Brain-computer interface exercise rehabilitation system parameter adaptive adjustment system and method

Abstract

The invention discloses a brain-computer interface exercise rehabilitation system parameter adaptive adjustment system and method. A data receiving and preprocessing module is configured to receive an electroencephalogram signal of a patient and perform preprocessing; the index calculation module is configured to determine a cognitive state index of the patient based on the preprocessed electroencephalogram signal, determine a deviation value of a cognitive state relative to a corresponding baseline, generate state change information if the deviation value exceeds a preset standard deviation, and send the state change information to the training parameter adjustment module; and the training parameter adjusting module is configured to call a pre-associated adjusting strategy according to the state change information, and adaptively adjust the exercise rehabilitation system parameters according to the adjusting strategy. According to the method, the cognitive state indexes are monitored in real time, so that the rehabilitation system parameters can be adaptively adjusted according to the rehabilitation process of the patient, the situation of rehabilitation interruption or invalid training caused by parameter discomfort is reduced, and the continuity and efficiency of rehabilitation training are improved.

Description

Technical Field

[0001] This application belongs to the technical field of brain-computer interfaces, and specifically relates to a system and method for adaptively adjusting the parameters of a motion rehabilitation system based on a brain-computer interface. Background Art

[0002] Brain-Computer Interface (BCI) technology aims to establish an information and control pathway between the human brain and the external environment that does not rely on peripheral nerves and muscle tissues, and it plays an important role in the rehabilitation treatment of upper limb movement disorders after stroke. BCI technology based on Motor Imagery (MI) Electroencephalogram (EEG) signals has been widely applied in clinics and has been proven to effectively promote the remodeling of nerve function. However, most existing MI-BCI rehabilitation systems are based on a fixed training process and cannot adaptively adjust the training parameters based on the patient's real-time cognitive state. Summary of the Invention

[0003] In view of the technical problem that the training parameters of the brain-computer interface motion rehabilitation system cannot be adaptively adjusted based on the patient's real-time cognitive state, this application provides a system and method for adaptively adjusting the parameters of a brain-computer interface motion rehabilitation system.

[0004] To achieve the above technical objectives, this application adopts the following technical solutions.

[0005] In a first aspect, an embodiment of this application provides a system for adaptively adjusting the parameters of a brain-computer interface motion rehabilitation system, including:

[0006] A data receiving and preprocessing module, configured to receive the patient's EEG signal and perform preprocessing;

[0007] An index calculation module, configured to determine the cognitive state index of the patient based on the preprocessed EEG signal, determine the deviation value of the cognitive state compared to the corresponding baseline, and if the deviation value exceeds the preset standard deviation, generate a state change information and send the state change information to the training parameter adjustment module;

[0008] The training parameter adjustment module, configured to retrieve the pre-associated adjustment strategy according to the state change information and adaptively adjust the parameters of the motion rehabilitation system according to the adjustment strategy.

[0009] Further, the cognitive state index includes a mental fatigue index, and the index calculation module is configured to determine that the expression of the mental fatigue index is:

[0010]

[0011] Among them, EI is the mental fatigue index, and E β is the ratio of the power of the β wave in the electroencephalogram signal to the total power of the preselected frequency band, and E α is the ratio of the power of the α wave in the electroencephalogram signal to the total power of the preselected frequency band, and E θ is the ratio of the power of the θ wave in the electroencephalogram signal to the total power of the preselected frequency band. The preselected frequency bands include the β wave frequency band, the α wave frequency band, and the θ wave frequency band.

[0012] Furthermore, the index calculation module is specifically configured to: compare the mental fatigue index with the mental fatigue index baseline to obtain a mental fatigue index deviation value. If the mental fatigue index deviation value exceeds the preset standard deviation, generate a first state change information, and send the first state change information to the training parameter adjustment module;

[0013] The training parameter adjustment module is specifically configured to: retrieve a first adjustment strategy associated with the first state change information, and adjust the parameters of the first exercise rehabilitation system according to the first adjustment strategy;

[0014] Furthermore, the cognitive state index includes a mental workload index;

[0015] The data reception and preprocessing module is configured to: perform data preprocessing on the electroencephalogram signal, intercept non-overlapping electroencephalogram signal segments according to a preset time length, perform high-pass filtering and band-pass filtering on each segment of the electroencephalogram signal in sequence, and divide it into four frequency bands: θ, α, β, and γ;

[0016] The index calculation module is configured to determine the mental workload index, including:

[0017] Determine the power spectral density of each frequency band based on the preprocessed electroencephalogram signal, determine the total power in each frequency band according to the power spectral density, and obtain a PSD feature matrix;

[0018] Input the PSD feature matrix into the trained regression model, use the regression model to obtain a mapping value, and use the mapping value as the mental workload index.

[0019] Furthermore, the index calculation module is also configured to:

[0020] Compare the mental workload index with the mental workload index baseline to obtain a mental workload index deviation value. If the mental workload index deviation value exceeds the preset standard deviation, generate a second state change information, and send the second state change information to the training parameter adjustment module;

[0021] The training parameter adjustment module is specifically configured to: retrieve a second adjustment strategy associated with the second state change information, and adjust the parameters of the second motion rehabilitation system according to the second adjustment strategy.

