Electroencephalogram guided adaptive adjustment method for hip-knee-ankle orthopedic rehabilitation exoskeleton

Abstract

The application discloses an electroencephalogram guided hip-knee-ankle orthopedic rehabilitation exoskeleton adaptive adjustment method, and relates to the technical field of intelligent rehabilitation engineering and biomedical engineering. On the basis of identifying the continuous electroencephalogram fluctuation characteristics in the intention transition period, the support response of the lower limb three joints is combined to construct a traceable transition identification zone. By extracting the driving dislocation point to set the joint response delay, the problem of early driving before the support is completed is avoided. The forward inclination trend is predicted in combination with the torso posture angle and the foot bottom pressure change, and a reverse support hierarchical matrix is dynamically arranged among the three joints to ensure the timing matching and compensation of the support forces among the joints when the abnormal driving trend occurs. Through the rhythmized inhibition and micro-assistance alternating mechanism, the joint output forms a flexible buffer gait, so that the gait stability, the neural participation degree and the safety of the patient in the rehabilitation training process are significantly improved.

Description

Technical Field

[0001] This invention relates to the fields of intelligent rehabilitation engineering and biomedical engineering technology, specifically to a brainwave signal-guided adaptive adjustment method for hip, knee and ankle orthopedic rehabilitation exoskeleton. Background Technology

[0002] EEG-guided adaptive adjustment of the hip, knee, and ankle orthopedic rehabilitation exoskeleton refers to an intelligent control method that, during lower limb rehabilitation, utilizes an intelligent sensing system and a brain-computer interface to collect the patient's motor intention signals in real time during EEG activity. These signals are decoded into identifiable control commands, which are then used to dynamically adjust the exoskeleton's assistance amplitude, gait rhythm, and joint compliance at the hip, knee, and ankle drive joints. Unlike traditional exoskeletons that rely on pressure, electromyography, or preset trajectory control, this method directly uses EEG signals as an active intention source, enabling the exoskeleton to execute movements under the guidance of the patient's thoughts, achieving a shift from passive following to active cooperation. Its core mechanism is as follows: when the EEG shows an intention to walk, flex or extend, or adjust balance, the system identifies this intention through the intelligent sensing system and brain-computer interface and maps it to the corresponding hip, knee, and ankle drive joints. The control algorithm automatically adjusts torque distribution and angle response, ensuring that the rhythm, support force, and gait morphology of rehabilitation training change in real time according to the patient's neural intentions. This approach not only enables individualized and compliant training during orthopedic rehabilitation but also promotes the reactivation of the brain-muscle neural pathway, enhancing neuroplasticity and rehabilitation outcomes.

[0003] The existing technology has the following shortcomings: In existing technologies, lower limb rehabilitation exoskeletons driven by EEG signals typically control the activation and deactivation of the hip and knee joints by recognizing the motor intention characteristics in the patient's EEG. However, during the transition from a standing intention to a walking intention, the EEG signals do not switch instantaneously but rather exhibit a brief, overlapping period. During this period, the EEG contains a mixture of static maintenance and dynamic initiation characteristics. When existing intention recognition algorithms fail to adequately distinguish the signal characteristics of this transitional phase, the system may misinterpret this mixed signal as a continuous motor intention. This causes the exoskeleton to prematurely release its support torque at the hip and knee joints, forcing the patient into a forward gait before establishing a stable center of gravity. This can lead to serious risks such as a forward shift of the body's center of gravity, delayed lower limb support, and forward leaning falls.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive adjustment method for hip, knee and ankle orthopedic rehabilitation exoskeleton guided by electroencephalogram (EEG) signals, so as to solve the problems in the background art mentioned above.

