An adaptive closed-loop brain-computer interface neurofeedback training system

Abstract

The application is particularly a self-adaptive closed-loop brain-computer interface neural feedback training system, and relates to the technical fields of brain-computer interface and neural feedback, comprising: a multi-modal signal acquisition module; a multi-modal feature extraction module; a coefficient fusion module; and a feedback training module.In the application, high-precision crystal oscillator clock is adopted to realize time synchronization of electroencephalogram, eye movement and behavior signals, and fusion distortion caused by signal misplacement is completely eliminated.Three types of signals cover cognitive state, visual attention and motion characteristics respectively, forming complementary state evaluation dimensions; and fine quantization algorithms are designed for each modality: electroencephalogram coefficients are screened by extreme value to eliminate instantaneous artifacts, eye movement coefficients calibrate pixel diameter to physical diameter to eliminate imaging interference, and behavior coefficients take into account large motion intensity and high-frequency posture micro-variation.

Description

Technical Field

[0001] This invention relates to the field of brain-computer interface and neurofeedback technology, and in particular to an adaptive closed-loop brain-computer interface neurofeedback training system. Background Technology

[0002] Neurofeedback training, by collecting users' physiological signals in real time and converting them into perceptible feedback information, helps users actively regulate their cognitive state and has been widely applied in scenarios such as attention enhancement and emotion regulation. Current mainstream systems mostly rely on single EEG signals, which suffer from poor anti-interference capabilities and low state recognition accuracy.

[0003] EEG signals are susceptible to interference from eye movement and electromyography artifacts, and single-modal data cannot fully reflect a user's cognitive and behavioral state.

[0004] While existing multimodal systems incorporate eye-tracking and behavioral signals, they suffer from three major drawbacks:

[0005] First, the signal synchronization accuracy is low, and the time references of EEG, eye movement, and behavioral signals are inconsistent, resulting in distorted fusion results.

[0006] Second, the methods for quantifying abnormality coefficients are crude. For example, eye movement signals are directly analyzed using pixel diameter, without eliminating imaging condition interference. Behavioral signals only focus on large-amplitude movements while ignoring high-frequency micro-changes in posture.

[0007] Third, the feedback strategy is rigid, using fixed parameters for training, which cannot adapt to changes in the user's real-time state and reduces the training effect.

[0008] Therefore, an adaptive closed-loop brain-computer interface neurofeedback training system is proposed to address the aforementioned problems. Summary of the Invention

[0009] The purpose of this invention is to provide an adaptive closed-loop brain-computer interface neurofeedback training system to solve the above-mentioned problems.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] An adaptive closed-loop brain-computer interface neurofeedback training system includes:

[0012] The multimodal signal acquisition module is configured to acquire EEG signals, eye movement signals, and behavioral signals, and performs preliminary preprocessing to eliminate noise and unify the time reference.

[0013] The multimodal feature extraction module is configured to analyze EEG signals, eye movement signals, and behavioral signals respectively to obtain EEG abnormality coefficients, eye movement abnormality coefficients, and behavioral abnormality coefficients.

[0014] The coefficient fusion module is configured to normalize the three obtained coefficients and then perform a weighted summation to obtain the comprehensive evaluation coefficient.

[0015] The feedback training module is configured to dynamically adjust the parameters of the neurofeedback task based on the comprehensive evaluation coefficient, and generate an adaptive neurofeedback training strategy that matches the user.

[0016] Preferably, the multimodal signal acquisition module specifically includes:

[0017] EEG signal acquisition:

[0018] A multi-channel EEG electrode cap, paired with a high-performance signal amplifier, is used to acquire electroencephalogram (EEG) signals.

[0019] Preprocessing: The raw EEG data is obtained through bandpass filtering, power frequency notch filtering, and artifact removal.

[0020] Eye movement signal acquisition:

[0021] Using an integrated or external video eye tracker, the pupil position coordinates, pupil diameter, blink frequency, and fixation point trajectory can be captured in real time.

[0022] Preprocessing: Identify and label typical eye-tracking events;

[0023] Behavioral signal acquisition:

[0024] The inertial measurement unit, with built-in accelerometer and gyroscope, can be integrated into a head-mounted device or worn independently on the limbs to collect motion data in real time.

[0025] Preprocessing: Calculate motion characteristics based on inertial measurement unit data;

[0026] A unified hardware clock is used to assign a unique timestamp to the signals from the three channels: EEG, eye movement, and behavior.

[0027] Preferably, the process of obtaining the EEG abnormality coefficient includes:

[0028] The raw EEG signal is segmented, a Hanning window is added, and then the spectrum is obtained through fast Fourier transform.

