompanied by the original acquisition time, transmission delay, and processing / rendering completion time. The system uses a multi-level timestamp mapping mechanism to precisely calibrate signals from different sources to the same reference time axis and constructs a signal timing alignment table. When the system detects... There is an overlap and conflict in execution time between acupuncture operations and VR training tasks. For example, when acupuncture operations temporarily limit the patient's motor ability and prevent them from completing VR training tasks, the system calculates the patient's current state parameters in real time, including the neural load index, physiological fluctuation coefficient, and degree of lag in action response. The system dynamically adjusts the signal processing priority through preset priority judgment rules. If it is determined that the patient is currently in a high-load state caused by acupuncture, the system prioritizes recording and parsing acupuncture-related signals, while temporarily suspending or downgrading VR task response signals. In addition, after the entire training and treatment process is completed, the system performs backtracking comparison on all signals to identify small time offsets caused by factors such as neural modulation delay, asynchronous acquisition equipment, or task response drift. The system performs time-series backtracking calibration on key stimulation points and interaction nodes to ensure accurate alignment of data from different modalities in the signal chain.

[0037] During acupuncture stimulation, the system simultaneously collects multidimensional physiological data from the patient via multi-channel sensors. This data includes, but is not limited to, electroencephalography (EEG), electromyography (EMG), skin conductance response (GSR), heart rate variability (HRV), local skin temperature, and respiratory rhythm. At least three of these signals are collected to comprehensively reflect the patient's neurological and physical state during acupuncture stimulation. In the data processing stage, the system prioritizes identifying low-amplitude, short-duration, continuous local fluctuation signals. These signals are often caused by local nerve reflexes near the acupuncture point, and traditional processing methods easily misinterpret them as noise or background disturbances. Therefore, a local signal enhancement mechanism is introduced. Within a narrow time window, the system dynamically amplifies the amplitude of weak physiological electrical signals and adaptively suppresses them using a noise floor threshold, achieving boundary enhancement without introducing spurious signals. The system then tracks and analyzes the enhanced signals across multiple continuous time windows to identify delayed neural or muscular responses following acupuncture stimulation. These delayed responses typically occur between hundreds of milliseconds and several seconds after stimulation. The system classifies direct and indirect responses based on the location, peak variation, and duration of different response signals within the window. Furthermore, to ensure that the extracted physiological feedback is indeed triggered by acupuncture, the system integrates these response results with VR... The behavioral performance during training is cross-validated, including dynamic indicators such as changes in the patient's movement rhythm, fluctuations in muscle coordination, or delayed responses during the training task. If the change in the peak physiological signal is highly correlated with a certain behavioral disturbance during the training task, the feedback is determined to have a causal attribute of acupuncture intervention. Based on this, the system removes pseudo-response data such as motion noise and environmental interference signals that are irrelevant to the training task, and finally extracts the target physiological feedback data with a strong causal correlation with acupuncture stimulation. At the same time, in the VR rehabilitation training, the system continuously monitors the patient's movement completion in real time, including multiple dimensions such as task rhythm consistency, movement strength stability, and task response delay. If... If the system detects abnormal fluctuations in the behavioral data, such as failed movements, sudden changes in muscle exertion, or rhythm deviations, it records the relevant data and marks the abnormal behavioral events. Subsequently, it performs correlation analysis by combining the acupuncture stimulation parameters (such as stimulation intensity, frequency, and location) at the time of the abnormality with the corresponding time-series information to determine whether the behavioral abnormality can be explained by the concurrent acupuncture operation. If a correlation is confirmed, the abnormal behavioral data is marked as an "acupuncture effect interference" sample and independently separated from the regular training behavioral data. This sample is used to construct a feature database of the impact of acupuncture stimulation on motor behavior, thereby improving the system's ability to respond to and analyze intervention effects and the accuracy of behavior classification.

[0038] The system collects real-time subjective feedback information from patients in a virtual reality (VR) rehabilitation training scenario through a multimodal input interface. This feedback information includes the patient's subjective rating or perceptual response to comfort, pain, and fatigue in the current training state. Feedback collection methods include non-contact human-computer interaction methods such as voice input, gesture selection, and eye-tracking control. The system presets confidence level weights for different feedback channels. After collection, the system performs time matching and data fusion with physiological signals (such as EEG, EMG, and heart rate variability) and behavioral data (such as task completion speed, movement amplitude, and stability) acquired synchronously at that moment. A correlation scoring matrix is ​​constructed to evaluate the consistency between subjective indicators and objective responses. If the subjective feedback and physiological behavioral data are consistent, the system maintains the original analysis weights for evaluation; if there is a significant deviation, the system adjusts the weights of the current data segment. The system dynamically adjusts key factors. For example, when a patient's subjective feedback of pain increases significantly but physiological signals do not fluctuate drastically, the system lowers the assessment credibility of that physiological data segment while increasing the reference weight of the subjective channel in the comprehensive evaluation, thereby improving the system's sensitivity to latent symptoms. In addition, after each round of feedback collection and analysis, the system continuously updates the historical feedback characteristic curves of individual patients and archives them with the stimulus parameters, task content, and physiological response records of the corresponding training stage. This gradually establishes a model of individualized response preferences and tolerance intervals for patients. In subsequent training, the system uses this individual model to perform contextual reasoning and trend comparison on new feedback data, achieving early identification of abnormal feedback and adaptive correction of analysis weights. Thus, through multiple rounds of training iterations, the system gradually optimizes the rehabilitation data processing strategy and improves the system's ability to accurately adapt to individual patient states and the depth of feedback understanding.

