Body line correction and neural remodeling method and system based on eeg signal closed-loop feedback

Abstract

The application discloses a body line correction and nerve remodeling method and system based on electroencephalogram closed-loop feedback, which comprises the following steps: collecting three-dimensional space posture data of a human body in a static or dynamic state, and calculating a force line offset between a current body line and an ideal force line model, so as to adjust the current force line to a target force line state corresponding to the ideal force line model; collecting a brain cortex electric activity signal of a user, obtaining electroencephalogram data, performing feature extraction, obtaining a quantitative index representing a neural coding intensity of the user to the current force line state, mapping the quantitative index into a multi-modal feedback signal, and adjusting a mechanical guiding strategy of a flexible actuator array until the user's brain establishes a stable neural representation of the target force line. According to the technical scheme, multi-modal feedback is used to enhance the somatosensory cognition of the user, the actuator strategy is dynamically adjusted, the brain forms a stable neural representation of the correct force line, and the user can still independently maintain a good body state after leaving the device.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of medical rehabilitation devices and neuroengineering, and relates to a method and system for body alignment correction and neural remodeling based on closed-loop feedback of electroencephalogram (EEG) signals. Background Technology

[0002] Body alignment refers to the axis of gravity transmission from the top of the head, through the spine and pelvis, to the soles of the feet during static standing or dynamic movement. Misalignment of this alignment can lead to muscle compensation, abnormal joint stress, and consequently, chronic pain and motor dysfunction. Currently, techniques for correcting body alignment are mainly divided into several categories:

[0003] Passive alignment devices (posture correction braces) only provide passive mechanical restraint and lack active control over the user's neuromuscular system. After the user removes the device, the brain has not yet established a neural cognitive representation of the correct alignment, making it difficult to maintain the corrective effect and prone to relapse of alignment deviation.

[0004] Open-loop EEG biofeedback training systems limit feedback to the visualization of EEG signal characteristics alone, failing to establish a real-time mapping between EEG states and body biomechanical posture. Users cannot directly perceive and correct their body alignment through EEG feedback, resulting in a disconnect between training effectiveness and everyday postural control scenarios.

[0005] Brain-computer interface rehabilitation training systems have shown potential in motor function reconstruction, but the system output only acts on peripheral actuators (such as robotic arms and electrical stimulation electrodes), failing to form a complete two-way closed loop from peripheral mechanical state to central neural representation. Specifically, when external devices assist the limbs in completing movements, the user's brain does not receive precise proprioceptive feedback corresponding to the correct movement pattern, resulting in limited efficiency in updating central motor representation.

[0006] In summary, existing technologies in the field of body alignment correction have the following common shortcomings:

[0007] ① Force alignment correction and neural remodeling are disconnected: mechanical correction devices only act on the peripheral musculoskeletal system, and EEG training only acts on the central nervous system. The two have not established a synergistic linkage mechanism.

[0008] ② Lack of closed-loop feedback: No system can convert the corrected force line state into neural feedback signals that the brain can learn from, and it is impossible to achieve a complete learning cycle of "adjustment → perception → reinforcement → internalization";

[0009] ③ The effect is not sustainable: After the device is removed, the corrective effect will decrease over time because the central nervous system has not formed a stable neural representation of the correct force line, and repeated passive correction is required.

[0010] ④ Insufficient personalization: The existing system does not perform joint modeling based on individual EEG characteristics and force line deviation characteristics, and the correction strategy lacks the ability to be individually adapted. Summary of the Invention

[0011] The purpose of this invention is to address the aforementioned problems in existing technologies by proposing a method and system for body alignment correction and neural remodeling based on closed-loop feedback of electroencephalogram (EEG) signals.

[0012] To achieve the above objectives, the basic solution of this invention is: a method for body alignment correction and neural remodeling based on closed-loop feedback of electroencephalogram (EEG) signals, comprising the following steps:

[0013] By deploying a distributed attitude sensor array at key nodes of the human body, the three-dimensional spatial attitude data of the human body under static or dynamic conditions is collected in real time, and the force line offset between the current body force line and the ideal force line model is calculated.

[0014] Based on the force line offset, a flexible actuator array arranged in key parts of the human body is driven to apply progressive mechanical guidance or constraint to the local area of ​​the human body, adjusting the current force line to the target force line state corresponding to the ideal force line model.

[0015] During and after the force line adjustment, the user's cerebral cortex electrical activity signals are simultaneously collected by a multi-channel EEG acquisition device to obtain EEG data.

[0016] Machine learning decoding algorithms are used to extract features from EEG data and obtain quantitative indicators that characterize the intensity of neural coding of the user’s current force line state.

[0017] The quantitative indicators are mapped to multimodal feedback signals, which include visual feedback, auditory feedback and / or proprioceptive feedback generated by the flexible actuator array;

[0018] Based on multimodal feedback signals, the mechanical guidance strategy of the flexible actuator array is adjusted until the user's brain establishes a stable neural representation of the target force line.

[0019] The working principle and beneficial effects of this basic solution are as follows: This technical solution simultaneously performs physical correction of force lines and neural coding training of the brain, breaking the limitation of traditional devices that "only adjust the body and not the brain." By enhancing the user's somatosensory cognition through multimodal feedback (visual, auditory, proprioceptive), it dynamically adjusts the actuator strategy until the brain establishes a stable neural representation, ensuring long-term effects and achieving complete closed-loop control from mechanical perception to neural feedback.

[0020] Furthermore, the ideal force line model is specifically as follows:

[0021] The head and neck alignment, the midpoint of the external auditory meatus, the acromion, the greater trochanter of the femur, and the anterior edge of the lateral malleolus are all located on the same vertical line of the coronal plane;

[0022] Shoulder girdle alignment: acromion, greater tuberosity of humerus, and radial head are vertically aligned in the coronal plane.

[0023] The hip joint force line is formed by three points: the center of the hip joint (femoral head), the center of the knee joint (tibial intercondylar spine), and the center of the ankle joint (talus top).

[0024] The ankle joint alignment, when standing in a neutral position, is such that the long axis of the tibia passes through the midpoint of the talus trochlea and falls between the weight-bearing area of ​​the heel and the first half of the metatarsal bones;

[0025] The trunk midline force line and gravity line pass through the external auditory meatus, the posterior edge of the cervical vertebral body, the acromion, the anterior edge of the thoracic vertebral body, the posterior edge of the lumbar vertebral body, the posterior aspect of the hip joint, the anterior aspect of the knee joint, and the anterior aspect of the ankle joint in the sagittal plane; and through the root of the nose, the suprasternal notch, the umbilicus, the pubic symphysis, between the two knees, and between the two ankles in the coronal plane.

[0026] The patellofemoral line is the angle formed between the line connecting the center of the patella to the tibial tuberosity and the line of tension of the quadriceps femoris muscle. The normal range for men is 10°-15° and for women is 12°-18°.

[0027] The hip, knee, and ankle mechanical axis is arranged in a straight line with the femoral head center, knee joint center, and talus center. The mechanical axis passes through the knee joint center (0° position) to prevent knee valgus and varus.

[0028] Foot alignment: When standing, the angle between the long axis of the tibia and the midline of the calcaneus is 0°-4°.

[0029] It provides precise anatomical force line reference standards, covering multiple dimensions such as head, shoulder, hip, ankle, lower limb, foot, and shoulder girdle, making the system biomechanically sound and providing quantitative basis for offset calculation.