[0029] The power of a frequency band is the sum of the power of all frequency points within that band.

[0030] Relative power is the proportion of the power in that frequency band to the total power of the entire frequency band;

[0031] The baseline data of users is collected in advance to obtain the relative power benchmark value and standard deviation of Beta wave. The deviation of the current Beta wave power is calculated by Z-score and recorded as the initial EEG coefficient.

[0032] Preferably, the method further includes:

[0033] The normal range of the initial EEG coefficient is preset, and the initial EEG coefficient that is not within the normal range is recorded as the EEG deviation coefficient.

[0034] The EEG deviation coefficients are sorted in descending order of numerical value to obtain a set of EEG deviation coefficients. The k largest and k smallest EEG deviation coefficients are extracted from the set to form a new set. The absolute values ​​of each coefficient in the set are taken and the mean is calculated to obtain the EEG abnormality coefficient.

[0035] Preferably, the process of obtaining the eye movement abnormality coefficient includes:

[0036] The pupil pixel diameter output by the eye tracker is converted into the actual physical diameter and recorded as the actual pupil diameter;

[0037] Obtain the maximum and minimum values ​​of the actual physical diameter of the pupil within a preset time period, as well as the duration of that time period; subtract the minimum value from the maximum value of the actual physical diameter of the pupil within the preset time period, and then divide by the duration of that time period to obtain the pupil diameter change rate.

[0038] A preset threshold for pupil diameter change rate is set, and pupil diameter change rates that are not within the threshold are recorded as abnormal pupil diameter change values.

[0039] Preferably, the method further includes:

[0040] Obtain all abnormal pupil diameter values ​​within a preset time period and sort them in descending order of numerical value;

[0041] Starting from the maximum pupil diameter change anomaly, the interval duration between each adjacent pupil diameter change anomaly is extracted, and the obtained durations are rearranged in the order they were obtained to form a duration set.

[0042] A preset interval duration threshold is set, and values ​​greater than the interval duration threshold are removed from the duration set to obtain a new duration set. Then, three interval durations are extracted from it in chronological order and used as the length, width, and height of a cuboid. The volume of the cuboid is calculated and recorded as the eye movement abnormality coefficient.

[0043] Preferably, the process of obtaining the behavioral anomaly coefficient includes:

[0044] Acceleration and angular velocity data of limbs / head within a preset time period are obtained through an inertial measurement unit;

[0045] Calculate the peak triaxial acceleration;

[0046] A preset triaxial acceleration peak threshold is set, and triaxial acceleration peak values ​​that exceed the triaxial acceleration peak threshold are recorded as abnormal peak values; and the difference between the maximum abnormal peak value and the triaxial acceleration peak threshold within a preset time period is obtained to obtain the abnormal peak difference value.

[0047] Preferably, the method further includes:

[0048] Obtain the Euler angles of the head / torso within a preset time period. The Euler angles include pitch angle, yaw angle, and roll angle.

[0049] Actions where the pitch angle, yaw angle, and roll angle change by more than 5° within a preset time period are marked as flagged actions.

[0050] The marking behavior and its duration corresponding to pitch angle, yaw angle and roll angle are obtained respectively and recorded as the marking duration; a standard duration is preset, and the marking duration that is less than the standard duration is recorded as the abnormal duration;

[0051] Extract the anomaly durations corresponding to pitch angle, yaw angle, and roll angle respectively;

[0052] The average duration of each anomaly corresponding to pitch angle, yaw angle and roll angle is calculated to obtain the average duration of anomalies corresponding to pitch angle, yaw angle and roll angle respectively.

[0053] After counting the number of abnormal durations of pitch angle, yaw angle, and roll angle respectively, divide them by the corresponding average abnormal duration to obtain the abnormal change frequency of pitch angle, abnormal change frequency of yaw angle, and abnormal change frequency of roll angle.

[0054] Preferably, the method further includes:

[0055] The frequency of abnormal pitch angle changes Frequency of abnormal changes in yaw angle Frequency of abnormal changes in roll angle Establish a three-dimensional Cartesian coordinate system, denoted as Cartesian space, with the X, Y, and Z axes as the axes. The coordinates of any point P in the coordinate system are: ;

[0056] Collect a predetermined number of subjects in the target training task Baseline data, calculate the mean of each dimension;

[0057] Centered on the mean point f,

[0058] Will Mapped to real-time points in Cartesian space ;

[0059] Calculate the Euclidean distance from the real-time point to the center f of the normal area, and denot it as the quantization distance. Set a quantization distance threshold. If the calculated quantization distance is greater than the quantization distance threshold, then record it as an abnormal quantization value.