[0039] Combined with appendix Figure 2To achieve high-precision synchronization between acupuncture stimulation signals and virtual reality (VR) training interaction signals, the system employs a multi-level timestamp comparison mechanism as the basis for signal association. Specifically, this includes independently recording the original acquisition timestamp generated during acupuncture operation signal acquisition, the transmission timestamp of the signal from the acquisition module to the processing center, and the rendering completion timestamps corresponding to each key node in the VR training task execution process (e.g., task instruction issuance, scene rendering completion, and action recognition initiation). The system compares and uniformly converts these timestamps by constructing a three-dimensional time mapping table, thereby achieving time synchronization and logical alignment of multi-source asynchronous signals. This enables subsequent data processing to dynamically fuse and determine causality based on a unified timeline. When acupuncture operations and VR training tasks conflict in time, for example, when a certain stimulation event corresponding to an acupuncture operation conflicts with the VR training task... When a key task action overlaps or interferes with each other, the system immediately invokes the patient's current state perception module based on preset rules to acquire multidimensional physiological indicators in real time, including EEG stability, electromyography amplitude change rate, and heart rate fluctuation coefficient. The system calculates the neural load index at the current moment, which comprehensively assesses whether the patient's nervous system is in an activated, inhibited, or easily fatigued state. If the neural load index exceeds the threshold, it indicates that the patient is in a state of high sensitivity to acupuncture or has a heavy cognitive processing burden. The system then prioritizes processing signal sources strongly correlated with this state, that is, it prioritizes retaining and analyzing acupuncture stimulation-related data. At the same time, it delays, downweights, or suspends the acquisition of VR task response signals to avoid confusion of physiological responses caused by training action interference, improves the cleanliness of data analysis and the accuracy of causal determination, thereby ensuring that the data processing logic conforms to the actual changes in the patient's state.

[0040] To comprehensively reflect the neuro-motor coupling state of patients during acupuncture rehabilitation training, the system synchronously collects multidimensional physiological data during VR interactive tasks. This physiological data includes at least three of the following six categories: electroencephalography (EEG), electromyography (EMG), skin conductance (GSR), heart rate variability (HRV), respiratory rhythm, and local skin temperature. EEG reflects central nervous system activity and cognitive load levels; EMG monitors muscle group activity characteristics to identify muscle response pathways; GSR reflects changes in sympathetic nerve excitability; heart rate variability and respiratory rhythm characterize the balance state of the autonomic nervous system; and local skin temperature, as an auxiliary indicator of blood circulation regulation, monitors temperature trends at the acupuncture site. All these multi-channel signals are synchronously sampled using a unified time reference to ensure cross-modal signal alignment consistency in the temporal dimension. During acupuncture stimulation... After the event occurs, the system sets multiple consecutive physiological response time windows based on the stimulus type and location. Each window has a fixed width but allows dynamic sliding to cover different response delay ranges. During signal processing, the system extracts the response peak, rise rate, and duration of each channel within each time window and classifies them based on the time location and characteristic differences of the response. If the response peak appears rapidly within a short period after stimulation, it is determined to be a direct response, such as immediate muscle twitching or sudden changes in brain electrical potential. If the response appears delayed and shows a slow rise or diffusion trend, it is determined to be an indirect response, such as autonomic nerve fluctuations or secondary muscle group responses. The system correlates the classification results with the motor behavior data in the VR training task to further identify the path of acupuncture stimulation on rehabilitation training performance and provide data support for subsequent personalized intervention strategies.

[0041] The system synchronously collects patients' subjective feedback information in VR rehabilitation training scenarios, specifically through a combination of three methods: voice input, gesture selection, and eye tracking. Voice input allows patients to actively describe their current subjective experiences such as pain, fatigue, or comfort. Gesture selection allows patients to quickly make choices among different evaluation levels through a visual interactive interface. Eye tracking is used to identify the patient's gaze duration and selection intent for the feedback options. The system presets a reliability level for each of the three collection methods. Voice input is assigned the highest level due to its rich semantics and high initiative, gesture selection has a medium reliability level, and eye tracking has a relatively low reliability level due to its susceptibility to attention drift. Within each feedback collection cycle, the system uniformly maps the three types of feedback signals to the time window corresponding to the current training task and physiological state, and synchronizes them with the EEG and myocardial infarction data collected during that time period. Physiological data such as electrophysiology and skin conductance are correlated and compared with task completion data such as movement precision and rhythm consistency. When there is a significant conflict between the patient's subjective pain score and physiological signals or training performance, such as increased pain but stable physiological signals or no decline in task completion quality, the system judges that there are factors such as errors, subjective amplification, or cognitive misjudgment in the current feedback. In order to avoid such conflicts from misleading the evaluation of training effect, the system triggers the analysis weight adjustment mechanism, which reduces the proportion of subjective feedback channels in the analysis of the current data segment, and dynamically adjusts its influence coefficient according to the credibility of the feedback source. This ensures that the overall data analysis results can more realistically reflect the patient's comprehensive status. At the same time, this weight adjustment mechanism has short-term memory and maintains reasonable sensitivity to subsequent short-term feedback, thereby achieving stable filtering and adaptive correction of abnormal fluctuations in subjective perception.