[0006] To achieve the above objectives, the present invention provides the following technical solution: An electroencephalogram (EEG)-guided adaptive adjustment method for hip, knee, and ankle orthopedic rehabilitation exoskeleton includes the following steps: S1 extracts continuous temporal fluctuation signals from the EEG signal during the transition from standing intention to stepping intention. Combined with the support response signals of the hip, knee and ankle joints, a transition recognition band unfolding along the time axis is constructed to mark unstable segments during gait initiation. S2, extract the temporal segments in the transition recognition zone where the EEG rhythm and the lower limb driving force response gradually decouple, splice the extracted segments into a continuous high-risk chain according to the order of appearance, and mark the driving misalignment points in the chain where the hip joint support is unstable but the knee joint driving has been released in advance; S3 sets response delay intervals for the hip, knee, and ankle joints based on the drive misalignment point, so that the joint drive release time lags behind the movement intention rise interval, so as to achieve support force priority over gait development and form adjustment buffer space. S4, within the response delay interval, uses a prediction mechanism driven by changes in trunk posture angle and plantar pressure to detect the initial tendency of the body to lean forward. It dynamically deploys a reverse support layered matrix between the hip, knee and ankle joints to counteract the premature release of the driving force output. S5 generates a continuous support rhythm on the time axis based on the reverse support layered matrix, consisting of alternating short-cycle inhibition segments and micro-amplitude assist segments. This allows the hip, knee, and ankle joints to dynamically adjust the assist output along the rhythmic path, forming a flexible buffer gait when the movement intention is not yet clear, and dispersing the risk of falls through multiple micro-balancing processes.

[0007] Optionally, step S1 includes: Continuous temporal fluctuation signals during the transition from standing intention to stepping intention are collected from electroencephalogram (EEG) signals to form data segments containing frequency mixing characteristics; By combining the support response data of the hip, knee and ankle joints, the micro-vibrations and asymmetrical force distribution in the loosening phase of the pre-stepping movement are extracted to form a temporal mixed data segment. The temporal mixed data segment is located on the time axis and extended forward and backward to construct a transition recognition zone that combines EEG and joint response; In the transition recognition zone, asynchronous segments where the EEG frequency has increased but the joint support has not decreased are marked, extracted as unstable segments, and recognition label blocks are established.

[0008] Optionally, step S2 includes: In the transition recognition zone, time-series segments where the frequency of EEG rhythms changes abruptly are selected, and asynchronous disconnection segments between the driving force responses of the hip, knee and ankle joints are extracted. The extracted multiple time-series disjoint segments are spliced ​​together in chronological order to construct a continuous high-risk signal chain; Mark drive misalignment points in the high-risk signal chain where the hip joint support is not yet stable but the knee or ankle joint has prematurely released the driving force; By establishing an index relationship between high-risk signal chains and driving misalignment sites, and archiving EEG fluctuation patterns and joint support curves, a micro-segment waveform archive for gait regulation is formed.

[0009] Optionally, the determination of the driving misalignment point is based on the fact that the hip joint support angle stability index has not dropped to the support end threshold, and that the knee or ankle joint has experienced a rapid decrease in output torque or a joint rotation trend.

[0010] Optionally, step S3 includes: Extract continuous signal segments of fixed duration before and after the driving misalignment point, and analyze the support force change curves and angle change curves of the hip joint, knee joint and ankle joint; Based on the time difference between the stable support angle range and the early release range of the driving force, response delay ranges are set for the hip, knee and ankle joints respectively; The response delay range is individually adjusted based on the gait characteristics of different individuals, and the delay value is fine-tuned in real time according to the changes in the intensity of EEG intention. Before the drive preparation action is about to be triggered, check whether the current intent timing is in a high-risk chain or drive misalignment segment, and activate the response delay control mechanism.

[0011] Optionally, the response delay interval is set based on the stability of the hip joint support angle and the starting moment of the decrease in driving force of the knee or ankle joint. It is calculated by the time difference between the two and the delay time percentage is dynamically adjusted according to the continuous upward trend of the EEG frequency.

[0012] Optionally, step S4 includes: Within the joint response delay interval, acquire changes in trunk posture angle and determine whether the forward tilt angle and angular acceleration reach the warning threshold. After identifying the initial signs of forward leaning, the plantar pressure distribution was collected and the pressure difference between the front and rear areas was analyzed to jointly determine the forward shift of the center of gravity. After meeting the conditions for predicting forward tilting, a reverse support layered matrix is ​​dynamically deployed between the hip, knee and ankle joints to construct an anti-countermeasure structure. After deployment, the three-joint support response is continuously fine-tuned based on the recovery of posture angle and the rebalancing of plantar pressure to form a dynamic closed-loop adjustment process.