[0022] Further, the training method of the regression model includes:

[0023] Collect historical electroencephalogram signals during the calibration phase;

[0024] Perform data preprocessing on the historical electroencephalogram signals, intercept non-overlapping electroencephalogram signal segments according to a preset time window, and perform high-pass filtering and band-pass filtering on each segment of the electroencephalogram signals in turn;

[0025] Determine the power spectral density based on the filtered electroencephalogram signals, determine the total power in each of the four divided frequency bands of θ, α, β, and γ according to the power spectral density, and obtain a historical PSD feature matrix;

[0026] Label the load category for each segment of the historical electroencephalogram signals;

[0027] Input the historical PSD feature matrix and the corresponding load category into the regression model to train the regression model. The regression model includes the LDS module and the Softmax layer. Use the LDA module to obtain a mapping value based on the input historical PSD feature matrix, and use the Softmax layer to determine the corresponding category according to the mapping value.

[0028] Further, the cognitive state index includes a mental effort index. The index calculation module is configured to determine the mental effort index, including: dividing the electroencephalogram signals into specific frequency bands, and determining the sum of the power within the set frequency bands as the mental effort index.

[0029] Still further, the cognitive state index further includes a stress perception index. The index calculation module is configured to determine the stress perception index, including:

[0030] For the electroencephalogram signals within each time window, use the Pearson correlation coefficient to evaluate the correlation between the power of β waves and δ waves;

[0031] Determine the difference in the average power of α waves on two preselected EEG channels;

[0032] Determine the difference in the fuzzy entropy of the energy on the two preselected EEG channels;

[0033] Use the principal component analysis method to map the correlation, the difference in the average power of α waves, and the difference in fuzzy entropy to a single-dimensional feature space to obtain a mapped feature, and use the mapped feature as the stress perception index.

[0034] Further, the index calculation module is specifically configured to: obtain a mental effort index deviation value by comparing the mental effort index with a mental effort index baseline, or obtain a stress perception index deviation value by comparing the stress perception index with a stress perception index baseline; generate third state change information if the mental effort index deviation value exceeds a preset standard deviation, generate fourth state change information if the stress perception index deviation value exceeds the preset standard deviation, and send the third state change information and the fourth state change information to the training parameter adjustment module;

[0035] The training parameter adjustment module is specifically configured to: retrieve an associated third adjustment strategy according to the state change information and the fourth state change information, and adjust the parameters of the third motion rehabilitation system according to the third adjustment strategy.

[0036] In a second aspect, an embodiment of the present application provides a method for adaptively adjusting parameters of a brain-computer interface motion rehabilitation system, including: receiving an electroencephalogram signal of a patient and performing preprocessing;

[0037] Determining a cognitive state index of the patient based on the preprocessed electroencephalogram signal, determining a deviation value of the cognitive state compared to a corresponding baseline, generating state change information if the deviation value exceeds a preset standard deviation, and sending the state change information to a training parameter adjustment module;

[0038] Retrieving a pre-associated adjustment strategy according to the state change information, and adaptively adjusting the parameters of the motion rehabilitation system according to the adjustment strategy.

[0039] Compared with the prior art, the system and method for adaptively adjusting parameters of a brain-computer interface motion rehabilitation system provided by the present application can deeply understand the instant psychological and cognitive conditions of each patient during the rehabilitation process by collecting the electroencephalogram signals of the patients. As the rehabilitation training progresses, the cognitive state and physical functions of the patients will continuously change. The present application monitors the cognitive state index in real time, enabling the parameters of the rehabilitation system to be adaptively adjusted according to the rehabilitation progress of the patients, reducing the situations of rehabilitation interruption or ineffective training caused by inappropriate parameters, thereby improving the continuity and efficiency of the rehabilitation training. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings described herein are for illustrative purposes only and are not intended to limit the scope of the present disclosure in any way. Additionally, the shapes and proportional dimensions of the components in the drawings are only schematic for facilitating the understanding of the present application and do not specifically limit the shapes and proportional dimensions of the components of the present application. Those skilled in the art can, under the teaching of the present application, select various possible shapes and proportional dimensions according to specific circumstances to implement the present application. In the drawings:

[0041] Figure 1Schematic structural diagram of the parameter adaptive adjustment system of the brain-computer interface motion rehabilitation system provided for the embodiment;

[0042] Figure 2 Schematic principle diagram of the index calculation module in the embodiment;

[0043] Figure 3 Schematic diagram of the upper limb rehabilitation training process of the parameter adaptive adjustment system of the brain-computer interface motion rehabilitation system provided for the embodiment;

[0044] Figure 4 Schematic diagram of the parameter adaptive adjustment method flow of the brain-computer interface motion rehabilitation system provided for the embodiment. Specific implementation mode

[0045] In order to enable those skilled in the art to better understand the technical solutions in this application, the following will clearly and completely describe the technical solutions in the embodiments of this application in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this application.

[0046] In the description of this application, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more features.