[0030] Furthermore, the method for calculating the force line offset between the current body force line and the ideal force line model is as follows:

[0031] Obtain the coordinates of key points on the human body. These coordinates are provided by a distributed attitude sensor array. We assume that the coordinates of the key points are obtained through the sensors:

[0032] Center point of external auditory meatus: ,

[0033] Anterior margin of cervical vertebral body: ,

[0034] acromion: ,

[0035] The junction of the thoracic and lumbar vertebrae:

[0036] ,

[0037] Highest point of the iliac crest: ,

[0038] greater trochanter of the femur: ,

[0039] Hip joint center: ,

[0040] Knee joint center: ,

[0041] Lateral ankle: ,

[0042] Constructing an ideal force line model:

[0043] The ideal force line of the head and neck is a straight line in space formed by the center point of the external auditory meatus and the anterior edge of the cervical vertebral body:

[0044] ,

[0045] The ideal force line of the shoulder girdle is a straight line formed by the intersection of the acromion and the thoracolumbar region:

[0046] ,

[0047] Ideal force lines for the hip, knee, and ankle are constructed by segmenting two spatial straight lines:

[0048] Hip to knee joint segment:

[0049] ,

[0050] From the knee to the lateral malleolus: ,

[0051] The ideal coronal plane is a standard vertical reference plane formed by the acromion, the highest point of the iliac crest, the greater trochanter of the femur, and the lateral malleolus.

[0052] Let the coordinates of each key point in the real-time human posture be:

[0053] The corresponding real-time body force line is: N sampling points are uniformly selected along a single real-time force line, and the spatial distance d from each sampling point to the corresponding ideal force line is calculated. i The average value is taken as the offset of a single force line:

[0054] The total overall posture offset is obtained by weighted summing of the force line offsets of the head, shoulder girdle, and lower limb segments: ,

[0055] in, , , The weighting coefficients satisfy the following constraints: + + =1.

[0056] It calculates the force line offset between the current body force line and the ideal force line model. The operation is simple and easy to use.

[0057] Furthermore, based on the force line offset, a flexible actuator array arranged at key parts of the human body is driven to apply progressive mechanical guidance or constraint to the local area of ​​the human body, adjusting the current force line to the target force line state. The specific steps are as follows:

[0058] On the medial side of the foot arch (navicular-subtalar joint area), use a flat pneumatic braided array to pull the arch fascia upward and inward, guiding the calcaneus to 0-4° eversion, correcting foot eversion / medial longitudinal arch collapse, and restoring the calcaneus to neutrality; apply 1N, 3N, and 5N sequentially, with an increment of 0.5N each time, increasing every 5 minutes;

[0059] Apply oblique tightening force from the superior lateral aspect of the patella to the oblique bundle of the quadriceps femoris using an arc-shaped flexible SMA braided band, reducing the Q angle beyond the preset range to prevent patellar lateral displacement; apply 0N, 2N, and 4N sequentially, activating in stages according to the gait cycle, with activation during the standing phase;

[0060] The lumbar spiral, the contralateral gluteus maximus, quadratus lumborum, latissimus dorsi fascia bridge, and the use of a cross-shaped elastic-pneumatic hybrid array (oblique along the thoracolumbar fascia) to apply rotational shear force from the contralateral posterior superior iliac spine to the ipsilateral inferior angle of the scapula to reconstruct the spiral energy storage and transmission efficiency; low pretension (1N) is applied, which is increased to 6N only during trunk rotation / walking.

[0061] Apply gentle pressure (without actively pulling the head) to the deep flexor muscles of the head and neck, specifically the projection area of ​​the longus colli / longus capitis muscles, using an ultra-thin dielectric elastomer patch (fitted to the submandibular triangle area) in a posterior and inferior direction to restore the cervical spine to a neutral position (vertical line from the tragus to the acromion); maintain a pressure of 0.8-1.5N, with only tactile feedback; reduce to 0.3N after more than 30 minutes.

[0062] Mechanical guidance is applied to the arch of the foot, patella, lumbar spiral, and deep flexor muscles of the head and neck in a phased, segmented manner, using a progressive increment to avoid defensive muscle contraction and improve comfort and compliance.

[0063] Furthermore, machine learning decoding algorithms are used to extract features from EEG data to obtain quantitative indicators representing the intensity of neural coding related to the user's current force line state. The specific method is as follows:

[0064] Collect the target force line, introduce known deviations, and mark the sensing time points;

[0065] Configure the EEG channel;

[0066] Offline preprocessing of EEG data;

[0067] Three types of features were extracted from the EEG data, including ERN (Error-Related Negative Wave) amplitude, theta ERS (Event-Related Synchronization), and wPLI (Weighted Phase Lag Index) (C3-CPz).

[0068] Training the ridge regression model;

[0069] The posture sensor continuously collects the coordinates of key points on the human body, compares them with the ideal force line model, and calculates the force line offset. If the offset exceeds the allowable threshold, a start signal is sent to the integrated real-time system to correct the target parameters and formally trigger the entire workflow.

[0070] Integrated Real-Time System: This includes configuring a real-time operating system foundation, EEG acquisition thread, feature extraction thread, model inference thread, actuator control thread, synchronization and latency measurement, integrity verification and debugging, specifically:

[0071] Upon receiving the start command, the integrated real-time system immediately activates the EEG hardware acquisition channel to synchronously record the subject's real-time EEG signals, aligning them with the posture sensor timestamps throughout the process to ensure a one-to-one correspondence between "posture state" and "EEG signal".

[0072] The raw EEG was filtered and preprocessed to remove artifacts, and neural features were extracted using a sliding window.

[0073] The features are fed into the offline-trained regression model to calculate the subject's current neural feedback state.

[0074] By combining the correction target given by the force line and the feedback value output by the regression model, actuator control commands are dynamically generated.

[0075] An integrated real-time system drives the actuator to complete posture correction; the posture sensor continuously refreshes the force line offset, and the EEG device continuously samples.

[0076] Repeat this process until the force line offset returns to the acceptable range, then stop data acquisition and correction.

[0077] Three types of neurophysiological features, ERN, theta ERS, and wPLI, are extracted. A regression model is trained to map EEG features to actuator gains (0-1), achieving real-time closed-loop drive with a latency of <100ms.

[0078] Furthermore, the quantitative indicators are mapped to multimodal feedback signals, including:

[0079] Real-time neural bias coding strength: This refers to the internal confidence level of the brain's continuous calculation of "the current posture deviates from the ideal force line" through multi-region neural activity.

[0080] Urgency of corrective needs: i.e., the need to correct biomechanical deviations;

[0081] Mapping is performed through vision (AR glasses / ambient light), hearing (speakers / headphones), and proprioception (flexible actuator array);

[0082] The warning phase uses visual and auditory senses, the corrective execution phase uses proprioception, hearing, and vision, and the successful reversion phase uses hearing and vision.

[0083] By distinguishing between two types of feedback logic—neural bias coding intensity and biomechanical correction needs—and switching feedback modalities (visual / auditory / proprioceptive) according to the three stages of warning, correction execution, and successful regression, training efficiency can be improved.

[0084] The present invention also provides a body alignment correction and neural remodeling system based on the method described herein, comprising:

[0085] The force line sensing module consists of a distributed attitude sensor array arranged at key nodes of the human body, used to collect human body three-dimensional spatial attitude data in real time and calculate force line offset.

[0086] The force line adjustment module consists of a distributed array of flexible actuators arranged in key parts of the human body. It is used to apply mechanical guidance or constraint to local parts of the human body according to control commands and adjust the body's force line.

[0087] The EEG acquisition module consists of a multi-channel non-invasive EEG electrode array, used to synchronously acquire electrical activity signals from the user's cerebral cortex;

[0088] The central processing module is connected to the force line sensing module, the force line adjustment module, and the EEG acquisition module. The central processing module receives EEG data, processes the data, and adjusts the mechanical guidance strategy of the flexible actuator array until the user's brain establishes a stable neural representation of the target force line.

[0089] This system integrates four major modules: force line sensing, force line adjustment, EEG acquisition, and central processing, to fully realize closed-loop neural remodeling function. The central processing module uniformly schedules data and control commands to ensure the system's real-time performance and coordination.