[0060] After normalizing the abnormal peak difference value and the abnormal quantification value, the abnormal behavior coefficient is obtained by weighted summation.

[0061] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0062] 1. This invention achieves time synchronization of EEG, eye movement, and behavioral signals by using a high-precision crystal oscillator clock, completely eliminating fusion distortion caused by signal misalignment. The three types of signals respectively cover cognitive state, visual attention, and motion characteristics, forming complementary state assessment dimensions. A refined quantization algorithm is designed for each modality: the EEG coefficient is filtered to remove transient artifacts through extreme value screening, the eye movement coefficient has its pixel diameter calibrated to the physical diameter to eliminate imaging interference, and the behavioral coefficient takes into account both large-scale motion intensity and high-frequency micro-changes in posture.

[0063] 2. This invention divides the task into three levels by a comprehensive evaluation coefficient. In the normal state, the task difficulty is gradually increased. In the mild abnormal state, the feedback frequency is finely adjusted. In the severe abnormal state, multi-sensory collaborative feedback is switched. Combined with the user's training history iteration parameter range, the duration of focus maintenance is improved and the efficiency of fatigue relief is enhanced. The quantitative algorithm logic is simple and reproducible, and no special hardware needs to be customized, which greatly reduces equipment costs and industrialization threshold. It not only meets the practical needs of clinical and training scenarios, but also has the engineering foundation for large-scale promotion. Attached Figure Description

[0064] Further details, features, and advantages of this application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0065] Figure 1 This is a system structure diagram of the present invention. Detailed Implementation

[0066] Several embodiments of this application will now be described in more detail with reference to the accompanying drawings to enable those skilled in the art to implement this application. This application may be embodied in many different forms and for various purposes and should not be limited to the embodiments set forth herein. These embodiments are provided to make this application thorough and complete, and to fully convey the scope of this application to those skilled in the art. The embodiments described do not limit this application.

[0067] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It will be further understood that terms such as those defined in commonly used dictionaries shall be interpreted as having a meaning consistent with their meaning in the relevant field and / or the context of this specification, and shall not be interpreted in an idealized or overly formal sense unless expressly defined herein.

[0068] Example 1

[0069] Its specific implementation method is combined with the appendix Figure 1 Please provide a detailed explanation.

[0070] Appendix Figure 1 This invention provides a structural block diagram of an adaptive closed-loop brain-computer interface neurofeedback training system, which shows the connection relationship between the multimodal signal acquisition module and the coefficient fusion module, and marks the main functional interaction flow of each module.

[0071] In this embodiment, it includes:

[0072] The multimodal signal acquisition module is configured to acquire EEG signals, eye movement signals, and behavioral signals. It also performs preliminary preprocessing to eliminate noise and unify the time reference, laying the data foundation for subsequent fusion analysis.

[0073] Specifically, it includes:

[0074] EEG signal acquisition:

[0075] A multi-channel EEG electrode cap (commonly 32 or 64 leads) is used in conjunction with a high-performance signal amplifier to acquire EEG signals;

[0076] Preprocessing: The clean raw EEG data is obtained by bandpass filtering (typically 0.5-45Hz, retaining the effective frequency band of EEG and filtering out low-frequency drift and high-frequency noise), power frequency notch filtering (such as 50Hz or 60Hz, to eliminate power grid interference), and artifact removal (targeted filtering out physiological interference signals such as electrooculography (EOG) and electromyography (EMG).

[0077] Eye movement signal acquisition:

[0078] Using an integrated or external video eye tracker, the pupil position coordinates, pupil diameter, blink frequency, and fixation point trajectory can be captured in real time.

[0079] Preprocessing: First, identify and label typical eye movement events such as blinking, saccades, and smooth tracking; then calculate key feature indicators such as pupil diameter change rate (reflecting cognitive load fluctuations), blink frequency (assessing fatigue level, increased frequency is often accompanied by fatigue), and fixation stability (measuring attention concentration, low stability suggests inattention).

[0080] Behavioral signal acquisition:

[0081] An inertial measurement unit (IMU) with built-in accelerometers and gyroscopes can be integrated into a head-mounted device or worn independently on the limbs (such as the wrist or torso) to collect motion data in real time.

[0082] Preprocessing: Calculate motion characteristics based on inertial measurement unit (IMU) data, including motion amplitude (reflecting the intensity of limb activity), frequency of head posture changes (excessive frequency may indicate restlessness or agitation), and motion consistency under specific tasks (such as the degree of matching between limb movements and task requirements during training).