[0013] Optionally, during the deployment of the reverse support layered matrix, the hip joint maintains angular stability and enhances the adduction direction resistance, the knee joint increases joint stiffness and enhances output torque, and the ankle joint adjusts the flexion and extension angles to share the load pressure generated by the forward shift of the center of gravity.

[0014] Optionally, step S5 includes: Based on the reverse support layered matrix, a rhythmic path consisting of alternating short-cycle suppression segments and micro-amplitude assist segments is generated on the time axis, and the angle change rate and output torque change rate of the three joints are set as the adjustment basis. After each rhythmic cycle ends, the ratio of the inhibition and assistance phases in the next cycle is adjusted based on the postural stability index and the plantar pressure balance. The output parameters of the hip, knee and ankle joints in each segment are adjusted in real time according to the rhythm path to achieve dynamic synergistic assistance adjustment; By comparing across cycles to determine the downward trend of risk, and by extending the proportion of the inhibition segment or increasing the output torque level of the hip joint when the center of gravity of the support changes abruptly, the support response can be strengthened.

[0015] Optionally, in the micro-assistance phase, the hip joint maintains an angle of stability and adjusts the output torque to maintain upper body support, the knee joint reduces its counter-stiffness and guides the thigh forward extension, and the ankle joint enhances its plantar flexion response to balance the distribution of ground reaction force. In the short-cycle inhibition phase, the three joints maintain a zero-angle difference locked state.

[0016] The beneficial effects of the technical solution provided by this invention include at least the following: This invention, based on the continuous EEG fluctuation characteristics during the intention transition period, combines the support response of the three lower limb joints to construct a traceable transition recognition zone. Furthermore, by extracting the drive misalignment points and setting joint response delays, it effectively avoids premature drive before support is completed. It predicts forward leaning trends by combining trunk posture angles and plantar pressure changes, and dynamically deploys a reverse support layered matrix between the three joints to ensure the temporal matching and compensation of support forces between joints when abnormal drive trends occur. Finally, through rhythmic inhibition and micro-assistance alternation mechanisms, it guides joint output to form a flexible, buffered gait. Even when the EEG motor intention is not fully clear, it can actively unload the risk of falls through multiple micro-balancing processes, thereby significantly improving gait stability, neural involvement, and patient safety during rehabilitation training. This scheme constructs a dynamic buffer interface between neural intention recognition and mechanical assistance control, providing an effective path for the safe adaptation of intelligent exoskeleton control in rehabilitation scenarios. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0018] Figure 1 This is a flowchart of the method for adaptive adjustment of the exoskeleton in hip, knee and ankle orthopedic rehabilitation guided by electroencephalogram (EEG) signals according to the present invention. Detailed Implementation

[0019] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0020] This invention provides, for example Figure 1 The electroencephalogram (EEG)-guided adaptive adjustment method for hip, knee, and ankle orthopedic rehabilitation exoskeletons, as shown, includes the following steps: S1 extracts continuous temporal fluctuation signals from the EEG signal during the transition from standing intention to stepping intention. Combined with the support response signals of the hip, knee and ankle joints, a transition recognition band unfolding along the time axis is constructed to mark unstable segments during gait initiation. By focusing on the temporal changes in EEG signals as the intention to stand gradually transitions to the intention to walk, a transition recognition band is constructed to identify and label unstable regions in the lower limb motor preparation state. This lays the signal analysis foundation for the prediction and intervention of subsequent high-risk movements. The specific steps are as follows:

[0021] A standard EEG acquisition device worn on the user's scalp was used to collect baseline EEG signals while the user was standing. The sampling frequency was set to 1000 times per second to ensure sufficient temporal resolution for subsequent temporal change analysis. During the acquisition process, the subject was asked to stand still for 15 seconds as a stable reference interval for the standing intention. Subsequently, the subject was instructed to transition from standing to a natural walking state, and the EEG signals during this intention transition were collected completely. The acquisition time lasted from 3 seconds before the intention was initiated to 2 seconds after the stepping action was completed, for a total of approximately 7 seconds of continuous EEG temporal data. This data includes frequency component changes during the transition from a static intention state to a dynamic intention state, especially concentrated in the mid-to-high frequency rhythms in the 8 to 30 Hz range. A frequency mixing interval was clearly observed before and after the formation of the motor intention, and this interval served as the basic data segment for subsequent construction of the transition recognition band.