[0090] Furthermore, the distributed attitude sensor array includes a three-axis accelerometer, a gyroscope, and a magnetometer, and the key human body nodes include the head, shoulders, thoracic vertebrae, lumbar vertebrae, pelvis, knee joint, and ankle joint;

[0091] The key parts of the human body include the shoulders, thorax, pelvic girdle, and the inner and outer sides of the knee joints.

[0092] Clearly define sensor nodes (head, shoulder, thoracic spine, lumbar spine, pelvis, knee, ankle) and actuator locations (shoulder, thorax, pelvis, knee) to ensure complete coverage of whole-body force line data and precise control of key correction points.

[0093] Furthermore, the distributed flexible actuator array includes adjustable straps driven by micro linear motors, pneumatic artificial muscles, and shape memory alloy wire actuators.

[0094] Employing a variety of flexible actuators such as miniature linear motors, pneumatic artificial muscles, and shape memory alloys, it provides diverse mechanical output methods to adapt to the mechanical needs of different parts and training stages.

[0095] Furthermore, the central processing module includes:

[0096] Force line calculation engine, used to calculate force line offset based on attitude data;

[0097] Adjust the control engine to generate control commands for the force line adjustment module;

[0098] The EEG decoding engine is used to decode and analyze EEG data and output a quantitative index of neural coding strength.

[0099] A feedback mapping engine is used to map the quantification metrics to multimodal feedback parameters.

[0100] Define the force line calculation engine, adjustment control engine, EEG decoding engine, and feedback mapping engine to achieve full-link automated processing of data fusion, decision-making, execution, and feedback.

[0101] Furthermore, it also includes a feedback output module, as well as a data storage and communication module;

[0102] The feedback output module includes a visual feedback device, an auditory feedback device, and a haptic feedback module, and the feedback output module is connected to the central processing module.

[0103] The data storage and communication module includes a local solid-state storage array and a wireless communication unit, wherein the local solid-state storage array is connected to the central processing module through the wireless communication unit.

[0104] Added feedback output modules (visual, auditory, haptic) and data storage and communication modules to support local data recording and remote transmission, providing a data foundation for model iteration and personalized training. Attached Figure Description

[0105] Figure 1 This is a flowchart illustrating the body alignment correction and neural remodeling method based on closed-loop feedback of electroencephalogram (EEG) signals according to the present invention. Detailed Implementation

[0106] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0107] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0108] In the description of this invention, unless otherwise specified and limited, it should be noted that the terms "installation", "connection" and "linking" should be interpreted broadly. For example, they can refer to mechanical or electrical connections, or internal connections between two components. They can be direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances.

[0109] This invention discloses a method for body alignment correction and neural remodeling based on closed-loop feedback of electroencephalogram (EEG) signals, used to achieve automated assessment and adjustment of human body alignment, synchronous acquisition and decoding analysis of EEG signals, and somatosensory cognitive remodeling based on neural feedback. Figure 1 As shown, the body alignment correction and neural remodeling method based on closed-loop feedback of EEG signals includes the following steps:

[0110] By deploying a distributed attitude sensor array at key nodes of the human body, the system collects three-dimensional spatial attitude data of the human body in static or dynamic states in real time, and calculates the force line offset between the current body force line and the ideal force line model, including multi-dimensional parameters such as coronal plane offset angle, sagittal plane tilt angle, and rotation angle.

[0111] Based on the force line offset, a flexible actuator array arranged in key parts of the human body is driven to apply progressive mechanical guidance or constraint to the local area of ​​the human body, adjusting the current force line to the target force line state corresponding to the ideal force line model.

[0112] During and after the force line adjustment, the user's cerebral cortex electrical activity signals are simultaneously collected by a multi-channel EEG acquisition device to obtain EEG data.

[0113] Machine learning decoding algorithms are used to extract features from EEG data and obtain quantitative indicators that characterize the intensity of neural coding of the user’s current force line state.

[0114] The quantitative indicators are mapped to multimodal feedback signals, including visual feedback, auditory feedback, and / or proprioceptive feedback generated by the flexible actuator array; the EEG decoding results are mapped to multimodal feedback signals—when the decoding results show that the user's brain produces stable neural representations of the correct force lines, the system generates positive reinforcement signals (such as pleasant auditory cues, positive changes in visual images, and a continuous sense of support from external devices); conversely, when the neural representations are unstable, the system outputs guiding feedback signals.

[0115] Based on multimodal feedback signals, the mechanical guidance strategy of the flexible actuator array is adjusted until the user's brain establishes a stable neural representation of the target force line.

[0116] The decoding score is mapped to visual feedback parameters; the higher the score, the closer the force line trajectory on the screen is to the ideal curve, and the color gradually changes from red to green. Simultaneously, it is mapped to auditory feedback: a positive prompt sound is triggered when the score exceeds a threshold of 0.7. If the score remains below the threshold for more than 5 seconds, the system automatically fine-tunes the actuator position to guide the user's attention.

[0117] In a preferred embodiment of the present invention, the ideal force line model is specifically as follows:

[0118] The head and neck alignment, the midpoint of the external auditory meatus, the acromion, the greater trochanter of the femur, and the anterior edge of the lateral malleolus are all located on the same vertical line of the coronal plane;

[0119] Shoulder girdle alignment: acromion, greater tuberosity of humerus, and radial head are vertically aligned in the coronal plane.

[0120] The hip joint force line is formed by three points: the center of the hip joint (femoral head), the center of the knee joint (tibial intercondylar spine), and the center of the ankle joint (talus top).

[0121] The ankle joint alignment, when standing in a neutral position, is such that the long axis of the tibia passes through the midpoint of the talus trochlea and falls between the weight-bearing area of ​​the heel and the first half of the metatarsal bones;

[0122] The trunk midline force line and gravity line pass through the external auditory meatus, the posterior edge of the cervical vertebral body, the acromion, the anterior edge of the thoracic vertebral body, the posterior edge of the lumbar vertebral body, the posterior aspect of the hip joint, the anterior aspect of the knee joint, and the anterior aspect of the ankle joint in the sagittal plane; and through the root of the nose, the suprasternal notch, the umbilicus, the pubic symphysis, between the two knees, and between the two ankles in the coronal plane.

[0123] The patellofemoral line is the angle formed between the line connecting the center of the patella to the tibial tuberosity and the line of tension of the quadriceps femoris muscle. The normal range for men is 10°-15° and for women is 12°-18°.

[0124] The hip, knee, and ankle mechanical axis is arranged in a straight line with the femoral head center, knee joint center, and talus center. The mechanical axis passes through the knee joint center (0° position) to prevent knee valgus and varus.

[0125] Foot alignment: When standing, the angle between the long axis of the tibia and the midline of the calcaneus is 0°-4°.

[0126] Based on the TensorFlow 2.12 framework, a pre-trained EEG-Conformer model (an EEG decoding model combining convolutional neural networks and Transformer architecture) was deployed. The model input consisted of 32 channels of EEG data (500 sampling points) within a 1-second time window, and the outputs were a "somatosensory attention concentration" score (a continuous value of 0-1) and a "body schema matching" score (a continuous value of 0-1). During the training phase, the model underwent supervised learning training using labeled data (EEG data when the user actively focused on the correct force line was used as positive samples, and distracted or unfocused states were used as negative samples).

[0127] In a preferred embodiment of the present invention, the method for calculating the force line offset between the current body force line and the ideal force line model is as follows:

[0128] Obtaining the coordinates of key points on the human body: Obtaining the coordinates of bony landmarks on the body surface: By using distributed posture sensors worn on the head, acromion, iliac crest, greater trochanter of the femur, lateral malleolus, and other parts of the human body, the three-dimensional coordinates of measurable bony landmarks on the body surface, such as the center point of the external auditory meatus, acromion, the highest point of the iliac crest, greater trochanter of the femur, and lateral malleolus, can be directly detected.