[0083] A unified hardware clock (such as a high-precision crystal oscillator clock) is used to assign a unique timestamp to the signals from the three channels of EEG, eye movement, and behavior, ensuring that the time alignment accuracy of the three reaches the millisecond level; if the synchronization deviation is too large, it will directly lead to the distortion of the fusion result.

[0084] The multimodal feature extraction module is configured to analyze EEG signals, eye movement signals, and behavioral signals respectively to obtain EEG abnormality coefficients, eye movement abnormality coefficients, and behavioral abnormality coefficients.

[0085] The process of obtaining the EEG abnormality coefficient includes:

[0086] The raw EEG signal (sampling rate set to Fs=250Hz) is segmented (each segment has N=256 sampling points, approximately 1s), a Hanning window is added (to reduce spectral leakage), and then the spectrum is obtained through Fast Fourier Transform (FFT).

[0087] This process is a direct reference to existing technology and will not be elaborated upon here;

[0088] The power of a frequency band (such as Beta wave 13-30Hz) is the sum of the power of all frequency points within that band;

[0089] Relative power is the proportion of the power in this frequency band to the total power of the entire frequency band (0.5-45Hz);

[0090] Pre-collecting user baseline data to obtain the Beta wave relative power reference value and standard deviation The deviation of the current Beta wave power is calculated using Z-score and recorded as the initial EEG coefficient. The initial EEG coefficient value is less than 0, indicating insufficient focus, and greater than 0, indicating fatigue tension.

[0091] Deviation of current Beta wave power: ;in, This represents the relative power (percentage of total power across the entire frequency band) of the Beta wave (13-30Hz) during the current acquisition period. The baseline mean of the relative power of the Beta wave under the target cognitive state; The standard deviation of the Beta wave relative to the baseline power data under the target cognitive state.

[0092] Multiple initial EEG coefficients are obtained by processing the raw EEG signals of each segment;

[0093] The normal range of the initial EEG coefficient is preset, and the initial EEG coefficient that is not within the normal range is recorded as the EEG deviation coefficient.

[0094] The EEG deviation coefficients are sorted in descending order of their numerical values ​​to obtain a set of EEG deviation coefficients. The k largest and k smallest EEG deviation coefficients are then extracted from this set to form a new set. The absolute values ​​of each coefficient in the set are taken and the mean is calculated to obtain the EEG abnormality coefficient; where k is set to 2.

[0095] Raw EEG signals are susceptible to transient artifacts (such as brief electromyographic interference) and environmental noise. The initial EEG coefficient of a single segment may not reflect the user's true cognitive state (such as occasional Beta wave fluctuations that are easily misjudged as lack of focus or fatigue).

[0096] By filtering out EEG deviation coefficients that exceed the normal range, extracting the two largest and two smallest deviation coefficients and taking the average of their absolute values, we can focus on the core features that significantly deviate from the normal baseline and eliminate the misleading effect of single data fluctuations. This ensures that the final EEG abnormality coefficient can accurately correspond to the user's continuous cognitive state abnormality, rather than transient signal noise.

[0097] On the one hand, the two types of anomalies, inattention (negative deviation) and fatigue (positive deviation), are treated by absolute value, eliminating the difference in the direction of deviation and transforming them into a directly quantifiable degree of anomaly; on the other hand, the design of fixing the k value to 2 takes into account both data representativeness and computational efficiency.

[0098] This approach avoids both introducing redundant information due to excessive data collection and lacking statistical significance due to insufficient data collection. The final output of a single mean EEG abnormality coefficient can be directly normalized and weighted with eye movement and behavioral abnormality coefficients, significantly improving the consistency and operability of multimodal coefficient fusion.

[0099] The process of obtaining the eye movement abnormality coefficient includes:

[0100] The pupil pixel diameter output by the eye tracker is converted to the actual physical diameter (pre-calibration required):

[0101] ;in, The actual physical diameter of the pupil; The pixel diameter of the pupil in the sensor image after the eye tracker images; the calibration distance is the actual distance from the user's eye to the eye tracker lens during data acquisition; the imaging sensor focal length is the optical focal length of the sensor built into the eye tracker camera.

[0102] Recorded as the actual diameter of the pupil;

[0103] Obtain the maximum and minimum values ​​of the actual physical diameter of the pupil within a preset time period, as well as the duration of that time period; subtract the minimum value from the maximum value of the actual physical diameter of the pupil within the preset time period, and then divide by the duration of that time period to obtain the pupil diameter change rate.

[0104] A preset threshold for pupil diameter change rate is set, and pupil diameter change rates that are not within the threshold are recorded as abnormal pupil diameter change values.