[0042] Combined with appendix Figure 3 To achieve the synergistic integration of subjective feedback and objective physiological behavior data in acupuncture rehabilitation training, a multimodal subjective feedback acquisition and processing mechanism based on VR and digital therapy was systematically constructed, specifically including voice input S v Gesture selection S h and eye-tracking trajectory S e Three feedback methods are used, and after each data collection, they are aligned with the physiological data P(t) and behavioral data B(t) using timestamps to construct a unified time mapping matrix T for multi-source feedback. map To dynamically evaluate the reliable consistency of each feedback channel, a high-order nonlinear response consistency strength function Ψ was designed for the system. sync (t), the specific calculation formula is as follows:

[0043]

[0044] in:

[0045] Ψ(t): Represents the consistency strength between subjective feedback and objective response at time point t. The larger the value, the higher the synchronization consistency of the three types of feedback.

[0046] Δ: is the half-width of the feedback analysis time window, controlling the alignment range of the feedback response;

[0047] The derivative of the speech feedback is coupled with the physiological modulation factor, where β p (τ) represents the modulation weight of the resonance between the heart rate or respiratory rhythm and the speech rhythm.

[0048] The acceleration and behavioral rhythm coupling term of gesture feedback, γ b (τ) is the motion consistency modulation coefficient;

[0049] φ e (τ): represents the nonlinear velocity change of the eye-tracking signal, extracted in conjunction with the temporal fluctuation features of visual gaze;

[0050] η(τ): is the transient phase perturbation term of the neural response channel, such as transient interference from sudden changes in local electromyography or electroencephalography peaks;

[0051] The generalized divergence operator is used to measure the intensity of the response density variation of a feedback channel in local time and space.

[0052] ∈: A small, stable term set up to prevent division by zero errors during integration.

[0053] Based on the above definition, the system uses Ψ sync (t) is an evaluation metric used to dynamically adjust the credibility level of various feedback channels.

[0054] And when a significant difference in the confidence level is detected between any two types of feedback channels, i.e.:

[0055] |L i -L j |>δ conflict

[0056] Where L i ,L j These represent the confidence levels of feedback channels i and j in the current sampling period, δ conflict If a conflict judgment threshold is set for the system, the feedback divergence diagnosis protocol will be triggered.

[0057] During this diagnostic process, the system first extracts the historical stability weights W of each feedback channel. iThis value represents the average consistency between the channel feedback results and the physiological / behavioral data over multiple past task cycles; subsequently, the degree of synergy C of the neural responses corresponding to each feedback channel in the current training state is calculated. i (t), such as whether voice feedback is accompanied by activation of specific frequency bands in the central cortex, and whether gesture feedback triggers synchronous fluctuations in the electromyographic peaks of the target muscle group, etc. The final system is comprehensively evaluated based on the following weighting function:

[0058] Score i =α·W i +(1-α)·C i (t)

[0059] Where α∈[0,1] is the weight allocation coefficient, and the system adjusts the priority according to the current task stage. The feedback channel with the highest score will be determined as the main channel, and the other feedback signals will be marked as "minor" or "masked" according to the degree of conflict, and will not participate in the key feature extraction of the current analysis window.

[0060] To enhance the consistency between subjective feedback and physiological response in acupuncture rehabilitation training based on VR and digital therapy, after collecting three types of subjective feedback signals—voice, gesture, and eye movement—the system first calculates the deviation between the three types of feedback results in each training cycle. This includes situations such as differences in score values, contradictory selection intentions, or inconsistencies between gaze behavior and the semantic direction of feedback. A preset deviation threshold δ is then defined. deviation When the system detects that the deviation between any two feedback methods exceeds this threshold, i.e., |R| is satisfied... i -R j |>δ deviation If a certain feedback signal is detected, the "Heterogeneous Distribution Judgment" module is triggered. This module analyzes each type of feedback signal based on the principle of specific matching between feedback and physiological channels to determine whether it can effectively couple with the physiological response channel it should trigger in terms of time and response mode. Specifically, the voice feedback signal is correlated with the fluctuation of the cortical potential of the central EEG, the gesture feedback signal is matched with the electromyographic response of the target muscle group, and the eye movement feedback is compared with the eye movement velocity trajectory and the visual stimulus sequence to determine the fixation focus. Furthermore, autonomic nerve indicators such as skin conductance changes and heart rate variability are introduced as auxiliary signal sources to determine whether each feedback signal has "physiological response supporting evidence". If a certain type of feedback signal does not detect a significant response in its expected physiological channel or the response delay exceeds the tolerance window, the feedback distribution is determined to be unreasonable and caused by false response, distraction, or abnormal state. The system marks the feedback signal as "heterogeneous" or "low confidence" and reduces its weight or removes it from the current data segment, thereby ensuring a reasonable mapping relationship between subjective feedback information and physiological response.

[0061] To address the issue of conflicts between subjective feedback signals collected during acupuncture stimulation and VR training and physiological data (such as EEG, EMG, and TENS) or VR behavioral data (such as action completion, interaction latency, and training trajectory) at certain times, the system first constructs a feedback consistency detection module for each data acquisition time window. When the system detects inconsistencies between the trend of subjective feedback signals (such as those indicating pain or discomfort) and changes in objective physiological indicators or behavioral performance within a certain feedback window (e.g., high pain feedback but stable EMG response and VR performance), the system marks this window as a "feedback-behavior conflict state." Subsequently, it dynamically adjusts the weighting coefficient w of the data from this window in the overall rehabilitation analysis. t This reduces the contribution of isolated abnormal feedback to the individual recovery assessment model, avoiding bias in rehabilitation decisions. Simultaneously, to prevent such conflicting fragments from affecting the continuity of analysis, the system further activates the "conflict propagation assessment" process. This process performs trend deviation analysis on adjacent windows (e.g., ±2 time windows) before and after the current feedback window. Specifically, it determines whether the feedback, physiological, and behavioral data within these adjacent windows exhibit similar shifts or signal inconsistencies. If a cross-window extension trend is detected, meaning adjacent windows also show a mismatch between subjective feedback and objective data, the system will adjust the basic analysis strategies for these windows based on the strength of the trend extension. This includes reducing weights, delaying response assessments, and reconstructing behavioral mapping paths. This ensures the continuity, robustness, and multi-channel consistency of data processing, preventing local feedback anomalies from misleading the overall analysis.