[0129] Coordinates of key anatomical points within the body were obtained indirectly based on the coordinates of nearby bony landmarks collected by surface sensors, combined with standard human anatomical parameters determined through previous anatomical calibration experiments. Specifically, the anterior margin of the cervical vertebral body was calculated from the surface coordinates of the spinous process of the 7th cervical vertebra, combined with cervical anteroposterior offset parameters; the thoracolumbar junction was determined by the midpoint between the coordinates of the spinous processes of the 12th thoracic vertebra and the 1st lumbar vertebra; and the center of the femoral head was calculated by inversely using pelvic posture data, the coordinates of the greater trochanter of the femur, and femoral anatomical length parameters.

[0130] Center point of external auditory meatus: ,

[0131] Anterior margin of cervical vertebral body: ,

[0132] acromion: ,

[0133] The junction of the thoracic and lumbar vertebrae:

[0134] ,

[0135] Highest point of the iliac crest: ,

[0136] greater trochanter of the femur: ,

[0137] Hip joint center: ,

[0138] Knee joint center: ,

[0139] Lateral ankle: ,

[0140] Constructing an ideal force line model:

[0141] The ideal force line of the head and neck is a straight line in space formed by the center point of the external auditory meatus and the anterior edge of the cervical vertebral body:

[0142] ,

[0143] The ideal force line of the shoulder girdle is a straight line formed by the intersection of the acromion and the thoracolumbar region:

[0144] ,

[0145] Ideal force lines for the hip, knee, and ankle are constructed by segmenting two spatial straight lines:

[0146] Hip to knee joint segment:

[0147] ,

[0148] From the knee to the lateral malleolus: ,

[0149] The ideal coronal plane is a standard vertical reference plane formed by the acromion, the highest point of the iliac crest, the greater trochanter of the femur, and the lateral malleolus.

[0150] Let the coordinates of each key point in the real-time human posture be:

[0151] The corresponding real-time body force line is: N sampling points are uniformly selected along a single real-time force line, and the spatial distance d from each sampling point to the corresponding ideal force line is calculated. i The average value is taken as the offset of a single force line:

[0152] The total overall posture offset is obtained by weighted summing of the force line offsets of the head, shoulder girdle, and lower limb segments: ,

[0153] in, , , The weighting coefficients satisfy the constraints. + + =1, the specific value can be determined through clinical statistical samples.

[0154] In a preferred embodiment of the present invention, based on the force line offset, a flexible actuator array arranged at key parts of the human body is driven to apply progressive mechanical guidance or constraint to the local area of ​​the human body, adjusting the current force line to the target force line state, so that the offset of a single force line is zero. The specific steps are as follows:

[0155] On the medial side of the foot arch (navicular-subtalar joint area), use a flat pneumatic braided array to pull the arch fascia upward and inward, guiding the calcaneus to 0-4° eversion, correcting foot eversion / medial longitudinal arch collapse, and restoring the calcaneus to neutrality; apply 1N, 3N, and 5N sequentially, with an increment of 0.5N each time, increasing every 5 minutes;

[0156] Apply oblique tightening force from the superior lateral aspect of the patella towards the oblique bundle of the quadriceps femoris using a curved flexible SMA braided band, reducing the Q angle beyond the preset range (male: 10° - 15°; female: 12° - 18°) to prevent patellar lateral displacement; apply 0N, 2N, and 4N sequentially, activating in stages according to the gait cycle, with activation during the standing phase;

[0157] The lumbar spiral, the contralateral gluteus maximus, quadratus lumborum, latissimus dorsi fascia bridge, and the use of a cross-shaped elastic-pneumatic hybrid array (oblique along the thoracolumbar fascia) to apply rotational shear force from the contralateral posterior superior iliac spine to the ipsilateral inferior angle of the scapula to reconstruct the spiral energy storage and transmission efficiency; low pretension (1N) is applied, which is increased to 6N only during trunk rotation / walking.

[0158] Apply gentle pressure (without actively pulling the head) to the deep flexor muscles of the head and neck, specifically the projection area of ​​the longus colli / longus capitis muscles, using an ultra-thin dielectric elastomer patch (fitted to the submandibular triangle area) in a posterior and inferior direction to restore the cervical spine to a neutral position (vertical line from the tragus to the acromion); maintain a pressure of 0.8-1.5N, with only tactile feedback; reduce to 0.3N after more than 30 minutes.

[0159] In a preferred embodiment of the present invention, a machine learning decoding algorithm is used to extract features from electroencephalogram (EEG) data to obtain a quantitative index representing the intensity of neural coding on the user's current force line state. The specific method is as follows:

[0160] Collect the target force line, introduce known deviations, and mark the sensing time points;

[0161] Configure EEG channels: focus on covering Cz, C3, CPz, and Fz, with a sampling rate ≥ 500 Hz;

[0162] Offline preprocessing of EEG data: filtering, artifact removal, segmentation, and baseline correction;

[0163] Three types of features were extracted by analyzing EEG data, including: ERN (Error-Related Negative Wave) amplitude (Cz / CPz, 200-400ms), theta ERS (Event-Related Synchronization) (FCz, 4-8 Hz), and wPLI (Weighted Phase Lag Index) (C3-CPz).

[0164] Training Ridge Regression Model: Collect labeled training data, perform feature preprocessing, divide into training / validation / test sets, train the ridge regression model based on the training set, perform model evaluation and selection, export the model, extract features in real time, and obtain the gain through model inference and send it to the adjustment control engine.

[0165] The attitude sensor continuously collects the coordinates of key points on the human body, compares them with the ideal force line model, and calculates the force line offset. As long as the offset exceeds the allowable threshold, it sends a start signal to the integrated real-time system to correct the target parameters and officially trigger the entire workflow.

[0166] The system integrates a real-time system, configures the basic real-time operating system, and includes EEG acquisition threads, feature extraction threads, model inference threads, actuator control threads, synchronization and latency measurement, integrity verification and debugging, specifically:

[0167] Upon receiving the start command, the integrated real-time system immediately activates the EEG hardware acquisition channel to synchronously record the subject's real-time EEG signals; the timestamps are aligned with the posture sensor timestamps throughout the process (allowing for a certain error, such as 0.5ms), ensuring a one-to-one correspondence between "posture state" and "EEG signal".

[0168] The raw EEG was filtered and preprocessed to remove artifacts, and neural features were extracted using a sliding window.

[0169] The features are fed into the offline-trained regression model to calculate the subject's current neural feedback state.

[0170] By combining the correction target given by the force line and the feedback value output by the regression model, actuator control commands are dynamically generated.

[0171] An integrated real-time system drives the actuator to complete posture correction; the posture sensor continuously refreshes the force line offset, and the EEG device continuously samples.

[0172] Repeat this process until the force line offset returns to the acceptable range, then stop data acquisition and correction.

[0173] When configuring an integrated real-time system, select and install a real-time kernel, isolate CPU cores and set affinity, increase thread priority, lock memory and disable swapping.

[0174] The EEG acquisition thread initializes the hardware by pre-allocating two equal-sized buffers in memory. The hardware writes data directly to one of these buffers via DMA. Once the buffer is full, an interrupt is triggered, switching buffers and notifying the software to handle the full buffer. The thread then enters a real-time loop, waiting for a hardware data ready signal. It obtains the current timestamp (microseconds or nanoseconds), encapsulates it along with a data block, and places the data block into a lock-free circular buffer. This buffer is used to pass the data to the feature extraction thread. Enqueue operations must be non-blocking. If the buffer is full, the oldest data block is discarded, and a semaphore is released to notify the feature extraction thread of new data. If data cannot be read consecutively or a hardware error occurs, an error flag is set, the user is notified via a separate monitoring thread, and an attempt is made to reinitialize the hardware.