[0105] The pupil pixel diameter directly output by an eye tracker is greatly affected by imaging parameters such as the distance between the user and the device and the lens focal length. Pupils with the same physical diameter will show different pixel values ​​under different imaging conditions. If used directly for analysis, it can easily lead to misjudgment of cognitive state.

[0106] By using the calibration formula of pixel diameter × calibration distance / imaging sensor focal length, the pixel value is converted into the actual physical diameter, eliminating the interference of imaging conditions and enabling the pupil dynamic data to truly reflect the user's physiological state. Based on this, the pupil diameter change rate (maximum and minimum difference / duration) is calculated, which transforms the static pupil size into a dynamic change indicator, accurately linking it to the core physiological characteristics of cognitive load fluctuations, and providing objective and reliable raw data support for subsequent abnormality judgment.

[0107] Obtain all abnormal pupil diameter values ​​within a preset time period and sort them in descending order of numerical value;

[0108] Starting from the maximum pupil diameter change anomaly, the interval duration between each adjacent pupil diameter change anomaly is extracted, and the obtained durations are rearranged in the order they were obtained to form a duration set.

[0109] A preset interval duration threshold is set, and values ​​greater than the interval duration threshold are removed from the duration set to obtain a new duration set. Then, three interval durations are extracted from it in chronological order and used as the length, width, and height of a cuboid. The volume of the cuboid is calculated and recorded as the eye movement abnormality coefficient.

[0110] Abnormal values ​​of pupil diameter change are discrete single-point data that can only reflect the abnormal state at a certain moment and cannot reflect key features such as the persistence and frequency of the abnormality. However, by extracting the interval between adjacent abnormal values, we can capture the temporal distribution pattern of the abnormality (such as high-frequency continuous abnormalities and isolated occasional abnormalities). Then, by using the interval duration threshold to remove excessively long unrelated intervals, we can further focus on high-frequency abnormal features that are strongly correlated with abnormal cognitive state.

[0111] The three selected intervals are used to quantify the degree of anomaly by forming a cuboid volume. This method can also condense multi-dimensional time series information into a single geometric quantity, which not only preserves the temporal rhythm characteristics of the anomaly occurrence, but also avoids the complexity of integrating multiple data points. This allows the eye movement anomaly coefficient to more comprehensively and accurately reflect the dynamic anomaly pattern of the user's cognitive state.

[0112] The nonlinear characteristics of cuboid volume quantization can more sensitively distinguish the degree of anomalies (e.g., anomalies with short intervals and high frequencies will be manifested as smaller volumes, or conversely, the density of anomalies can be accurately mapped).

[0113] The entire process (outlier sorting, interval extraction, threshold filtering, and volume calculation) is based on clear rules and measurable time parameters, with no subjective judgment steps, ensuring that the calculation results are consistent across different devices and user scenarios. At the same time, the output single volume value can be directly normalized and fused with EEG and behavioral abnormality coefficients.

[0114] The process of obtaining the behavioral anomaly coefficient includes:

[0115] Acceleration and angular velocity data of limbs / head within a preset time period are acquired through an inertial measurement unit (IMU).

[0116] Calculate the peak triaxial acceleration: ;in, , , These are the three-axis accelerations output by the IMU;

[0117] A preset triaxial acceleration peak threshold is set, and triaxial acceleration peak values ​​that exceed the triaxial acceleration peak threshold are recorded as abnormal peak values; and the difference between the maximum abnormal peak value and the triaxial acceleration peak threshold within a preset time period is obtained to obtain the abnormal peak difference value.

[0118] By acquiring three-axis acceleration data through IMU and synthesizing the three-dimensional accelerations to extract the peak value, it avoids the limitation that single-axis data cannot reflect the overall movement of the limbs and can accurately capture the maximum movement intensity of the limbs / head. By filtering abnormal peak values ​​through preset acceleration peak thresholds and calculating the difference between the maximum abnormal peak value and the threshold (abnormal peak difference), it can not only distinguish between normal slight movements and large-amplitude interference movements, but also quantify the degree of excess of large-amplitude movements.

[0119] Obtain the Euler angles of the head / torso within a preset time period. The Euler angles include pitch angle, yaw angle, and roll angle.

[0120] Actions where the pitch angle, yaw angle, and roll angle change by more than 5° within a preset time period are marked as flagged actions.

[0121] The marking behavior and its duration corresponding to pitch angle, yaw angle and roll angle are obtained respectively and recorded as the marking duration; a standard duration is preset, and the marking duration that is less than the standard duration is recorded as the abnormal duration;

[0122] Extract the anomaly durations corresponding to pitch angle, yaw angle, and roll angle respectively;

[0123] The average duration of each anomaly corresponding to pitch angle, yaw angle and roll angle is calculated to obtain the average duration of anomalies corresponding to pitch angle, yaw angle and roll angle respectively.