[0022] While acquiring continuous EEG temporal data, multiple angle and torque sensors on the exoskeleton simultaneously collect support response data of the user's three major lower limb joints—hip, knee, and ankle—within this time interval. This includes joint rotation rate of change, ground reaction force response changes, and angular stability indices of the supporting foot relative to the ground. The data frequency is kept consistent with the EEG data to ensure accurate correspondence on the timeline. During the extraction process, special attention is paid to the slight tremors and asymmetrical force distribution generated by the joints during the pre-step fine-tuning phase. These phenomena typically occur within the time window when the movement intention is not yet fully clear but the movement has already begun to loosen. By combining the frequency mixing segments in the EEG signal, overlapping signal segments that simultaneously contain static maintenance intention and dynamic stepping intention within the intention transition interval can be accurately extracted, forming a temporal mixed data segment.

[0023] The temporal mixed data segments are located on the time axis and extended forward and backward in units of at least 500 milliseconds to form a joint transition recognition band covering EEG and three-joint support response signals. This recognition band is continuous and traceable, facilitating the observation of whether there is a synchronous shift between changes in EEG intention and joint response, paying particular attention to segments where joint support force has not yet decreased but EEG frequency characteristics show a tendency to initiate movement. For example, in an experiment with a subject, the EEG rhythm showed a high-frequency enhancement feature at 3.2 seconds, while the hip joint support angle remained stable until 3.8 seconds before changing. This asynchronous behavior of intention preceding support was marked as a high-risk precursor. The transition recognition band can be extracted by averaging multiple experiments to form a unified and comparable template for subsequent risk prediction modeling.

[0024] Based on the established transition recognition band, the stability characteristics of its time segments are further analyzed, and regions with obvious intentional support decoupling are selected as the basis for labeling unstable segments. The criteria for determining unstable segments include situations where the EEG frequency continuously increases above 200 milliseconds but no synchronous decrease occurs in any of the three joint support indicators, and segments where any of the three joints exhibits inertial micro-vibration signals before the support angle change begins. These unstable segments are labeled in the form of time indexes, forming recognition tag blocks with dynamic contradictions between EEG drive and support response. In subsequent training or use, the recognition band and unstable segments will serve as a reference for dynamic adjustment, establishing a risk warning window for upcoming joint drive processes and providing basic temporal anchors for subsequent movement rhythm buffering and support response delays.

[0025] S2, extract the temporal segments in the transition recognition zone where the EEG rhythm and the lower limb driving force response gradually decouple, splice the extracted segments into a continuous high-risk chain according to the order of appearance, and mark the driving misalignment points in the chain where the hip joint support is unstable but the knee joint driving has been released in advance; Based on the established transition recognition band, we further identify time segments where asynchronous changes occur between changes in EEG rhythm and the driving force responses of the three major lower limb joints. On this basis, we extract high-risk signal chains and precisely label driving misalignment points to improve the foresight and accuracy of gait safety control. The specific steps are as follows:

[0026] The study selected time segments within the transition recognition zone where EEG rhythms exhibited discontinuous acceleration trends, focusing particularly on periodic regions with frequency abrupt changes in the 8 Hz to 30 Hz frequency band. These segments were then subdivided into micro-segments with a time resolution of one millisecond. Further analysis was conducted on the relative response delays between these EEG rhythm abrupt changes and the support data of the three major joints, paying particular attention to time windows where high-frequency enhancement occurred first, but the hip joint driving force remained within the support stability range. For example, at 3.5 seconds recorded by the subject in the experiment, the EEG began to show a continuous rise in energy in the frequency band above 15 Hz, continuing until 4.1 seconds, but the hip joint angle stability index remained unchanged until 3.9 seconds. During this period, a disjoint state was formed where the EEG signal increased while the driving force response lagged. This disjoint state was extracted as a continuous temporal disjoint segment on the time axis as a preliminary basis for identifying high-risk potential segments.