[0175] The feature extraction thread maintains a global circular buffer internally to store the most recent second of raw EEG data. This buffer is at least one second long and covers the maximum time window required for feature extraction. Each time, one or more data blocks are retrieved from the circular buffer of the EEG thread and appended to the sliding window buffer in chronological order, while discarding older data exceeding one second. The extracted data segments are bandpass filtered (1-40Hz) using an IIR filter while maintaining the filter's internal state to ensure filtering continuity between continuous data streams. The mean of the data segments is calculated and subtracted, or a moving average is used for detrending. The maximum-minimum range of the segments is calculated; if it exceeds 150 microvolts (…),… If the issue is due to eye movement or electromyography (EMG) contamination, the previous feature vector is reused without further calculations. Fast ICA or thresholding is used to reject artifacts. A 200-400ms time window (corresponding to sampling point indices 100-200) is extracted from the Cz and CPz channels. The minimum voltage value (negative peak) within this range is found and subtracted from the baseline (average value from -200 to 0ms) to obtain the ERN amplitude. The Welch method (fast Fourier transform piecewise averaging) is used to calculate the power spectral density of the FCz channel. The average power in the 4-8Hz band is extracted, divided by the pre-stored resting-state baseline power (calculated from seated data collected before the experiment), and then subtracted by 1 to obtain the percentage change in power. The consistency of the C3 and CPz channels in the theta / alpha band is calculated. First, a Hilbert transform is performed on the two channel signals to obtain the instantaneous phase. The sign of the sine of the phase difference is calculated, and then a weighted average is performed to obtain the connection strength between 0 and 1.

[0176] The model inference thread loads the pre-trained ridge regression model during system startup. For the ridge regression model, the weight vector and bias can be parsed and implemented directly in the code using matrix multiplication. It performs an empty inference or warms up using typical feature vectors to ensure the model file is cached, allocates memory for input and output tensors, and blocks waiting for new feature vectors in the feature extraction thread queue with a timeout (e.g., 100ms). If no new features are added within the timeout, the previous inference result is used to prevent the executor from stopping output. The feature vectors are standardized by subtracting the mean and dividing by the standard deviation. The mean and standard deviation are derived from training statistics and must be embedded in the code or read from a configuration file. The system performs inference, performs temporal smoothing on the output gain, and places the smoothed gain value (range 0-1) into the queue of the executor control thread. It records timestamps at the inference entry and exit points, calculates the time taken for a single inference, and if it exceeds 30ms (which may be slower for neural networks), triggers a warning and considers reducing the frequency of subsequent inferences or switching to a lightweight model. It periodically (every second) calculates the average gain value of the past 100 inferences. If the gain is consistently higher than 0.8 and there is no biomechanical bias, model drift may exist, prompting the user to recalibrate. If the model inference fails consecutively (e.g., outputting NaN or outliers), or if the delay exceeds 50ms for 10 consecutive times, it automatically switches to safe mode: ignoring EEG input and calculating the correction force solely based on biomechanical sensors, while simultaneously sending an error report.

[0177] The actuator control thread initializes the driver board (such as a PWM generator or DAC) of the flexible actuator array, sets the PWM frequency (typically 200-500Hz) and resolution (8-12 bits); sets the output of all actuators to 0, starts an independent high-priority timer, checks every 100ms whether the actuator thread is updating instructions normally, and automatically cuts off the actuator power if there is no update within 500ms; in addition to receiving the EEG inference gain, it also needs to obtain the biomechanical deviation from the force line calculation engine, which is calculated based on sensors such as IMU and plantar pressure, reflecting the actual physical deviation; according to the spatial distribution of the force line deviation (e.g., foot eversion, knee valgus, etc.), the comprehensive gain is allocated to the actuators at different positions; the allocation coefficient is predefined in the configuration table or dynamically calculated based on real-time inverse kinematics, mapping the target tension of each actuator to the PWM duty cycle (0-1). If the actuator has nonlinear force-duty cycle characteristics, it is compensated using a lookup table or a polynomial calibration curve.

[0178] Synchronization and latency measurement involves recording high-precision timestamps at key points in the data stream. These markers include: EEG sampling time (when triggered by hardware), data block enqueue time, feature extraction start / end time, inference end time, and executor command transmission time. A data structure with a timestamp array is designed and transmitted along with the data stream. At each processing stage, the current timestamp is appended to the array. The total latency is obtained by subtracting the corresponding EEG sampling time from the executor command transmission timestamp. Latency is calculated in stages: feature extraction time = feature end - feature start; inference time = inference end - feature end; control transmission time = command transmission - inference end. Latency data is reported to the user interface or storage via an independent low-priority thread. The current average latency and maximum latency are displayed on the interface. A first warning signal is displayed if the average latency exceeds 80ms, and a second warning signal is displayed if it exceeds 100ms.

[0179] For integrity verification and debugging, a simulated EEG data generator was used to test the correctness of feature extraction, model inference, and actuator instruction mapping sequentially. The entire system was run without connecting to a real EEG and actuator. Using data playback or a sine wave generator as EEG input, deadlocks or queue overflows were checked in data transfer between threads, along with CPU utilization and memory leaks. An EEG simulator (capable of generating standard test signals) and a dummy load (replacing the actuator) were connected to verify the complete closed loop from EEG to actuator instructions. In a controlled environment, healthy subjects wore the system, and a pure open-loop test was conducted (the actuator was not driven, only data was recorded) to confirm that the EEG decoding gain was basically consistent with subjective feelings.

[0180] Gradually increase the actuator tension, starting from 0 and in increments of 0.5N. Have the subjects rate any discomfort to confirm that the maximum set tension is within the individual's tolerance range. Adjust the tools and logs, and generate a performance benchmark report.

[0181] In a preferred embodiment of the present invention, mapping the quantification index to a multimodal feedback signal includes:

[0182] Real-time neural deviation coding strength: that is, the confidence level of the user's brain in detecting "a difference between the current force line and the ideal force line" (the brain continuously calculates the internal confidence level of "the current posture deviates from the ideal force line" through neural activity in multiple areas).

[0183] Urgency of correction needs: that is, the need to correct biomechanical deviations (based on the difference from the ideal force line);

[0184] Mapping is achieved through vision (AR glasses / ambient light), hearing (speakers / headphones), and proprioception (flexible actuator array) (reminding or correcting through audiovisual and actuator devices);

[0185] The warning phase uses visual and auditory senses, the corrective execution phase uses proprioception, hearing, and vision, and the successful reversion phase uses hearing and vision.

[0186] Traditional corrective devices only physically constrain body posture, and their effectiveness rapidly diminishes after the user removes the device. This invention, by simultaneously acquiring electroencephalogram (EEG) signals during the force alignment adjustment process and decoding the intensity of neural representations, guides the brain to actively establish stable neural codes for the correct force alignment, achieving "learning the correct posture" at the central level. Preliminary clinical trials show that after 20 closed-loop training sessions, the user's force alignment retention rate 72 hours after removing the device is approximately 40% higher than that of the traditional passive correction group.

[0187] Existing brain-computer interface systems are mostly unidirectional control (from brain to machine), and existing force line correction devices are open-loop constraints. This invention is the first to incorporate force line state as an input variable into the EEG acquisition context and use the EEG decoding result as a feedback signal to drive the external device to adjust its strategy, forming a complete closed-loop neural modulation chain, which conforms to the physiological principle of neural plasticity.

[0188] Traditional force line correction relies on therapists' visual assessment or users' subjective reports, lacking objective neurological evaluation indicators. This invention provides two quantitative indicators: "somatosensory attention concentration" and "body schema matching degree," which can objectively reflect the depth of neural processing of the correct force line by the user, providing data support for adjusting the training program.