[0062] To address the handling mechanism when subjective feedback signals conflict with physiological or VR behavioral data during acupuncture rehabilitation training, this paper further proposes that before performing conflict propagation assessment between adjacent feedback windows, the system first calls a dataset from non-adjacent windows (i.e., a time period spanning two or more windows) that is separated from the currently conflicting data window to construct a cross-temporal consistency verification module. The system compares the deviation characteristics of subjective feedback within the current window (such as abrupt changes in feedback level or significant discrepancies with objective data) with the trend of the feedback-physiological / behavioral data relationship in non-adjacent windows to determine whether the conflict exhibits a periodic or systematic anomaly, for example, across multiple window intervals. If a similar type of feedback conflict occurs again, and the analysis results show that the conflict characteristics of the current window are highly consistent with the conflict trend in windows far away (such as t-3 or t+4), the system determines that the current feedback anomaly is not an occasional interference, but an effective signal disturbance with a continuous or periodic behavioral background. Therefore, the system increases the influence coefficient in the current adjacent window propagation adjustment mechanism, that is, it implements a greater synchronous adjustment on the processing strategies such as feedback weights, physiological response interpretation, and behavioral data mapping relationships involved in the data windows on the left and right sides of the current window, to ensure that, under the premise that the conflict trend exists across time periods, adjacent windows can respond more sensitively to the potential impact of conflict spread.

[0063] To address the conflict between subjective feedback signals and objective physiological or behavioral data that arises during acupuncture rehabilitation training combining VR and digital therapy, the system employs a propagation assessment mechanism based on the severity of the conflict. This mechanism dynamically adjusts the range of data windows affected during data processing by determining the conflict level. The severity of the conflict is determined based on a comprehensive score, including but not limited to the degree of deviation between feedback signals and physiological signals, the duration of the conflict, the magnitude of deviation from historical feedback trends, and the actual interference with the accuracy of rehabilitation task execution. When the conflict is determined to be mild (e.g., a brief, slight deviation between feedback and physiological signals, or the appearance of localized abnormal behavior with no substantial impact), the system only processes data within the data window where the current conflict occurs. The system performs local parameter adjustments, including fine-tuning of analysis weights and correction of feedback credibility levels. When a conflict is determined to be moderate (e.g., the deviation lasts for a long time or the interference range extends to multiple physiological dimensions), the system synchronously adjusts the data processing strategies of adjacent windows on both sides of the conflict window, forming a local segment-level correction chain to prevent the impact of the conflict from spreading laggingly in a short period of time. When a conflict is determined to be severe (e.g., feedback information is significantly abnormal, accompanied by strong physiological inconsistencies, or a sharp reversal of the trend direction of multiple historical windows), the system triggers a cross-cycle propagation processing mechanism, expanding the affected range to cover multiple windows before and after the complete rehabilitation task cycle, and even re-analyzing the feedback-physiological coupling logic and data validity of the task cycle, forming a task-level conflict reconstruction and adjustment.

[0064] Example 1:

[0065] Combined with appendix Figure 4 In this embodiment, the patient, Mr. Zhang, a 45-year-old male, is undergoing rehabilitation treatment based on the system of this invention due to decreased muscle function in his left upper limb following a stroke. During the rehabilitation process, the system combines acupuncture with VR motion tasks for synergistic intervention. The treatment plan includes: the doctor performing electroacupuncture stimulation on the "Quchi" acupoint on the patient's left upper limb every 15 minutes, while the patient completes a "virtual grasp-lift-release" training task in a VR scene, with each round lasting approximately 120 seconds.

[0066] In the 5th round of training, the VR task started at:

[0067] T VR,start =215.340s

[0068] The start time of acupuncture electrical stimulation is recorded by the acquisition module as follows:

[0069] T acupuncture,raw =216.078s

[0070] Since the signal transmission delay and rendering delay are respectively:

[0071] ΔT transmission =0.065s,ΔR render =0.120s

[0072] The system calculates the acupuncture signal timing after uniform alignment as follows:

[0073] T acupuncture,aligned =T acupuncture,raw +ΔT transmission +ΔT render =216.263s

[0074] The current rendering frame completion time for the VR task is:

[0075] T VR,frame =215.980s

[0076] Because the time difference between the two is:

[0077] ΔT=|T acupuncture,aligned -T VR,frame |=0.283s

[0078] The system determines that there is a synchronization offset conflict in this round of interaction and enters the neural load index judgment mechanism.

[0079] Example of neural load index calculation and priority decision-making:

[0080] The system collected Zhang's current EEG alpha wave, electromyographic frequency, and heart rate variability (HRV), with the following parameters:

[0081] EEG alpha wave intensity: A α =37.5μV

[0082] Electromyographic activity frequency: f EMG =23Hz

[0083] HRV standard deviation: σ HRV =41ms

[0084] The neural load index is defined as follows:

[0085]

[0086] Setting initial baseline data for patients:

[0087]

[0088] f baseline =18Hz;

[0089] σ baseline =55ms;

[0090] Let the weighting factor be:

[0091] ω1=0.4,ω2=0.35,ω3=0.25

[0092] Substitute into the calculation:

[0093]

[0094] The system sets the neural load threshold to 0.9, and the current load index is... The neural response is determined to be under high load, and the acupuncture signal should be processed first.