[0047] Aiming at the current situation that the brain-computer interface motion rehabilitation system cannot adaptively adjust the training parameters based on the real-time cognitive state of the patient, the embodiment of this application provides a parameter adaptive adjustment system for the brain-computer interface motion rehabilitation system, which can recognize the motion intention based on the electroencephalogram signal generated when the patient performs motor imagery, determine the cognitive state of the patient, and adaptively adjust the parameters of the motion rehabilitation system according to the recognition result and the real-time cognitive state of the patient.

[0048] As Figure 1 shown, the system provided by the embodiment includes a data reception and preprocessing module, an index calculation module, and a training parameter adjustment module.

[0049] The data reception and preprocessing module is configured to receive the EEG data transmitted by the EEG acquisition module in real time and perform preprocessing (which may include filtering, dividing time windows, or dividing into different frequency bands according to frequency).

[0050] The index calculation module is configured to determine the cognitive state index of the patient based on the preprocessed electroencephalogram signal, determine the deviation value of the cognitive state compared to the corresponding baseline, and if the deviation value exceeds the preset standard deviation, generate a state change message and send the state change message to the training parameter adjustment module.

[0051] The training parameter adjustment module is configured to retrieve the pre-associated adjustment strategy according to the state change message and adaptively adjust the parameters of the motor rehabilitation system according to the adjustment strategy. The training parameter adjustment module can realize the adjustment of the motor intention recognition parameters and / or the adjustment of the gamification interaction paradigm parameters.

[0052] In some embodiments, the system for adaptively adjusting the parameters of the brain-computer interface motor rehabilitation system further includes a data storage and report generation module, which can generate report data and send it to the gamification interaction paradigm after the training is completed.

[0053] As Figure 2 shown, the cognitive state indexes determined by the index calculation module may include a mental fatigue calculation module, a mental workload calculation module, a mental effort calculation module, and a stress perception calculation module. The core objective of the index calculation module is to map the electroencephalogram signal into a cognitive state curve and compare the cognitive state curve with the corresponding baseline to obtain the state change information. Specifically, the data reception and preprocessing module extracts electroencephalogram signal segments of a specific time window length at every fixed time interval Δt; the index calculation module uses the mental fatigue calculation module, the mental workload calculation module, the mental effort calculation module, and the stress perception calculation module to calculate the index values of these electroencephalogram segments in the four dimensions of mental fatigue, mental workload, mental effort, and stress perception respectively, and compares each index value with the corresponding one to obtain the deviation value. If the deviation value exceeds the preset standard deviation, it sends the state change message to the training parameter adjustment module.

[0054] The cognitive state index calculated based on the electroencephalogram signal in this application can capture the state change information in real time, so it can provide personalized intervention measures for each individual, which helps to improve the training effect or work efficiency.

[0055] In some embodiments, the cognitive state index includes a mental fatigue (MF) index.

[0056] Mental fatigue occurs when a task requires a high level of attention and concentration. Research shows that there are mainly four categories of factors that trigger mental fatigue: 1) Drowsiness, which refers to a decrease in cognitive attention level accompanied by a desire to sleep; 2) Transition phase, which is defined as the transition from wakefulness to sleep; 3) Long-duration tasks, where as the time spent on a task increases, human performance declines and mental fatigue begins to appear; 4) Task engagement, which refers to a positive and excited state, influenced by workload status, task duration, motivation, and emotion. Specifically in the context of rehabilitation training, the focus is on mental fatigue caused by task engagement.

[0057] As an example, the calculation steps of the Mental Fatigue (MF) index are as follows:

[0058] 1) Obtain signal segments at four electrode positions, namely O3, O4, F3, and F4, within a time window (e.g., the time window is 20 seconds);

[0059] 2) Smooth the signal segments using a Hann window;

[0060] 3) Calculate the power spectrum using FFT, and combine the band powers to calculate the total power in three bandwidths (θ: 4 - 7Hz, α: 8 - 12Hz, β: 13 - 30Hz); 4) Divide the bandwidth power by the total power to generate the percentage power E θ 、E α and E β ; 5) The Engagement Index EI is calculated by the following formula:

[0061]

[0062] In an actual application embodiment, the index calculation module uses the EI curve calculated from the EEG signals collected during the calibration phase of the brain-computer interface motion rehabilitation system parameter adaptive adjustment system as a baseline, and continuously measures the deviation of the EI value from the baseline during formal training. Every time it rises or falls by 0.2 standard deviations, a first state change information is generated, and this first state change information is sent to the training parameter adjustment module as the basis for parameter adjustment.

[0063] The training parameter adjustment module can be configured to: retrieve the first adjustment strategy associated with the first state change information, and adjust the first motion rehabilitation system parameters according to the first adjustment strategy; the first adjustment strategy includes: if it is determined according to the first state change information that the mental fatigue index is greater than the mental fatigue index exceeding the preset standard deviation, then increase the set proportion of the rest time between motor imagery tasks; if it is determined according to the first state change information that the mental fatigue index is less than the mental fatigue index exceeding the preset standard deviation, then decrease the set proportion of the rest time between motor imagery tasks.

[0064] In some embodiments, the cognitive state indicator includes a Mental Workload (MW) indicator.