[0189] Because this invention operates at the level of neural representation, the training effect has long-term enhancement characteristics. After a complete training cycle, users can maintain a relatively correct force alignment state autonomously even without wearing the device, reducing the time and economic costs of repeated treatments.

[0190] The system collects individualized force line reference values ​​and EEG characteristics through initial calibration, enabling personalized training with a "one person, one model" approach. Furthermore, the architecture of this invention can be extended to the correction of other postural problems, such as scoliosis, genu varum / valgum, and forward head posture. It can also be integrated with virtual reality technology for proprioceptive enhancement training in athletes and postural stability training for high-pressure professions such as pilots and surgeons.

[0191] The non-invasive design ensures safety and accessibility. All modules utilize non-invasive technology (dry electrode EEG acquisition, external flexible actuator), requiring no surgical implantation or drug intervention. This results in high user acceptance and can be widely adopted in rehabilitation departments, health check-up centers, gyms, and even home settings.

[0192] The present invention also provides a body alignment correction and neural remodeling system based on the method described herein, comprising:

[0193] The force line sensing module consists of a distributed array of attitude sensors deployed at key nodes of the human body. It is used to acquire real-time three-dimensional spatial attitude data of the human body and calculate force line offsets. It can employ Xsens DOT miniature inertial sensors (36×30×11mm, 11.2g), with a total of 8 nodes located at: occipital protuberance (head), C7 spinous process (cervicothoracic junction), T7 spinous process (mid-thoracic vertebra), L3 spinous process (lumbar vertebra), S1 sacrum (pelvis), greater trochanter of femur (hip), tibial tuberosity (knee), and lateral malleolus (ankle). The sensors transmit quaternion attitude data to the central processing module via Bluetooth 5.0 at a frequency of 100Hz.

[0194] The force line adjustment module consists of a distributed array of flexible actuators positioned at key points on the body. It applies mechanical guidance or constraint to specific areas of the body according to control commands, adjusting the body's force line. A customized flexible corrective vest can be used, integrating four miniature linear stepper motors (Haydon Kerk 19000 series, 50mm stroke, maximum thrust 80N), positioned behind both shoulders and at both anterior superior iliac spines. The motors drive flexible fabric straps, adjusting the shoulder abduction angle and pelvic tilt angle through contraction / release. The motor driver uses a Trinamic TMC5130 and communicates with the central module via a CAN bus.

[0195] The EEG acquisition module consists of a multi-channel non-invasive EEG electrode array used to simultaneously acquire electrical activity signals from the user's cerebral cortex. The number of electrodes ranges from 8 to 64 channels, arranged according to the international 10-20 or 10-10 system, focusing on covering the sensorimotor cortex (C3, C4, Cz), the parietal somatosensory association area (P3, P4, Pz), and the prefrontal cortex (F3, F4, Fz). Signal acquisition parameters include a sampling rate of 250-1000Hz, bandpass filtering of 0.1-100Hz, and a notch filter to suppress 50Hz power frequency interference. The acquisition module transmits raw EEG data to the central processing module in real time via USB or PCIe interface. The module's function is to acquire cortical electrical activity reflecting the brain's processing of somatosensory and body schemata. Preferably, a 32-channel active dry electrode EEG cap (g.tecg.Nautilus PRO) is used, with electrodes arranged according to the international 10-20 system, focusing on covering C3, C4, Cz, P3, P4, Pz, and Fz. The sampling rate is 500Hz, the ADC resolution is 24-bit, and the data is wirelessly transmitted to the central processing module.

[0196] The central processing module is connected to the force line sensing module, the force line adjustment module, and the EEG acquisition module. The central processing module receives EEG data, processes the data, and adjusts the mechanical guidance strategy of the flexible actuator array until the user's brain establishes a stable neural representation of the target force line.

[0197] The central processing unit consists of a high-performance embedded processor or industrial control computer, including: a multi-core CPU (such as Intel Core i7 or ARM Cortex-A76 and above), a GPU acceleration unit (such as NVIDIA Jetson series), memory (at least 8GB), and solid-state storage (at least 256GB). An industrial embedded computer (Advantech ARK-3532, Intel Core i7-10700, 32GB DDR4, NVIDIA RTX A2000 GPU, 1TB NVMe SSD) can be used. It runs the Ubuntu 20.04 LTS operating system.

[0198] In a preferred embodiment of the present invention, the distributed attitude sensor array includes a three-axis accelerometer, a gyroscope and a magnetometer, and the key human body nodes include the head, shoulders, thoracic vertebrae, lumbar vertebrae, pelvis, knee joint and ankle joint; the key human body parts include the shoulders, thorax, pelvic girdle and inner and outer sides of the knee joint.

[0199] The sensor nodes are electrically connected to the central processing module via flexible wires or Bluetooth 5.0 wireless protocol. Each sensor has a sampling frequency of 100-200Hz, and the data is output in quaternion or Euler angle format. The module's function is to acquire the spatial orientation information of each segment of the body in real time, providing raw data for force line calculation.

[0200] Clearly define sensor nodes (head, shoulder, thoracic spine, lumbar spine, pelvis, knee, ankle) and actuator locations (shoulder, thorax, pelvis, knee) to ensure complete coverage of whole-body force line data and precise control of key correction points.

[0201] Preferably, the distributed flexible actuator array includes adjustable straps driven by miniature linear motors, pneumatic artificial muscles, and shape memory alloy wire actuators. The actuators are positioned at key points for force line adjustment, such as the shoulder, thorax, pelvic girdle, and inner and outer sides of the knee joint. Each actuator connects to the central processing module via a CAN bus or RS-485 bus, receiving position / force control commands and providing feedback on the actual execution status. The module's function is to apply precise position guidance or mechanical support to specific body parts according to control commands.

[0202] For the control of flexible actuator arrays (micro linear motors, pneumatic artificial muscles, SMA actuators), an existing event-driven progressive control strategy based on finite state machines can be adopted. The control rhythm and operation are jointly determined by the "force line offset" and "usage time". The force line offset is used as the trigger source to capture body deformation in real time and adjust the actuator output. The cumulative usage time is used as a progressive constraint condition to gradually increase the correction load in stages. Based on the finite state machine, four working states are divided: standby, pre-tension, active correction, and maintenance. The state is automatically switched according to the real-time values ​​of the two parameters, which takes into account both instantaneous correction response and long-term progressive adaptation, and avoids instantaneous strong damage to the human body.

[0203] The total offset serves as the transition criterion for the four working states of the finite state machine (standby, pre-tightening, active correction, and maintenance). When the total offset is less than the preset qualified threshold, the system remains in standby or maintains the conformal working state. When the total offset exceeds the qualified threshold, the system triggers the pre-tightening step and then enters the active correction state.

[0204] In a preferred embodiment of the present invention, the central processing module includes:

[0205] The force line calculation engine is used to calculate the force line offset based on attitude data; based on a multi-sensor fusion algorithm (extended Kalman filter or complementary filter), it estimates the joint angles of the whole-body rigid chain model in real time and calculates the deviation between the actual force line and the ideal force line.

[0206] Adjust the control engine to generate control commands for the force line adjustment module; run the PID controller or model predictive controller to calculate the target displacement / force commands for each actuator based on the force line deviation, specifically:

[0207] Identify primary and compensatory offsets, and determine the primary offset source (such as anterior pelvic tilt) and secondary compensatory offset (such as lumbar hyperextension and knee hyperextension) of the current posture. For example, the value with the largest single force line offset is the primary offset source, and the others are secondary compensatory offsets.

[0208] Plan the shortest energy path from the current attitude to the ideal attitude, with the minimum total output force, and prioritize correcting the key parts that have the greatest impact on the overall force line, avoiding mutual antagonism between actuators. For example, select and adjust each actuator sequentially, first adjusting the dominant offset source, and then adjusting the secondary compensating offset.