[0095] The system pauses the current VR task action trigger response (such as grasping judgment), freezes the VR feedback frame update for a short time, and prioritizes the inclusion of acupuncture electrical stimulation signals into the physiological analysis process. After synchronously recognizing EEG and EMG, it extracts their induced response features. Subsequently, the system performs VR scene task progress backtracking for 0.3 seconds to ensure complete alignment of interactive data.

[0096] Further, the system incorporates simultaneous acquisition of multidimensional physiological data and delayed neural response detection, and combines three subjective feedback mechanisms—voice, gesture, and eye movement—to achieve dynamic correction of data processing. In the sixth round of training, after Zhang received acupuncture stimulation at the "Quchi" acupoint, the system simultaneously acquired EEG, EMG, and EDA signals. The time windows were set as three consecutive windows, W1, W2, and W3, within 0–5 seconds after stimulation, each lasting 1.5 seconds. In W1, the system detected a significant increase in the theta wave power in the right prefrontal EEG band (an increase of 42%, with a baseline power of 5.2 μV). 2 The current amplitude is 7.4 μV2. EMG detected a muscle twitching peak in the left arm flexor muscles at 0.7 s (amplitude 0.95 mV, baseline 0.48 mV). The EDA signal showed a skin conductance spike in phase W2 (jumping from 0.35 μS to 0.62 μS, exceeding the background noise by 2.1 times). Based on these characteristics, the system labeled the EEG and EMG signal peaks in W1 as direct responses, and the EDA delay changes in W2 as indirect responses. The system thus determined that the current stimulus elicited a central-peripheral combined response within 1 second, strengthening the weight of this data segment in subsequent feature analysis.

[0097] Meanwhile, during this round of VR training, Zhang exhibited delayed grasping movements accompanied by saccades in both eyes, triggering a "pain level" feedback request within the VR scene. Zhang responded via voice, stating that "the stinging sensation is increasing." The system recorded the rhythmic acceleration and intonation enhancement in his audio, and assigned a subjective pain score of 7 (out of 10). However, during this period, EMG signals showed normal fluctuations in muscle electrical activity (no sudden high-frequency muscle twitching or jerking). Although the hand gestures in the behavioral data showed slight swaying, they did not reach the set threshold. Based on this, the system determined that there was a significant deviation between the subjective score and the objective physiological and behavioral data. Due to the high reliability of the voice feedback (a confidence score of 0.82 given by the intonation recognition module), the system did not directly reject the score but reduced the weight coefficient of this data segment from the standard weight of 1.0 to 0.65, while simultaneously triggering parallel verification of eye movement trajectory and hand gesture feedback. The results showed that there was a 1.2-second continuous saccade with multiple momentary gaze interruptions in the eye movement indicators. The system increased the reliability level of the eye movement feedback and retained the score at the medium level of 6, forming a fusion-corrected score result. In subsequent rounds of training, based on Zhang's feedback of frequent fluctuations, the system gradually collected and formed his individualized subjective bias feature profile, and dynamically matched his historical response stability weights during the analysis process.

[0098] Example 2:

[0099] Combined with appendix Figure 5 In this embodiment, patient Wang, 50 years old, suffers from spinal problems due to prolonged sitting at work and also has mild nerve damage. He is undergoing rehabilitation training based on the acupuncture and VR combination of this invention. Wang's treatment plan includes: the doctor using acupuncture to stimulate the "Jianjing" acupoint on his shoulder with electroacupuncture; simultaneously, the patient performs a grasping and releasing task in a VR training scenario. This training task requires the patient to complete the grasping action through visual and gesture interaction, and each training cycle is 120 seconds.

[0100] S1. Subjective feedback is collected synchronously:

[0101] During Wang's VR rehabilitation training, the system simultaneously collected Wang's subjective feedback signals through a voice recognition module, gesture capture, and eye tracking. At the start of the training, Wang inputted "slight shoulder pain" via voice, and the system recorded the voice signal S. v The timestamp is T voice =120.350s. Simultaneously, the system detected Wang's grasping motion via the gesture capture module, and the gesture signal S... h The timestamp is T gesture =120.375s, the eye-tracking module recorded Wang's eye movement trajectory and generated an eye movement signal S. e The timestamp is T eye= 120.355s. All these signals are time-aligned based on a unified time reference standard (e.g., the system's internal clock), constructing a time correspondence matrix T for the three types of feedback signals. map :

[0102]

[0103] This matrix ensures that different feedback channels collect and process data synchronously within the same time window, laying the foundation for subsequent feedback consistency assessment.

[0104] S2. Construct the response consistency level mapping function:

[0105] The system is based on three types of collected feedback signals (speech S). v Gesture S h Eye movement S e The feedback consistency strength function Ψ is calculated by comparing the response consistency between the data and real-time physiological data (such as EEG, EMG, and EDA). sync (t), used to dynamically adjust the confidence level of each type of feedback signal. Set within the current time window [t-Δ, t+Δ], the system calculates using the following formula:

[0106]

[0107] In this formula:

[0108] This represents the coupling relationship between the derivative of the speech feedback and the physiological rhythm modulation factor;

[0109] This indicates the coupling relationship between the acceleration of gesture feedback and the modulation factor of the behavior response;

[0110] φ e (τ) is the nonlinear velocity expression for eye-tracking feedback, which describes the change in the eye-tracking trajectory;

[0111] η(τ) is the instantaneous phase perturbation factor of the neurophysiological response;

[0112] It is a generalized divergence calculation of the feedback coupling term, reflecting the response density distribution of the feedback signal.