[0027] Multiple extracted temporal disjoint segments are linearly spliced ​​together according to their natural occurrence order on the timeline to form a continuous and cumulative high-risk signal chain. This signal chain is not merely a simple combination of multiple time periods, but is connected based on the transitional relationship between each disjoint segment, ensuring the temporal logical consistency of signal fluctuations within the chain. For example, in the continuous data of a subject, the EEG changes from 2.8 seconds to 3.1 seconds are disjointed from the knee joint initiation response, followed by ankle joint driving force changes from 3.3 seconds to 3.7 seconds that precede the support posture stabilization signal, together forming a continuous high-risk chain covering 0.9 seconds. Each segment in this chain not only contains response deviations of a single joint, but also forms a structural risk accumulation trend of multi-joint response imbalance, providing a complete temporal early warning basis for subsequent interventions.

[0028] Further locating drive misalignment points within the constructed high-risk signal chain involves identifying critical time points in the three major joints where a drive response precedes the transfer of support action. Specifically, using the hip joint as the core support benchmark, the point in time where the knee or ankle joint shows a tendency to rotate or a rapid change in output torque, even before the support angle stability index has decreased to the end-of-support threshold, is identified as a drive misalignment point. For example, in one experiment, at 4.0 seconds, the hip joint support index remained above 90% stable, but the knee joint torque had rapidly decreased to 40% of its initial value, indicating that support had not been completed but drive action had been released, thus confirming this time point as a typical misalignment point. Multiple misalignment points can be marked in the same chain to observe the degree and frequency of relative coordination imbalance between joints.

[0029] A synchronous indexing relationship is established between high-risk signal chains and their internally marked misalignment points, forming a data benchmark structure for gait control response adjustment. The EEG fluctuation patterns and joint support curve changes within 500 milliseconds before and after each misalignment point are uniformly archived, forming a micro-segment waveform file for the misalignment point. During actual training or use, real-time comparison of EEG and joint data quickly locates whether an area has entered an identified high-risk chain region and determines whether it is approaching a known misalignment point threshold region, thereby triggering rhythm delay and assist buffering operations in the control system in advance. In this way, prediction and adjustment intervention can be completed at an early stage, before the EEG intention signal is fully clear but the driving rhythm has already shown an early trend, building a higher safety margin and controllability for the gait execution process.

[0030] S3 sets response delay intervals for the hip, knee, and ankle joints based on the drive misalignment point, so that the joint drive release time lags behind the movement intention rise interval, so as to achieve support force priority over gait development and form adjustment buffer space. Based on the identified actuation points, response delay intervals are set for the hip, knee, and ankle joints on the time axis. This ensures a stable support transition before joint actuation, avoiding safety hazards caused by premature release of support force during gait development. The specific steps are as follows:

[0031] Based on the identified drive misalignment points in the high-risk chain, continuous signal segments of 500 milliseconds in length were extracted before and after each misalignment point on the time axis. The support force and angle change curves of the hip, knee, and ankle joints within these segments were analyzed. Taking a subject as an example, a hip-knee misalignment occurred at 3.6 seconds. For the preceding 400 milliseconds, the hip support angle remained stable at approximately 45 degrees, while the knee joint output torque gradually decreased from 3.52 seconds, reaching 35% of its initial value by 3.6 seconds, showing a premature release trend. Within this segment, the stable support period of the hip joint was used as the support baseline, and the premature release point of the knee joint as the drive prolapse reference. The time difference between the two was calculated, yielding a response offset time of 80 milliseconds. Therefore, near this misalignment point, if the knee joint assist release is triggered immediately according to the upward trend of the EEG intention, there will be significant insufficient support. Therefore, around such misalignment points, the response delay time of the knee joint should be set no less than this offset value to ensure that the drive behavior does not precede the completion of support.

[0032] After extracting the initial response delay time, the response delay is further personalized based on the gait characteristics of different individuals to avoid sluggish or inconsistent joint responses due to a single threshold setting. In practice, comparative analysis of transition recognition bands and high-risk chain data from multiple subjects revealed that hip-knee dislocations occurred more frequently than knee-ankle dislocations, and the hip joint support time was typically longer than the knee joint actuation time by approximately 60 to 90 milliseconds. Therefore, for hip-knee combinations, a hip joint delay setting of zero and a knee joint delay setting of around 70 milliseconds are recommended; while for knee-ankle combinations, the ankle joint response advance is smaller, typically requiring only a buffer delay of 30 to 50 milliseconds. By differentiating the response time differences between different joints, the delay range setting ensures that it better matches the functional center of gravity of different joints at different stages, thereby improving the overall coordination accuracy of gait control.