[0209] When adjusting a specific part (such as traction on the inversion of the foot arch), the sensor detects the displacement of the key part in real time and acquires the changes in secondary compensatory offset, that is, the coupling offset it produces on the upstream force line (such as the knee and hip) (for example, inversion of the foot may cause internal rotation of the tibia). While issuing commands to the foot actuator, the controller sends feedforward compensation commands (usually a reverse preload of 5-10%) to the knee and hip actuators based on the predicted coupling amount, thereby maintaining the stability of the overall force line. The reverse preload can be determined as a compensation ratio of 5%-10% based on the total offset value. In a preferred embodiment, 10% reverse preload compensation is used when the total offset is greater than the upper threshold, and 5% reverse preload compensation is used when the total offset is less than the lower threshold. When the total offset is within the upper and lower limit range, the real-time compensation ratio is calculated by linear interpolation between 5% and 10%. Specifically, the upper threshold is higher than the qualified threshold, and the qualified threshold is higher than the lower threshold. Specific values ​​can be set, for example, the upper threshold is 10mm, the qualified threshold is 5mm, and the lower threshold is 1mm. The EEG decoding engine is used to decode and analyze EEG data, outputting a quantitative index of neural coding strength. Based on a deep learning framework (such as TensorFlow or PyTorch), it runs a pre-trained convolutional neural network or recurrent neural network model to perform feature extraction and classification / regression analysis on EEG signals.

[0210] The feedback mapping engine is used to map the quantification index into multimodal feedback parameters. It receives posture data from the force line sensing module and raw EEG data from the EEG acquisition module. After processing, it outputs control commands to the force line adjustment module, outputs feedback parameters to the feedback output module, and stores the entire process data in the data storage module.

[0211] In a preferred embodiment of the present invention, the body alignment correction and neural remodeling system further includes a feedback output module and a data storage and communication module. The feedback output module includes a visual feedback device, an auditory feedback device, and a somatosensory feedback module, and is electrically connected to the central processing module. Preferably, the feedback output module includes a head-mounted display or large-screen display (for visual feedback), bone conduction headphones or stereo speakers (for auditory feedback), and the alignment adjustment module itself (for proprioceptive feedback). The feedback output device is connected to the central processing module via an HDMI / DisplayPort interface, an audio interface, and the aforementioned actuator bus. The module's function is to convert EEG decoding results into sensory signals perceptible to the user, establishing a causal relationship between neural activity and external feedback.

[0212] Visual feedback can be provided using a 27-inch LCD display (Dell P2723D, 2560×1440 resolution) to present real-time visualization of the human force line; auditory feedback uses bone conduction headphones (AfterShokz Aeropex); proprioceptive feedback reuses the actuator of the force line adjustment module.

[0213] The data storage and communication module includes a local solid-state storage array and a wireless communication unit. The local solid-state storage array is electrically connected to the central processing module via the wireless communication unit. The data storage and communication module has a storage capacity of at least 512GB and is used to record raw EEG data, posture data, force line calculation results, decoding results, and training logs. The communication unit supports Wi-Fi 6 and 5G / 4G mobile networks and can encrypt and upload data to a cloud server for model iteration and optimization. This module connects to the central processing module via a SATA / PCIe interface. Local SSD storage can be used, and the module connects to the hospital's internal server via Wi-Fi 6.

[0214] The specific embodiments described herein are merely illustrative examples of the present invention. Those skilled in the art can make various modifications or additions to the described embodiments or use similar methods to substitute them, without departing from the technology of the present invention or exceeding the scope defined by the appended claims.

[0215] In the embodiments of this application, terms such as "fixed," "fixed connection," and "fixed connection" refer to common fixing methods in the prior art, such as welding, riveting, and screws. "Rotary connection" refers to common rotary connection methods in the prior art, such as hinges and bearing rotation. If electrical components are provided, the functions, control, and power supply methods of all electrical components are common technical means in the prior art. This application has not improved them and they are not within the protection scope of this application. Therefore, this application will not elaborate on them.

[0216] Furthermore, the selection of materials and strength limitations for all components in this application can be made and arranged by those skilled in the art based on the site environment and the requirements of relevant national or industry standards, and are not within the scope of protection of this application. Therefore, this application will not elaborate on these points.

Claims

1. A method for body alignment correction and neural remodeling based on closed-loop feedback of electroencephalogram (EEG) signals, characterized in that, Includes the following steps: By deploying a distributed attitude sensor array at key nodes of the human body, the three-dimensional spatial attitude data of the human body under static or dynamic conditions is collected in real time, and the force line offset between the current body force line and the ideal force line model is calculated. Based on the force line offset, a flexible actuator array arranged in key parts of the human body is driven to apply progressive mechanical guidance or constraint to the local area of ​​the human body, adjusting the current force line to the target force line state corresponding to the ideal force line model. During and after the force line adjustment, the user's cerebral cortex electrical activity signals are simultaneously collected by a multi-channel EEG acquisition device to obtain EEG data. Machine learning decoding algorithms are used to extract features from EEG data and obtain quantitative indicators that characterize the intensity of neural coding of the user’s current force line state. The quantitative indicators are mapped to multimodal feedback signals, which include visual feedback, auditory feedback and / or proprioceptive feedback generated by the flexible actuator array; Based on multimodal feedback signals, the mechanical guidance strategy of the flexible actuator array is adjusted until the user's brain establishes a stable neural representation of the target force line.

2. The method for body alignment correction and neural remodeling based on closed-loop feedback of electroencephalogram (EEG) signals according to claim 1, characterized in that, The ideal force line model is specifically as follows: The head and neck alignment, the midpoint of the external auditory meatus, the acromion, the greater trochanter of the femur, and the anterior edge of the lateral malleolus are all located on the same vertical line of the coronal plane; Shoulder girdle alignment: acromion, greater tuberosity of humerus, and radial head are vertically aligned in the coronal plane. The hip joint force line is formed by three points: the center of the hip joint (femoral head), the center of the knee joint (tibial intercondylar spine), and the center of the ankle joint (talus top). The ankle joint alignment, when standing in a neutral position, is such that the long axis of the tibia passes through the midpoint of the talus trochlea and falls between the weight-bearing area of ​​the heel and the first half of the metatarsal bones; The trunk midline force line and gravity line pass through the external auditory meatus, the posterior edge of the cervical vertebral body, the acromion, the anterior edge of the thoracic vertebral body, the posterior edge of the lumbar vertebral body, the posterior aspect of the hip joint, the anterior aspect of the knee joint, and the anterior aspect of the ankle joint in the sagittal plane; and through the root of the nose, the suprasternal notch, the umbilicus, the pubic symphysis, between the two knees, and between the two ankles in the coronal plane. The patellofemoral line is the angle formed between the line connecting the center of the patella to the tibial tuberosity and the line of tension of the quadriceps femoris muscle. The normal range for men is 10°-15° and for women is 12°-18°. The hip, knee, and ankle mechanical axis is arranged in a straight line with the femoral head center, knee joint center, and talus center. The mechanical axis passes through the knee joint center (0° position) to prevent knee valgus and varus. Foot alignment: When standing, the angle between the long axis of the tibia and the midline of the calcaneus is 0°-4°.