[0065] In high-intensity load scenarios, an individual's cognitive resources are often consumed to a great extent. This not only leads to desensitization to auditory warning signals but also causes overall neglect of all input information, a significant slowdown in the decision-making process, and an overall decline in the alertness level. Especially in a rehabilitation environment, this resource depletion is specifically manifested as ignoring task cues or difficulty in performing motor imagery. Some studies have shown that as the task complexity increases, the power spectral density (PSD) of the alpha band in the parietal region shows a downward trend, while the PSD of the theta band in the frontal region increases correspondingly.

[0066] As an example, the calculation steps of the mental workload indicator include:

[0067] 1) The data reception and preprocessing module preprocesses the electroencephalogram (EEG) signals. It intercepts non-overlapping EEG signal segments according to a preset time length (such as 2 seconds), and sequentially performs high-pass filtering (cutoff frequency 0.5 Hz) and band-pass filtering (frequency range 2 Hz to 50 Hz) on each segment of the EEG signals to effectively remove the DC drift and enhance the frequency signals closely related to cognitive activities. Four frequency bands, namely θ, α, β, and γ, are divided.

[0068] 2) PSD estimation: The indicator calculation module uses the Welch method to calculate the PSD of the filtered EEG signals, and then integrates the PSD values over four key frequency bands (θ: 4 - 8 Hz, α: 8 - 12 Hz, β: 12 - 30 Hz, γ: 30 - 40 Hz) to obtain a PSD feature matrix with the shape of (C, 4), laying a solid foundation for subsequent analysis, where C is the number of EEG measurement channels.

[0069] 3) The indicator calculation module inputs the PSD feature matrix as a feature vector into the trained regression model, and uses the regression model to obtain a mapped value based on the input feature vector, and takes the mapped value as the mental workload indicator.

[0070] As an example, during the formal rehabilitation training process, the PSD feature matrix of the EEG signal segment is mapped into specific numerical values by using the established regression model. Taking the mean of the mapped numerical values of the low-load and high-load signals in the calibration stage as the baseline, the index calculation module continuously monitors and evaluates the deviation degree of the mapped numerical values of the EEG signals during rest and motor imagery tasks in the formal training relative to this baseline to obtain the deviation value of the mental workload index. Whenever the deviation value of the mental workload index reaches ±0.2 standard deviations, the second state change information is generated, and this second state change information will be immediately transmitted to the training parameter adjustment module as an important basis for dynamically adjusting the training parameters, so as to achieve precise control and optimization of the rehabilitation training. Among them, the baseline of the mental workload index is the mean of the mapped numerical values obtained by using the regression model based on the PSD feature matrix of different categories of electroencephalogram signals.

[0071] The training parameter adjustment module can be configured to: retrieve the second adjustment strategy associated with the second state change information, and adjust the parameters of the second motor rehabilitation system according to the second adjustment strategy; the second adjustment strategy includes: if it is judged according to the second state change information that the mental workload index is greater than the mental workload index baseline by more than the preset standard deviation, increase the set ratio of the movement distance of the rehabilitation device; if it is judged according to the second state change information that the mental workload index is less than the mental workload index baseline by more than the preset standard deviation, decrease the set ratio of the movement distance of the rehabilitation device.

[0072] In the embodiment, the training method of the regression model includes: collecting historical electroencephalogram signals in the calibration stage; performing data preprocessing on the historical electroencephalogram signals, intercepting non-overlapping electroencephalogram signal segments according to the preset time window, and performing high-pass filtering and band-pass filtering on each segment of the electroencephalogram signal in turn; determining the power spectral density based on the filtered electroencephalogram signals, and determining the total power in each of the four divided frequency bands of θ, α, β, and γ according to the power spectral density to obtain the historical PSD feature matrix.

[0073] Label the load category for each segment of the historical electroencephalogram signal according to the activity state corresponding to the EEG signal in the calibration stage. As an example, the signals during rest and preparation are marked as low load, while the signals during the execution of motor imagery tasks (such as left and right hand motor imagery) are regarded as high load. To ensure the effectiveness and representativeness of the data, 5 times of left and right hand motor imagery tasks are collected respectively during calibration, each lasting for 4 seconds, and each sample is a 2-second signal segment, a total of 20 high-load samples are obtained; at the same time, the same number of samples are also collected in the rest and preparation stages for constructing and training the regression model.

[0074] Input the historical PSD feature matrix and the corresponding load categories into the regression model for training. The regression model includes an LDS module and a Softmax layer. Use the LDA module to obtain the mapped values based on the input historical PSD feature matrix, and use the Softmax layer to determine the corresponding category according to the mapped values.

[0075] The regression model uses the Linear Discriminant Analysis (LDA) method to determine the optimal subspace projection, aiming to reduce the historical PSD feature matrix of high-dimensional EEG data to one dimension, that is, the key dimension reflecting the user's workload. Further, use the Softmax layer to optimize this linear projection through logistic regression technology, assuming that the class conditional probability of the given projection follows the logistic model:

[0076]

[0077] where the weight vector w and the bias b are adjusted by maximizing the data likelihood to ensure a high degree of consistency between the data and the logistic model distribution of the labels. The leave-one-out cross-validation method is used to comprehensively evaluate the performance of the regression model to ensure its stability and reliability.