[0124] After counting the number of abnormal durations of pitch angle, yaw angle, and roll angle respectively, divide them by the corresponding average abnormal duration to obtain the abnormal change frequency of pitch angle, abnormal change frequency of yaw angle, and abnormal change frequency of roll angle.

[0125] Traditional behavioral signal analysis often focuses on large-amplitude movements, but easily overlooks high-frequency micro-changes in the head / trunk (such as frequent nodding when agitated, or small head turns when distracted), which are important indicators of abnormal cognitive states. By using a precise threshold of 5° change amplitude to label postural behavior, invalid micro-fluctuations such as physiological tremors are eliminated, while meaningful postural changes are accurately identified.

[0126] Furthermore, defining the marking duration shorter than the standard duration as abnormal duration can effectively distinguish between normal deliberate posture adjustments (long-duration marking behavior) and abnormal high-frequency micro-changes (short-duration abnormal duration). Finally, by calculating the three-dimensional abnormal change frequency through the number of abnormal durations / mean, the frequency of posture micro-changes is transformed into a quantifiable indicator. The intensity of large-scale motion complements each other, constructing a complete behavioral anomaly analysis dimension of large-scale motion and posture micro-changes.

[0127] The frequency of abnormal pitch angle changes Frequency of abnormal changes in yaw angle Frequency of abnormal changes in roll angle Establish a three-dimensional Cartesian coordinate system, denoted as Cartesian space, with the X, Y, and Z axes as the axes. The coordinates of any point P in the coordinate system are: ;

[0128] Collect a predetermined number of subjects in the target training task Baseline data, calculate the mean of each dimension;

[0129] With the mean point f Centered on,

[0130] Will Mapped to real-time points in Cartesian space ;

[0131] The Euclidean distance from the real-time point to the center f of the normal region is calculated and denoted as the quantization distance. A preset quantization distance threshold is set. If the calculated quantization distance is greater than the quantization distance threshold, it is recorded as an abnormal quantization value. The calculation of the Euclidean distance is a direct reference to the prior art and will not be elaborated here.

[0132] After normalizing the abnormal peak difference value and the abnormal quantification value, the abnormal behavior coefficient is obtained by weighted summation.

[0133] A comprehensive quantitative model for behavioral anomalies involving minute attitude changes and large-amplitude movements was constructed, addressing the key pain points of fragmented multi-dimensional features and subjective quantification in traditional behavioral analysis. This claim maps the abnormal change frequencies of pitch, yaw, and roll angles to real-time points in three-dimensional Cartesian space, and transforms the three-dimensional attitude anomaly features into a single spatial geometric quantity by calculating the Euclidean distance from this point to the center of the normal baseline.

[0134] This quantification method does not require subjective allocation of weights for each angle. Instead, it objectively reflects the degree of multi-dimensional synergistic anomalies through spatial distance, thus avoiding human bias caused by weighted summation.

[0135] Meanwhile, it integrates the geometric quantization value with the abnormal peak difference value, achieving full coverage of two types of behavioral anomalies: high-frequency attitude micro-changes (three-dimensional frequency) and large-amplitude motion intensity (abnormal peak difference value). This enables the behavioral anomaly coefficient to fully characterize the user's overall behavioral state, rather than isolated local features.

[0136] The coefficient fusion module is configured to normalize the three obtained coefficients and then perform a weighted summation to obtain the comprehensive evaluation coefficient.

[0137] The feedback training module is configured to dynamically adjust the parameters of the neurofeedback task based on the comprehensive evaluation coefficient, and even switch the training paradigm to generate an adaptive neurofeedback training strategy that matches the user, ensuring that the feedback process always adapts to the user's current state.

[0138] Specifically, it includes:

[0139] The system pre-defines a three-tiered classification rule for the comprehensive evaluation coefficient and identifies the core adjustment directions for each tier:

[0140] Normal state: The core goal is to maintain training intensity and consolidate the current state. The strategy adjustment is mainly to gradually increase the difficulty in small steps, and to avoid excessive intervention that may cause fluctuations in the state.

[0141] Mild abnormality: The core goal is to correct slight deviations in the state and return to the normal training track. The strategy adjustment mainly focuses on targeted parameter fine-tuning and low-intensity intervention, and prioritizes guiding the state recovery through feedback form / frequency adjustment.

[0142] Severe anomalies: The core objectives are emergency intervention, termination of the abnormal state, and prevention of training ineffectiveness / damage. The strategy adjustment focuses on significantly reducing the difficulty / switching paradigms and providing strong multi-sensory feedback. If necessary, the training pause mechanism will be triggered.