[0113] Set φ v (τ), φ h (τ), φ e The specific values ​​of η(τ) and η(τ) in the time window [120,121] are as follows:

[0114] (The rate of change of the speech signal);

[0115] β p (τ) = 1.2 (physiological rhythm modulation factor);

[0116] (Acceleration from gesture feedback);

[0117] γ b (τ) = 1.3 (behavioral modulation factor);

[0118] φ e (τ) = 0.32 (speed change of eye-tracking feedback);

[0119] η(τ) = 0.5 (neural perturbation factor).

[0120] By substituting these values, we can calculate:

[0121]

[0122] Further calculations yielded:

[0123]

[0124] If we set ∈ = 0.01, then:

[0125]

[0126] This function value indicates a high degree of consistency between the three types of feedback signals and the physiological data response, suggesting strong reliability. The system adjusts the weights of the feedback signals based on this value, enhancing the influence of voice feedback in the current data segment.

[0127] Feedback Disagreement Diagnosis Agreement:

[0128] During the current training process, the voice feedback L is set to... v =0.85 (credibility level), gesture feedback L h =0.92, eye-tracking feedback L e =0.72. The system judges the credibility difference between the speech and eye-tracking feedback to be greater than the preset threshold δ. conflict =0.1, that is, |L v -L e |=|0.85-0.72|=0.13>0.1, initiate the feedback discrepancy diagnosis protocol.

[0129] The system analyzes the stability weight W of historical feedback signals. v =0.78, W e =0.67 and the current neural response coordination level C v (120)=0.84, C e (120) = 0.60 for evaluation. Therefore, the system calculates the stability score of the main channel as follows:

[0130] Score v =αW v +(1-α)C v (120) = 0.5·0.78 + 0.5·0.84 = 0.81

[0131] Score e =αW e +(1-α)C e (120) = 0.5·0.67 + 0.5·0.60 = 0.635

[0132] Due to Score v =0.81>Score e =0.635, the system will provide voice feedback S v As the main channel, eye-tracking feedback S is blocked or downweighted. e This will have an impact, thus ensuring the accuracy of feedback data analysis.

[0133] During the treatment, Wang performed grasping and placement tasks using VR equipment while simultaneously receiving electroacupuncture stimulation. The system simultaneously collected three types of subjective feedback signals from Wang (voice feedback S...). v Gesture feedback h and eye-tracking feedback S e Simultaneously, multidimensional physiological data such as electroencephalography (EEG), electromyography (EMG), and electrical conductance on skin (EDA) were monitored. During a certain time period, Wang provided voice feedback stating "I feel a persistent tingling sensation in my left hand." This voice feedback signal was collected and recorded, with the timestamp T. voice = 252.745 seconds; the gesture selection was confirmed by the VR system to be 90% accurate in Wang's current movement (corresponding to the gesture signal timestamp T). gesture =252.765 seconds); Eye-tracking feedback timestamp T eye = 252.755 seconds. The system records the timestamps of these signals and constructs the corresponding time alignment matrix T. map :

[0134]

[0135] This ensures that all feedback signals are time-aligned, and the system can perform subsequent feedback consistency calculations.

[0136] In Wang's treatment, the system assessed the reliability of the subjective feedback by calculating the deviation between the feedback signals. Voice feedback S vThe system indicated a strong stinging sensation, while gesture and eye-tracking feedback showed that Wang's gestures were accurate (90% precision) and his eye-tracking trajectory followed a normal pattern. However, a discrepancy was observed between the verbal feedback and physiological signals (such as electromyography and electroencephalography). For example, the system detected:

[0137] The pain score based on voice feedback was P. v =8 (out of 10), while the electromyography (EMG) showed stable muscle electrical activity. peak =0.4mV.

[0138] The electroencephalogram (EEG) data did not show any significant abnormal fluctuations within 15 seconds after stimulation (the EEG power remained at the baseline level).

[0139] Therefore, the significant deviation δ between the pain score from voice feedback (8 / 10) and physiological data (EMG, EEG) is not significant. conflict =|8-0.4|=7.6, which exceeds the system's set deviation threshold δ. conflict =2.5, the system triggered the "Heterogeneous Distribution Judgment" module.

[0140] During the heterogeneous distribution determination process, the system examines the coupling degree between each feedback signal and physiological data based on a preset feature extraction algorithm. The system is configured to specifically match each type of feedback signal with the physiological response channel, utilizing neural response co-analysis for judgment. In Wang's feedback, the system compares the following data:

[0141] Relationship between voice feedback and EEG signals: The system found through spectrum analysis that when the voice feedback score was 8 (strong pain), the EEG signal did not show the expected increase (alpha waves did not decrease significantly).

[0142] Matching gesture feedback with electromyographic signals: Although the accuracy of the gesture task was high (90%), the amplitude of changes in electromyographic signals was small, only slightly higher than that of EMG signals. peak =0.4mV, which indicates that hand muscle activity was not significantly affected.

[0143] Therefore, the system found inconsistencies in the matching degree between feedback signals and physiological responses (especially the deviation between voice feedback and physiological data). Through specific matching analysis, the system determined that the current feedback distribution was unreasonable, especially since the pain score of voice feedback deviated significantly from the physiological data. The system then decided to reduce the weight of the voice feedback signal.

[0144] After adjusting the signal in the current feedback window, the system performs a conflict propagation assessment on adjacent windows. The system sets the time window for the current data segment to [t-Δ, t+Δ], where Δ = 1. sWithin this time period, the system analyzes whether there are consistency issues in the feedback signals between adjacent windows (e.g., W1 is from t-1 seconds to t seconds, W2 is from t seconds to t+1 seconds).