[0078] In other embodiments, the cognitive state indicator further includes a Mental Effort (ME) indicator.

[0079] Mental effort is a key indicator to measure the degree of cognitive resource allocation of an individual when performing a specific task, and it is closely related to the degree of investment and participation in the task. From the perspective of cognitive science, mental effort is not only closely related to the degree of attention concentration on external stimuli, but also involves the in-depth processing of task-related information, which is accompanied by the enhancement of activities in specific brain regions, thereby promoting the improvement of neural efficiency.

[0080] As an example, determining the mental effort indicator includes: based on the frequency bands divided by the electroencephalogram signal by the data reception and preprocessing module, the indicator calculation module determines the sum of the powers in the set frequency bands as the mental effort indicator.

[0081] Specifically, it includes the following steps: Signal acquisition and frequency band division Select the electroencephalogram (EEG) signals at four key positions, namely Fz, Cz, Pz, and O1, as the analysis objects. Subsequently, these signals are subdivided into five specific frequency bands, namely the δ frequency band (0-3Hz), the θ frequency band (4-7Hz), the α frequency band (8-12Hz), the β1 frequency band (13-22Hz), and the β2 frequency band (23-30Hz), and calculate the sum of the powers in each frequency band to comprehensively capture the characteristics of brain activities.

[0082] As an example, to establish an accurate reference baseline, the EEG signals during the preparation period are segmented without overlap with a time window of 2 seconds, and the power values of the above five frequency bands within each window are calculated as the baseline data for subsequent analysis.

[0083] Existing research has pointed out that an increase in mental effort is often accompanied by significant changes in the power of the θ and β2 frequency bands in EEG. In the embodiment, with a time interval of 2 seconds, the power dynamics of these two frequency bands are continuously monitored. Once a difference of 0.2 standard deviations in the mental effort index compared to the baseline is detected, third state change information is generated and sent to the training parameter adjustment module. This mechanism provides a scientific basis for dynamically adjusting the rehabilitation training parameters, aiming to achieve precise control and optimization of the rehabilitation training process, thereby maximizing the training effect.

[0084] In some embodiments, the cognitive state index further includes a stress perception index (Stress index).

[0085] Emotions play a crucial role in human overall performance as they significantly affect cognitive functions, decision-making, and personal performance, while stress often stems from emotional burdens. In the context of rehabilitation training, continuous monitoring of stress and timely adjustment of treatment parameters are of great importance for improving the rehabilitation effectiveness of patients.

[0086] Based on EEG-based stress level monitoring technology, commonly used methods include analyzing the asymmetry of left and right frontal lobe activities and the correlation between β waves and δ waves. It is known that the power of frontal lobe α waves is negatively correlated with task-related activities. Therefore, by observing the asymmetry of α wave power, the stress state can be effectively evaluated.

[0087] As an example, the method for determining the stress perception index includes the following steps:

[0088] 1) The data reception and preprocessing module performs data preprocessing: with a time window of 5 seconds and an interval of 2 seconds, segments for calculating the stress index are intercepted from the EEG signals, and the F3 and F4 channels are specifically selected as the signal sources.

[0089] 2) The index calculation module performs spectral analysis: applying the FFT of the Welch method, the power changes of the α, β, δ, and θ frequency bands are extracted from each 5-second segment, with the window length set to 1 second.

[0090] 3) The index calculation module performs correlation analysis: for the δ and β frequency bands, by dividing their original power by the sum of the total spectral power, the relative power changes are obtained, and the Pearson correlation coefficient is used to evaluate the correlation between the power of β waves and δ waves. The calculation formula is:

[0091]

[0092] Among them, x and y respectively represent the relative power change curves of β waves and δ waves in the 5-second signal segment.

[0093] 4) The index calculation module conducts alpha wave asymmetry evaluation: By calculating the difference in the average power of alpha waves and the difference in fuzzy entropy (Fuzzy Entropy, FuzzEn) on the F3 channel and the F4 channel, to measure the asymmetry of alpha waves in terms of energy and complexity. Research shows that under stress conditions, the asymmetry of the prefrontal lobe increases, and the specific calculation is as follows:

[0094] AlphaAsym(t) = P F4 (t) - P F3 (t) (4)

[0095] FuzzEnAsym(t) = FuzzEn F4 (t) - FuzzEn F3 (t) (5) Among them, the fuzzy entropy of the information sequence is an effective tool for measuring the complexity of time series.

[0096] 5) After the above processing, each 5s EEG signal segment is mapped to a 3D feature vector, and the signals during the calibration period are mapped to a feature sequence, specifically as follows:

[0097]

[0098] Where Stress(t) fit is the feature sequence at time t, r(t) is the correlation between the power of β waves and δ waves at time t, AlphaAsym(t) is the difference in the average power of alpha waves on the two preselected EEG channels (F3 channel and F4 channel), and FuzzEnAsym(t) is the difference in the fuzzy entropy of energy on the two preselected EEG channels (F3 channel and F4 channel).