[0143] Quantitative adjustment rules were designed for the core parameters of the neurofeedback task. All adjustment magnitudes were positively correlated with the degree of abnormality of the comprehensive evaluation coefficient. Specific dimensions included:

[0144] Task difficulty parameters:

[0145] Cognitive tasks (such as attention-focusing tasks): The target recognition threshold (such as the target stimulus presentation time) is gradually adjusted from 500ms (increased by 50ms each time in normal state, decreased by 50ms in mild abnormality, and directly reduced to 300ms in severe abnormality), and the number of interference items (increased by 1 each time in normal state, decreased by 1 in mild abnormality, and cleared to zero in severe abnormality).

[0146] Motor imagery tasks (such as feedback on limb motor imagery): accuracy threshold of imagined movements, time limit for completing the target movement;

[0147] Feedback frequency parameter:

[0148] Under normal conditions, the feedback frequency is once every 2 seconds (matching the user's cognitive rhythm); for minor anomalies, it is increased to once every 1 second (strengthening feedback guidance); for severe anomalies, it is adjusted to once every 0.5 seconds (high-frequency feedback to recall attention), while reducing the complexity of feedback information (only retaining core status prompts);

[0149] Target threshold parameter:

[0150] Based on baseline thresholds of EEG Beta waves, eye movement fixation stability, and behavioral motion amplitude, the task achievement threshold is dynamically adjusted.

[0151] It incorporates three core neurofeedback training paradigms and pre-sets clear switching trigger conditions. The switching logic combines a comprehensive evaluation coefficient and the dominant abnormal modality (EEG / eye movement / behavior).

[0152] Visual guidance paradigm (suitable for mild EEG attention deficit and stable eye movement fixation): Using dynamic visual targets on the screen (such as floating balls or progress bars) as the core feedback carrier, the intensity of EEG Beta waves is fed back through the target's motion state (such as the ball rising / falling). It is suitable for users with normal or mild abnormal states and normal visual perception.

[0153] Auditory guidance paradigm (suitable for eye movement fatigue (high blinking frequency) and brain electrical fatigue tension): switch to natural soundscape / pure tone feedback, use tone frequency / soundscape loudness to provide feedback on relaxation (Alpha wave intensity), and disable high-frequency visual stimulation;

[0154] Motor calibration paradigm (adapted to high-frequency micro-changes in behavior and posture, mild abnormalities in EEG / eye movement): Pause cognitive feedback tasks and switch to limb / head posture calibration tasks (such as guiding users to maintain a head pitch angle of ±2° and a yaw angle of ±2°), and use tactile vibration to provide feedback on posture deviations.