[0145] By calculating the difference in the feedback signal within two adjacent time windows, the system detected the following problem:

[0146] Window W1: The consistency between voice feedback and electromyographic signals is high, and the gesture feedback is displayed normally;

[0147] Window W2: The voice feedback showed a deviation, especially when the gesture signal and EEG signal were not synchronized. The system found that the signal in window W2 fluctuated greatly, especially the deviation between the voice feedback and the behavioral data.

[0148] Therefore, based on the calculated trend deviation, the system decides to adjust the basic analysis strategy for window W2. Specifically, the system increases the signal propagation influence coefficients of windows W1 and W2 (e.g., w1 = 1.1 and w2 = 1.3) to make more precise adjustments to the conflicts, while appropriately reducing the weight of other signal sources.

[0149] During the analysis, the system adjusts the data weights of the feedback signals in real time. The initial weight of the speech signal is set to w. voice =1.0, gesture feedback weight is W gesture =1.0, eye-tracking feedback weight is w eye =0.8, due to the deviation between speech and physiological signals, the system decides to reduce the weight of speech feedback in the current data segment:

[0150] w′ voice =0.65 (weight after deweighting)

[0151] In this way, the system avoids error propagation caused by abnormal deviations in the voice signal, thus improving the accuracy of data processing.

[0152] In Wang's 10th round of treatment, acupuncture was applied to the "Jianjing" acupoint, while the VR training task required Wang to grasp and place objects. The system, through a multi-channel data acquisition module, simultaneously recorded Wang's three subjective responses (voice, gestures, and eye movements), as well as physiological data such as electroencephalography (EEG), electromyography (EMG), and electrical conductance analysis (EDA). Wang's grasping accuracy was set at 88%, but at the 20-second mark of the training, the patient reported "some numbness in the shoulder" via voice feedback, and the voice feedback showed a pain score of 6 (out of 10). The system recorded the following feedback data:

[0153] Voice feedback S v =6, timestamp T voice = 235.480 seconds;

[0154] Gesture Feedback S h =88%, timestamp T gesture = 235.485 seconds;

[0155] Eye-tracking feedback S e =3 (indicating small changes in eye movement), timestamp T eye = 235.470 seconds.

[0156] Meanwhile, the physiological data monitoring revealed the following information:

[0157] Electroencephalogram (EEG): Brain electrical activity showed no significant changes within 5 seconds after stimulation, with small power fluctuations. α =4.2μV 2 ;

[0158] Electromyography (EMG): Muscle electrical activity manifests as slight tremors in the shoulder flexors. peak =0.45mV;

[0159] Electrodermal conductivity (EDA): The peak value of the electrodermal conductivity response is relatively low. peak =0.33μS.

[0160] Because of the significant discrepancy between Wang's voice feedback and physiological data (especially EEG and EMG), the system initiated cross-consistency verification, particularly extending the analysis of the current time window [t-Δ, t+Δ]. The system-defined time window was set to Δ = 1 second, meaning the current conflict window spanned [235.480, 235.480+1]. The system not only processed the data within the current window but also performed cross-consistency verification on data from non-adjacent windows separated by two or more. The first two adjacent windows were set as W1 = [234.430, 234.480] and W2 = [236.480, 236.530], and the system compared the data within these two windows as follows.

[0161] Window W1: Voice feedback is S v =5 (lower pain level), gesture feedback is S h =90%, and the power fluctuation of the EEG signal is relatively small;

[0162] Window W2: Voice feedback is S v =6, gesture feedback is S h =88%, and eye movement feedback and EEG signal fluctuations remained stable.

[0163] The system found that no significant abnormalities were observed between verbal feedback and physiological data in W1 and W2. However, in the current window W3 (i.e., [235.480, 235.530]), the pain score of the verbal feedback was as high as 6, which did not match the low fluctuation of the EEG signal. By comparison, the system detected that the feedback signal in the current window had low consistency with the trend away from the window, and therefore decided to correct it in the current time period.

[0164] After completing cross-consistency verification, the system begins to assess the severity of the conflict and determine how to adjust the weights of adjacent windows. The system first judges based on the degree of anomaly in the voice feedback; if the deviation exceeds a preset threshold δ... conflict If the value is 1.5, the system enters the conflict propagation assessment phase.

[0165] Based on actual data, the voice feedback score in the current window is S. v =6 showed a significant deviation, and the EEG signal did not show the expected changes. The system calculated the deviation to be:

[0166] δ feedback =|S v -EEGresponse|=|6-0.45|=5.55

[0167] Therefore, the system determines that this is a severe conflict and decides to expand the conflict assessment scope to a window level spanning multiple task cycles. That is, in addition to the current window W3, all signal processing strategies need to be adjusted, including those for multiple time periods before and after (e.g., W1, W2, W4).

[0168] Minor conflict: The system only adjusts the data weights of a single window;

[0169] Moderate conflict: The system adjusts the data weights of the two side windows;

[0170] Severe conflict: The system extends to window data processing across task cycles.

[0171] In this case, the system detected that the conflict was severe, therefore it selected to adjust the two adjacent windows W1 and W2 before and after the conflict, and then processed the subsequent window W4. The specific operations are as follows:

[0172] w1=1.2,w2=1.15,w3=0.65,w4=1.1

[0173] These adjustments will ensure that the system maintains consistency with signal data in the current time period and its adjacent time periods, and reduce the interference of conflicts on data analysis results by strengthening the correction of conflict windows.