[0099] As an example, use PCA to map the above feature sequence Stress(t) at time t fit to a one-dimensional feature space as the baseline, and use the same mapping matrix to map the signals in the formal training to this feature space to obtain the stress perception index, monitor the deviation of the stress perception index from the stress perception index baseline, obtain the stress perception index deviation value. When the stress perception index deviation value reaches ±0.2 standard deviations, generate the fourth state change information, and immediately transmit the fourth state change information to the training parameter adjustment module as the key basis for dynamically adjusting the training parameters, so as to achieve precise control and optimization of the rehabilitation training process.

[0100] The training parameter adjustment module can be configured to retrieve the associated third adjustment strategy according to the third state change information and the fourth state change information, and adjust the parameters of the third motion rehabilitation system according to the third adjustment strategy. The third adjustment strategy includes: if it is determined according to the third state change information that the mental effort index is greater than the baseline of the mental load index by more than the preset standard deviation, or if it is determined according to the fourth state change information that the stress perception index is greater than the baseline of the stress perception index by more than the preset standard deviation, then reduce the setting ratio of the motion intention recognition threshold; if it is determined according to the third state change information that the mental effort index is less than the baseline of the mental load index by more than the preset standard deviation, and if it is determined according to the fourth state change information that the stress perception index is less than the baseline of the stress perception index by more than the preset standard deviation, then increase the setting ratio of the motion intention recognition threshold.

[0101] As an example, the training process in the formal training stage is as Figure 3 shown. The index calculation module calculates the baselines of the aforementioned 4 cognitive indexes, including the mean and the standard deviation, in the preparation stage and the model calibration stage. In the formal training stage, the monitoring results (i.e., state change information) are continuously sent to the training parameter adjustment module. The changes recorded in the state change information are in units of 0.2 standard deviations, that is, when the change is less than ±0.2 standard deviations, the value of the previous moment is maintained. The training parameter adjustment module starts to continuously receive the cognitive monitoring information x(t) = {MF(t), MW(t), ME(t), Stress(t)} transmitted by the index calculation module in the formal training stage after completing the model calibration stage, and performs dynamic parameter adjustment accordingly.

[0102] Some training parameters to be adjusted in some embodiments include: 1) the rest time t between tasks rest ; 2) the motion intention recognition threshold t; 3) the motion distance S of the upper limb rehabilitation device. By adjusting the above parameters, the training intensity and difficulty can be effectively regulated, thereby improving the training experience of the patient. The specific adjustment strategy is as follows:

[0103] (1) The rest time t between tasks rest Adjustment strategy

[0104] Adjustment basis: This strategy mainly dynamically adjusts the rest time t according to the change of the mental fatigue (MF) index baseline . As an important index for evaluating the fatigue state of the patient, the increase of the MF index usually means that the patient begins to feel fatigued. Therefore, it is necessary to effectively relieve fatigue by increasing the rest time to ensure the training effect and the safety of the patient.

[0105] Adjustment Strategy: 1) Initial Setting: Before the start of training, a professional sets a standard rest time and a threshold for the increase in the MF index according to the characteristics of the training task and the patient's physical condition. This threshold serves as a benchmark for determining whether the patient has started to feel fatigued. 2) Fatigue Monitoring and Increase in Rest Time: During training, continuously monitor the change in MF. When the MF index rises and exceeds the preset threshold, it indicates that the patient has started to show signs of fatigue. At this time, the rest time should be gradually increased, and it is recommended to increase it by 5% to 10% after each trial until the MF index shows a downward trend. This step aims to relieve the patient's fatigue by increasing rest and ensure the continuous progress of training. 3) Fatigue Relief and Decrease in Rest Time: As the MF index starts to decline from a high level, it indicates that the patient's fatigue state has been relieved to a certain extent. At this time, the rest time should be gradually decreased, and it is recommended to decrease it by 5% to 10% after each trial to promote the improvement of training efficiency. 4) Recovery Phase and Rapid Decrease in Rest Time: When the MF index returns to the baseline level or lower, it indicates that the patient's fatigue state has been significantly relieved. In this phase, the rest time should be decreased more quickly, and it is recommended to decrease it by 10% to 20% after each trial until it returns to the initially set standard rest time. This step aims to resume the normal training rhythm as soon as possible. 5) Optimization Adjustment Below the Baseline Level: When the MF index is at or below the baseline level and the mental workload (MW) index is also below the baseline level, it indicates that the patient is in a good current state and has the potential to withstand a higher training intensity. At this time, the rest time should be gradually decreased, and it is recommended to decrease it by 5% to 10% after each trial until the rest time is reduced to half of the standard rest time or the MF index returns to the baseline level. This step aims to further optimize the training effect and improve the patient's training efficiency.

[0106] (2) Adjustment Strategy for the Threshold (T) of Movement Intention Recognition

[0107] Basis for Adjustment: The adjustment of the threshold (T) for movement intention recognition is mainly based on the changes in the patient's mental effort (ME) index and stress perception (Stress) index. When the ME index or Stress index rises, it may mean that the patient is experiencing high cognitive stress. At this time, lowering the threshold can make the system more sensitive to capture the patient's movement intention, thereby reducing the patient's burden and enhancing the training experience.