[0155] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0156] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0157] It should be noted that, in this document, the use of relational terms such as "first" and "second" is merely for distinguishing one entity or operation from another, and does not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "include," "contain," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the statement "includes a…" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0158] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0159] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0160] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0161] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0162] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0163] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0164] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. An adaptive closed-loop brain-computer interface neurofeedback training system, characterized in that, include: The multimodal signal acquisition module is configured to acquire EEG signals, eye movement signals, and behavioral signals, and performs preliminary preprocessing to eliminate noise and unify the time reference. The multimodal feature extraction module is configured to analyze EEG signals, eye movement signals, and behavioral signals respectively to obtain EEG abnormality coefficients, eye movement abnormality coefficients, and behavioral abnormality coefficients, specifically including: The process of obtaining EEG abnormality coefficients includes: The raw EEG signal is segmented, a Hanning window is added, and then the spectrum is obtained through fast Fourier transform. The power of a frequency band is the sum of the power of all frequency points within that band. Relative power is the proportion of the power in that frequency band to the total power of the entire frequency band; Pre-collect user baseline data to obtain the relative power benchmark value and standard deviation of Beta waves, calculate the deviation of the current Beta wave power using Z-score, and record it as the initial EEG coefficient; Also includes: The normal range of the initial EEG coefficient is preset, and the initial EEG coefficient that is not within the normal range is recorded as the EEG deviation coefficient. The EEG deviation coefficients are sorted in descending order of numerical value to obtain a set of EEG deviation coefficients. The k largest and k smallest EEG deviation coefficients are extracted from the set to form a new set. The absolute values ​​of each coefficient in the set are taken and the mean is calculated to obtain the EEG abnormality coefficient. The process of obtaining the eye movement abnormality coefficient includes: The pupil pixel diameter output by the eye tracker is converted into the actual physical diameter and recorded as the actual pupil diameter; Obtain the maximum and minimum values ​​of the actual physical diameter of the pupil within a preset time period, as well as the duration of that time period; subtract the minimum value from the maximum value of the actual physical diameter of the pupil within the preset time period, and then divide by the duration of that time period to obtain the pupil diameter change rate. A preset threshold for pupil diameter change rate is set, and pupil diameter change rates that are not within the threshold are recorded as abnormal pupil diameter change values. Also includes: Obtain all abnormal pupil diameter values ​​within a preset time period and sort them in descending order of numerical value; Starting from the maximum pupil diameter change anomaly, the interval duration between each adjacent pupil diameter change anomaly is extracted, and the obtained durations are rearranged in the order they were obtained to form a duration set. A preset interval duration threshold is set, and values ​​greater than the interval duration threshold in the duration set are removed to obtain a new duration set. Three interval durations are extracted from the set in chronological order and used as the length, width, and height of a cuboid. The volume of the cuboid is calculated and recorded as the eye movement abnormality coefficient. The process of obtaining the behavioral anomaly coefficient includes: Acceleration and angular velocity data of limbs / head within a preset time period are obtained through an inertial measurement unit; Calculate the peak triaxial acceleration; A preset triaxial acceleration peak threshold is set, and triaxial acceleration peak values ​​that exceed the triaxial acceleration peak threshold are recorded as abnormal peak values; and the difference between the maximum abnormal peak value and the triaxial acceleration peak threshold within a preset time period is obtained to obtain the abnormal peak difference value; Also includes: Obtain the Euler angles of the head / torso within a preset time period. The Euler angles include pitch angle, yaw angle, and roll angle. Actions where the pitch angle, yaw angle, and roll angle change by more than 5° within a preset time period are marked as flagged actions. The marking behavior and its duration corresponding to pitch angle, yaw angle and roll angle are obtained respectively and recorded as the marking duration; a standard duration is preset, and the marking duration that is less than the standard duration is recorded as the abnormal duration; Extract the anomaly durations corresponding to pitch angle, yaw angle, and roll angle respectively; The average duration of each anomaly corresponding to pitch angle, yaw angle and roll angle is calculated to obtain the average duration of anomalies corresponding to pitch angle, yaw angle and roll angle respectively. After counting the number of abnormal durations of pitch angle, yaw angle, and roll angle respectively, divide them by the corresponding average abnormal duration to obtain the abnormal change frequency of pitch angle, abnormal change frequency of yaw angle, and abnormal change frequency of roll angle. Also includes: The frequency of abnormal pitch angle changes Frequency of abnormal changes in yaw angle Frequency of abnormal changes in roll angle Establish a three-dimensional Cartesian coordinate system, denoted as Cartesian space, with the X, Y, and Z axes as the axes. The coordinates of any point P in the coordinate system are: ; Collect a predetermined number of subjects in the target training task Baseline data, calculate the mean of each dimension; with the mean point f as the center, Will Mapped to real-time points in Cartesian space ; Calculate the Euclidean distance from the real-time point to the center f of the normal area, and denot it as the quantization distance. Set a quantization distance threshold. If the calculated quantization distance is greater than the quantization distance threshold, then record it as an abnormal quantization value. After normalizing the abnormal peak difference value and the abnormal quantification value, the abnormal behavior coefficient is obtained by weighted summation. The coefficient fusion module is configured to normalize the three obtained coefficients and then perform a weighted summation to obtain the comprehensive evaluation coefficient. The feedback training module is configured to dynamically adjust the parameters of the neurofeedback task based on the comprehensive evaluation coefficient, and generate an adaptive neurofeedback training strategy that matches the user.

2. The adaptive closed-loop brain-computer interface neurofeedback training system according to claim 1, characterized in that, The multimodal signal acquisition module specifically includes: EEG signal acquisition: A multi-channel EEG electrode cap, paired with a high-performance signal amplifier, is used to acquire electroencephalogram (EEG) signals. Preprocessing: The raw EEG data is obtained through bandpass filtering, power frequency notch filtering, and artifact removal. Eye movement signal acquisition: Using an integrated or external video eye tracker, the pupil position coordinates, pupil diameter, blink frequency, and fixation point trajectory can be captured in real time. Preprocessing: Identify and label typical eye-tracking events; Behavioral signal acquisition: The inertial measurement unit, with built-in accelerometer and gyroscope, can be integrated into a head-mounted device or worn independently on the limbs to collect motion data in real time. Preprocessing: Calculate motion characteristics based on inertial measurement unit data; A unified hardware clock is used to assign a unique timestamp to the signals from the three channels: EEG, eye movement, and behavior.

Images