[0174] Based on the processing in the first two steps, the system further dynamically adjusts the analysis weight of the current data segment. The initial weight of the voice feedback is set to w. voice =1.0, the weight of gesture feedback is w gesture =1.0, the weight of eye-tracking feedback is w eye =0.8, due to the deviation between voice feedback and physiological data, the system adjusts the weight of voice feedback by reducing its value.

[0175] w′ voice =0.7

[0176] Meanwhile, due to the severity of the conflict, the system appropriately increased the weight of gesture feedback and eye-tracking feedback:

[0177] w′ gesture =1.05,w′ eye =0.85

[0178] These weighting adjustments ensure that the system focuses more on the analysis of other signal sources (such as gestures and eye-tracking feedback), thereby reducing the potential misleading effect of voice feedback on the analysis results of rehabilitation training.

[0179] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A data processing method for acupuncture rehabilitation training based on the combination of VR and digital therapy, characterized in that... include: Acquire acupuncture stimulation signals and VR training interaction signals, and synchronize and correlate the two based on a unified time reference. When there is a conflict between acupuncture operation and virtual training task, dynamically determine the signal processing priority according to the patient's status, and backtrack and calibrate the signals with deviation after acquisition. Multidimensional physiological data of patients were collected during acupuncture stimulation. Low-amplitude, short-term local physiological signals were enhanced. Delayed neural or muscular responses after acupuncture stimulation were detected by continuous sampling. Irrelevant noise was eliminated by cross-validation with VR training performance to obtain physiological feedback data induced by acupuncture stimulation. In VR rehabilitation training, the rhythm, intensity and completion of patients' task movements were monitored in real time. Data on abnormal behaviors were collected and recorded. The abnormal behaviors were analyzed by combining acupuncture stimulation parameters and time series information to determine whether they were caused by acupuncture stimulation. Data confirmed to be interference from acupuncture effects were marked as independent samples. In a VR environment, real-time subjective feedback from patients is collected. Subjective indicators of comfort, pain, and fatigue are correlated with physiological and behavioral data. The analysis weights of corresponding data segments are adjusted based on the feedback, and individualized rehabilitation data processing strategies for patients are formed based on historical feedback data.

2. The acupuncture rehabilitation training data processing method based on VR and digital therapy according to claim 1, characterized in that... The synchronization between the acupuncture stimulation signal and the VR interaction signal is achieved through multi-level timestamp comparison, including the original acquisition time, signal transmission time, and VR task rendering completion time. When there is a conflict between acupuncture operation and VR training task, the patient's neural load index is calculated in real time and signals that affect neural response are processed first.

3. The acupuncture rehabilitation training data processing method based on VR and digital therapy according to claim 1, characterized in that... The multidimensional physiological data includes at least three of the following: electroencephalography (EEG), electromyography (EMG), electrodermal conductance (TEA), heart rate variability, respiratory rhythm, and local skin temperature. These data are collected synchronously during the VR interactive task to reflect the combined state of nerves and movement. The delayed nerve or muscle response detection after acupuncture stimulation is achieved by setting multiple continuous time windows, and direct and indirect responses are classified according to the peak position of the signal response in different windows.

4. The acupuncture rehabilitation training data processing method based on VR and digital therapy according to claim 1, characterized in that... The subjective feedback is collected through a combination of voice input, gesture selection, and eye tracking in the VR scene, and different credibility levels are assigned according to the collection method. When there is a significant conflict between the patient's subjective pain score and physiological signals or VR task performance, the system adjusts the analysis weight of the data segment.

5. The acupuncture rehabilitation training data processing method based on VR and digital therapy according to claim 4, characterized in that... The subjective feedback collection includes three feedback methods: voice, gesture, and eye movement signals. After each collection, a time correspondence matrix between the data and physiological and behavioral data is constructed to detect the consistency level of responses between feedback sources in real time and to dynamically adjust the credibility level accordingly. When any two of the three feedback methods conflict in level, the system executes a feedback divergence diagnosis protocol. The divergence diagnosis protocol determines to retain the main channel of subjective feedback and block deviation feedback signals based on the stability weight and neural response coordination degree in past feedback performance.

6. The acupuncture rehabilitation training data processing method based on VR and digital therapy according to claim 5, characterized in that... If the deviation between the results given by the three feedback methods exceeds a preset threshold, heterogeneous distribution determination is enabled, and specific matching analysis is performed on the physiological response channels corresponding to each type of feedback signal to verify whether the feedback distribution is reasonable.

7. The acupuncture rehabilitation training data processing method based on VR and digital therapy according to claim 6, characterized in that... When the feedback signal conflicts with physiological or VR behavior data, the analysis weight of the current data segment is adjusted, and the conflict propagation assessment is performed on the adjacent feedback windows. If a trend deviation is found, the basic analysis strategy of the adjacent windows is adjusted synchronously.

8. The acupuncture rehabilitation training data processing method based on VR and digital therapy according to claim 7, characterized in that... Before conducting the conflict propagation assessment of the adjacent feedback windows, cross-consistency verification is performed using data from two or more non-adjacent windows that are separated from the current window. If it is found that the current conflict is highly consistent with the trend of the windows far away, the influence coefficient of the adjacent window propagation adjustment is increased.

9. The acupuncture rehabilitation training data processing method based on VR and digital therapy according to claim 8, characterized in that... In the conflict propagation assessment, the affected adjacent window range is expanded or reduced according to the severity of the conflict. Mild conflicts only adjust a single window, moderate conflicts adjust both windows simultaneously, and severe conflicts are expanded to the window level across task cycles.

Images