[0108] Adjustment Method: 1) Initial Calibration: Before the start of training, a professional sets an initial threshold (T) for movement intention recognition according to the patient's movement ability and training goals. baseline2) High-load response: During the training process, if ME or Stress continues to increase and exceeds the preset high threshold, it indicates that the patient is in a high-load state. At this time, the motion intention recognition threshold should be gradually reduced, and it is recommended to adjust and reduce it by 1% to 5% each time until the ME index and Stress index return to a level below the threshold. This step aims to enable the patient to complete the movements more easily and relieve physical and mental stress by lowering the recognition threshold. 3) Load stabilization and threshold fine-tuning: When the patient is continuously judged to be imagining correctly k times and the training process is smooth, maintain the current threshold without further reduction. 4) Recovery stage and threshold increase: As the training progresses, if the ME index and Stress index return to the baseline level or are lower than the baseline level, it indicates that the patient has adapted to the current training intensity. At this time, the motion intention recognition threshold can be gradually increased, with each adjustment increasing by 5% to 10% until the initial set T baseline This step aims to gradually restore the normal recognition sensitivity as the patient's ability improves, and avoid a decline in training effectiveness caused by over-adaptation.

[0109] (3) Adjustment strategy for the movement distance (S) of the upper limb rehabilitation device

[0110] Basis for adjustment: This strategy mainly dynamically adjusts the movement distance S of the upper limb rehabilitation device based on the patient's mental workload (MW) index. Different from the traditional approach that mainly considers physical load, this strategy particularly focuses on the patient's psychological burden, aiming to effectively relieve the patient's psychological stress by adjusting the patient's passive movement distance, while promoting the recovery of upper limb function.

[0111] Adjustment strategy: 1) Baseline setting and monitoring: Before the start of training, set a basic movement distance S according to the patient's rehabilitation stage, upper limb function status, and psychological tolerance. During the training process, continuously monitor the patient's mental workload (MW) index to promptly capture changes in their mental state. 2) When the MW index increases and exceeds the preset threshold (set based on the baseline, exceeding the baseline standard deviation), it indicates that the patient has a high mental workload and may feel anxious, tense, or fatigued. At this time, contrary to traditional concepts, this strategy recommends appropriately increasing S to provide a "relaxed" rehabilitation experience by extending the patient's passive movement time, thereby indirectly relieving the patient's psychological stress. Specifically, when the MW index exceeds the threshold, the movement distance can be gradually increased, and it is recommended to adjust and increase it by 5% to 10% per trial until the MW index no longer rises or reaches 200% of S baseline 3) When the MW index drops to the baseline level or is lower than the baseline level, gradually reduce S until the MW index starts to exceed the threshold again or S drops to 50% of S baseline and increase the training intensity of active movement to promote more comprehensive rehabilitation. baseline

[0112] Based on the same inventive concept as the brain-computer interface motion rehabilitation system parameter adaptive adjustment system provided in the above embodiments, the embodiments of the present application also provide a method for adaptively adjusting the parameters of a brain-computer interface motion rehabilitation system, as Figure 4 shown, including:

[0113] Step S1: Receive the electroencephalogram signal of the patient and perform preprocessing;

[0114] Step S2: Determine the cognitive state index of the patient based on the preprocessed electroencephalogram signal, determine the deviation value of the cognitive state compared to the corresponding baseline, and if the deviation value exceeds the preset standard deviation, generate a state change information and send the state change information to the training parameter adjustment module;

[0115] Step S3: Retrieve the pre-associated adjustment strategy according to the state change information, and adaptively adjust the parameters of the motion rehabilitation system according to the adjustment strategy.

[0116] It can be understood that in this embodiment, the implementation methods of each step can refer to the implementation methods of each module in the above embodiments, and will not be elaborated here.

[0117] The above has introduced in detail the brain-computer interface motion rehabilitation system parameter adaptive adjustment system and method provided by the present application. Specific examples are used in this article to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the concept of the present application, and should not be construed as a limitation on the protection scope of the present application.

Claims

1. A parameter adaptive adjustment system for a brain-computer interface motor rehabilitation system, characterized in that: The system comprises: A data receiving and preprocessing module is configured to receive and preprocess the patient's electroencephalogram signal; an index calculation module, configured to determine the patient's cognitive state index based on the preprocessed electroencephalogram signal, determine a deviation value of the cognitive state compared to a corresponding baseline, and if the deviation value exceeds a preset standard deviation, generate state change information, and send the state change information to a training parameter adjustment module; The training parameter adjustment module is configured to retrieve a pre-associated adjustment strategy according to the state change information, and adaptively adjust the exercise rehabilitation system parameters according to the adjustment strategy.

2. The parameter adaptive adjustment system of the brain-computer interface motor rehabilitation system according to claim 1 is characterized in that: The cognitive state index includes a mental fatigue index, and the index calculation module is configured to determine the expression of the mental fatigue index as: Among them, EI is the mental fatigue index, E β is the ratio of the beta wave power in the EEG signal to the total power of the preselected frequency band, E α is the ratio of α wave power in the EEG signal to the total power of the preselected frequency band, E θ It is the ratio of the θ wave power in the EEG signal to the total power of the pre-selected frequency bands, which include the β wave band, the α wave band and the θ wave band.