3. The method for body alignment correction and neural remodeling based on closed-loop feedback of electroencephalogram (EEG) signals according to claim 2, characterized in that, The method for calculating the force line offset between the current body force line and the ideal force line model is as follows: Obtain the coordinates of key points on the human body: Center point of external auditory meatus: , Anterior margin of cervical vertebral body: , acromion: , The junction of the thoracic and lumbar vertebrae: ， Highest point of the iliac crest: , greater trochanter of the femur: , Hip joint center: , Knee joint center: , Lateral ankle: , Constructing an ideal force line model: The ideal force line of the head and neck is a straight line in space formed by the center point of the external auditory meatus and the anterior edge of the cervical vertebral body: ， The ideal force line of the shoulder girdle is a straight line formed by the intersection of the acromion and the thoracolumbar region: ， Ideal force lines for the hip, knee, and ankle are constructed by segmenting two spatial straight lines: Hip to knee joint segment: ， From the knee to the lateral malleolus: , The ideal coronal plane is a standard vertical reference plane formed by the acromion, the highest point of the iliac crest, the greater trochanter of the femur, and the lateral malleolus. Let the coordinates of each key point in the real-time human posture be: The corresponding real-time body force line is: ; N sampling points are uniformly selected along a single real-time force line, and the spatial distance d from each sampling point to the corresponding ideal force line is calculated one by one. i The average value is taken as the offset of a single force line: The total overall posture offset is obtained by weighted summing of the force line offsets of the head, shoulder girdle, and lower limb segments: , in, , , The weighting coefficients satisfy the following constraints: + + =1.

4. The method for body alignment correction and neural remodeling based on closed-loop feedback of electroencephalogram (EEG) signals according to claim 1, characterized in that, Based on the force line offset, a flexible actuator array arranged at key parts of the human body is driven to apply progressive mechanical guidance or constraint to the local area of ​​the human body, adjusting the current force line to the target force line state. The specific steps are as follows: On the medial side of the arch, use a flat pneumatic braided array to pull the arch fascia upward and inward, guiding the calcaneus to 0-4° eversion, correcting foot eversion / medial longitudinal arch collapse, and restoring the calcaneus to neutrality; apply 1N, 3N, and 5N sequentially, with an increment of 0.5N each time, increasing every 5 minutes; In the direction of the oblique bundle of the quadriceps femoris muscle on the upper lateral side of the patella, an arc-shaped flexible SMA braided band is used to apply an oblique tightening force from the upper lateral side of the patella toward the medial epicondyle of the femur to reduce the Q angle beyond the preset range and prevent patellar lateral displacement. Apply 0N, 2N, and 4N sequentially to activate the gait cycle in stages, with activation occurring during the standing phase. The lumbar spiral, the contralateral gluteus maximus, quadratus lumborum, latissimus dorsi fascia bridge, and the use of a cross-shaped elastic-pneumatic hybrid array (oblique along the thoracolumbar fascia) to apply rotational shear force from the contralateral posterior superior iliac spine to the ipsilateral inferior angle of the scapula to reconstruct the spiral energy storage and transmission efficiency; low pretension (1N) is applied, which is increased to 6N only during trunk rotation / walking. Apply gentle pressure to the deep flexor muscles of the head and neck, specifically the projection area of ​​the longus colli / longus capitis muscles, using an ultra-thin dielectric elastomer patch in a posterior and inferior direction to restore the cervical spine to a neutral position; maintain a pressure of 0.8-1.5N, with only tactile feedback; reduce to 0.3N after more than 30 minutes.

5. The method for body alignment correction and neural remodeling based on closed-loop feedback of electroencephalogram (EEG) signals according to claim 3, characterized in that, Machine learning decoding algorithms are used to extract features from EEG data and obtain quantitative indicators representing the intensity of neural coding that characterizes the user's current force line state. The specific method is as follows: Collect the target force line, introduce known deviations, and mark the sensing time points; Configure the EEG channel; Offline preprocessing of EEG data; Three types of features were extracted from the EEG data, including ERN (Error-Related Negative Wave) amplitude, theta ERS (Event-Related Synchronization), and wPLI (Weighted Phase Lag Index) (C3-CPz). Training the ridge regression model; The posture sensor continuously collects the coordinates of key points on the human body, compares them with the ideal force line model, and calculates the force line offset. If the offset exceeds the allowable threshold, a start signal is sent to the integrated real-time system to correct the target parameters and formally trigger the entire workflow. Integrated Real-Time System: This includes configuring a real-time operating system foundation, EEG acquisition thread, feature extraction thread, model inference thread, actuator control thread, synchronization and latency measurement, integrity verification and debugging, specifically: Upon receiving the start command, the integrated real-time system immediately activates the EEG hardware acquisition channel to synchronously record the subject's real-time EEG signals, aligning them with the posture sensor timestamps throughout the process to ensure a one-to-one correspondence between "posture state" and "EEG signal". The raw EEG was filtered and preprocessed to remove artifacts, and neural features were extracted using a sliding window. The features are fed into the offline-trained regression model to calculate the subject's current neural feedback state. By combining the correction target given by the force line and the feedback value output by the regression model, actuator control commands are dynamically generated. An integrated real-time system drives the actuator to complete posture correction; the posture sensor continuously refreshes the force line offset, and the EEG device continuously samples. Repeat this process until the force line offset returns to the acceptable range, then stop data acquisition and correction.

6. The method for body alignment correction and neural remodeling based on closed-loop feedback of electroencephalogram (EEG) signals according to claim 1, characterized in that, Mapping quantitative indicators to multimodal feedback signals includes: Real-time neural bias coding strength: This refers to the internal confidence level of the brain continuously calculating the "deviation of the current posture from the ideal force line" through multi-region neural activity. Urgency of corrective needs: i.e., the need to correct biomechanical deviations; Mapping is performed through vision (AR glasses / ambient light), hearing (speakers / headphones), and proprioception (flexible actuator array); The warning phase uses visual and auditory senses, the corrective execution phase uses proprioception, hearing, and vision, and the successful reversion phase uses hearing and vision.

7. A body alignment correction and neural remodeling system based on the method of any one of claims 1-6, characterized in that, include: The force line sensing module consists of a distributed attitude sensor array arranged at key nodes of the human body, used to collect human body three-dimensional spatial attitude data in real time and calculate force line offset. The force line adjustment module consists of a distributed array of flexible actuators arranged in key parts of the human body. It is used to apply mechanical guidance or constraint to local parts of the human body according to control commands and adjust the body's force line. The EEG acquisition module consists of a multi-channel non-invasive EEG electrode array, used to synchronously acquire electrical activity signals from the user's cerebral cortex; The central processing module is connected to the force line sensing module, the force line adjustment module, and the EEG acquisition module. The central processing module receives EEG data, processes the data, and adjusts the mechanical guidance strategy of the flexible actuator array until the user's brain establishes a stable neural representation of the target force line.

8. The body alignment correction and neural remodeling system according to claim 7, characterized in that, The distributed attitude sensor array includes a three-axis accelerometer, a gyroscope, and a magnetometer; Key nodes in the human body include the head, shoulders, thoracic vertebrae, lumbar vertebrae, pelvis, knee joint, and ankle joint; The key parts of the human body include the shoulders, thorax, pelvic girdle, and the inner and outer sides of the knee joints; The distributed flexible actuator array includes adjustable straps driven by micro linear motors, pneumatic artificial muscles, and shape memory alloy wire actuators.

9. The body alignment correction and neural remodeling system according to claim 7, characterized in that, The central processing module includes: Force line calculation engine, used to calculate force line offset based on attitude data; Adjust the control engine to generate control commands for the force line adjustment module; The EEG decoding engine is used to decode and analyze EEG data and output a quantitative index of neural coding strength. A feedback mapping engine is used to map the quantification metrics to multimodal feedback parameters.

10. The body alignment correction and neural remodeling system according to claim 7, characterized in that, It also includes a feedback output module, as well as a data storage and communication module; The feedback output module includes a visual feedback device, an auditory feedback device, and a haptic feedback module, and the feedback output module is connected to the central processing module. The data storage and communication module includes a local solid-state storage array and a wireless communication unit, wherein the local solid-state storage array is connected to the central processing module through the wireless communication unit.

Images