3. The method for adaptively adjusting parameters of a brain-computer interface motor rehabilitation system according to claim 2 is characterized in that: The index calculation module is specifically configured to: compare the mental fatigue index with the mental fatigue index baseline to obtain a mental fatigue index deviation value, if the mental fatigue index deviation value exceeds a preset standard deviation, generate first state change information, and send the first state change information to the training parameter adjustment module; The training parameter adjustment module is specifically configured to: retrieve a first adjustment strategy associated with the first state change information, and adjust the first exercise rehabilitation system parameter according to the first adjustment strategy.

4. The method for adaptively adjusting parameters of a brain-computer interface motor rehabilitation system according to claim 1, characterized in that: The cognitive state index includes a mental load index; The data receiving and preprocessing module is configured to: perform data preprocessing on the EEG signal, intercept non-overlapping EEG signal segments according to a preset time length, perform high-pass filtering and band-pass filtering on each segment of the EEG signal in turn, and divide the EEG signal into four frequency bands of θ, α, β, and γ; The index calculation module is configured to determine the mental load index, including: Determine the power spectral density of each frequency band based on the preprocessed EEG signal, determine the total power in each frequency band according to the power spectral density, and obtain a PSD feature matrix; The PSD feature matrix is ​​input into the trained regression model, and the mapping value is obtained by using the regression model, and the mapping value is used as the mental load index.

5. The method for adaptively adjusting parameters of a brain-computer interface motor rehabilitation system according to claim 4, characterized in that: The indicator calculation module is also configured as follows: Comparing the mental load index with the mental load index baseline to obtain a mental load index deviation value, if the mental load index deviation value exceeds a preset standard deviation, generating second state change information, and sending the second state change information to the training parameter adjustment module; The training parameter adjustment module is specifically configured to: retrieve a second adjustment strategy associated with the second state change information, and adjust the second motion rehabilitation system parameter according to the second adjustment strategy.

6. The method for adaptively adjusting parameters of a brain-computer interface motor rehabilitation system according to claim 4, characterized in that: The training method of the regression model includes: Historical EEG signals were collected during the calibration phase; Performing data preprocessing on the historical EEG signals, intercepting non-overlapping EEG signal segments according to a preset time window, and sequentially performing high-pass filtering and band-pass filtering on each segment of the EEG signal; Determine the power spectral density based on the filtered EEG signal, determine the total power in each of the four divided frequency bands of θ, α, β, and γ according to the power spectral density, and obtain a historical PSD feature matrix; Labeling the load category of each segment of the historical EEG signal; The historical PSD feature matrix and the corresponding load category are input into the regression model to train the regression model. The regression model includes the LDS module and the Softmax layer. The LDA module is used to obtain mapping values ​​based on the input historical PSD feature matrix, and the Softmax layer is used to determine the corresponding category according to the mapping values.

7. The method for adaptively adjusting parameters of a brain-computer interface motor rehabilitation system according to claim 1, characterized in that: The cognitive state index includes a mental effort index, and the index calculation module is configured to determine the mental effort index, including: dividing the electroencephalogram signal into a specific frequency band, and determining the sum of the power in the set frequency band as the mental effort index.

8. The method for adaptively adjusting parameters of a brain-computer interface motor rehabilitation system according to claim 7, characterized in that: The cognitive state index also includes a stress perception index, and the index calculation module is configured to determine the stress perception index, including: For the EEG signals in each time window, the Pearson correlation coefficient was used to evaluate the correlation between the beta and delta wave powers; Determine the difference in the average power of alpha waves on two preselected EEG channels; Determining the fuzzy entropy difference of energy on the two preselected EEG channels; The correlation, α wave average power difference and fuzzy entropy difference are mapped to a single-dimensional feature space using a principal component analysis method to obtain mapping features, and the mapping features are used as the pressure perception index.

9. The method for adaptively adjusting parameters of a brain-computer interface motor rehabilitation system according to claim 8, characterized in that: The indicator calculation module is specifically configured to: compare the mental effort indicator with the mental effort indicator baseline to obtain a mental effort indicator deviation value, or compare the stress perception indicator with the stress perception indicator baseline to obtain a stress perception indicator deviation value, if the mental effort indicator deviation value exceeds a preset standard deviation, then generate third state change information, if the stress perception indicator deviation value is less than, then generate fourth state change information, and send the third state change information and the fourth state change information to the training parameter adjustment module; The training parameter adjustment module is specifically configured to: retrieve the associated third adjustment strategy according to the state change information and the fourth state change information, and adjust the third exercise rehabilitation system parameter according to the third adjustment strategy.

10. A method for adaptively adjusting parameters of a brain-computer interface motor rehabilitation system, characterized in that: include: Receive the patient's EEG signal and pre-process it; Determine the patient's cognitive state index based on the preprocessed electroencephalogram signal, determine the deviation value of the cognitive state compared to the corresponding baseline, and if the deviation value exceeds a preset standard deviation, generate state change information, and send the state change information to a training parameter adjustment module; A pre-associated adjustment strategy is retrieved according to the state change information, and the parameters of the motor rehabilitation system are adaptively adjusted according to the adjustment strategy.

Images