[0033] After setting the response delay interval, a dynamic adaptive adjustment process is introduced to ensure its adaptability to dynamic changes in actual use. This process fine-tunes the delay value in real time based on changes in the intensity of the EEG intention. In practice, when a rapid rise in the EEG rhythm is detected in a specific segment—for example, when the subject's EEG frequency jumps from 13 Hz to 20 Hz between 3.5 and 3.8 seconds and lasts for more than 300 milliseconds—it can be inferred that the intention change is a clear gait initiation signal. In this case, the set delay time can be appropriately shortened by about 20% to ensure that the system does not exhibit excessive lag when the gait intention is clear. If the EEG frequency fluctuates only for a very short time and does not form a significant peak, the original delay setting remains unchanged. This approach ensures a balance between safety and sensitivity in the response delay, avoiding both premature release of support and excessive delay in action execution.

[0034] After setting and dynamically adjusting the joint response delay interval, and combining the misalignment index results, the system pre-checks whether the current intention timing falls within a marked high-risk chain or a segment near a misalignment before each drive preparation action is triggered. If the search results indicate that it is within a high-risk area, the response delay interval control mechanism is activated, delaying the drive release time by the set delay value until the support index is confirmed to enter a stable range. Taking a specific test as an example, at 4.2 seconds, the subject's EEG frequency continuously increased to above 20 Hz, but the hip joint support index had not yet fallen below 70%. After the system identified that this time point was at the second misalignment in the eighth high-risk chain, the knee joint drive time was delayed by 70 milliseconds, allowing the hip joint sufficient time to complete a stable transition and avoiding the risk of imbalance caused by premature release. This method ensures that in the early stages of gait initiation, the execution of each joint is always based on stable support, making exoskeleton-assisted control more reliable, smooth, and in line with human movement rhythms.

[0035] S4, within the response delay interval, uses a prediction mechanism driven by changes in trunk posture angle and plantar pressure to detect the initial tendency of the body to lean forward. It dynamically deploys a reverse support layered matrix between the hip, knee and ankle joints to counteract the premature release of the driving force output. To accurately identify the initial tendency of body forward leaning within the response delay interval, a coordinated prediction is made by combining changes in trunk posture angle and plantar pressure. A reverse support layered matrix is ​​dynamically deployed among the three joints of the lower limbs to counteract unstable torques and abrupt posture changes that may be caused by premature release of the actuator. The specific steps are as follows:

[0036] Within the joint response delay range, changes in the wearer's upper body posture angle were continuously acquired. Dual-set triaxial accelerometers and angular velocity meters, installed in the midline of the chest and above the lumbar spine, were used to monitor forward and backward tilt angles and angular acceleration information in real time. During the experiment, the basic standing posture was set as the zero reference point. If the forward tilt angle increased by more than three degrees from the zero reference point, and the angular acceleration reached more than 50 degrees squared per second, it was judged as an initial sign of forward tilting tendency. Simultaneously, to improve the timeliness of the prediction, the above conditions were required to be met within two hundred milliseconds consecutively before entering the initial warning state. Taking the subject as an example, in the stage where the intention to step was just formed but the support had not yet fully transferred, the forward tilt angle at the thoracic spine reached 3.7 degrees at 4.1 seconds and did not decrease within 0.25 seconds, which was recognized by the system as a clear initial signal of forward tilting tendency.

[0037] Based on the confirmed forward tilting trend, changes in plantar pressure are further superimposed to enhance the accuracy of the forward tilting trend assessment. Four pressure sensors are installed on each foot, located at the heel, below the metatarsals, in the center of the arch, and at the base of the toes, respectively. By continuously comparing the dynamic changes in pressure distribution on the left and right feet, it is determined whether there are signs of a forward shift in the center of gravity. Typical precursors to forward tilting are a rapid increase in pressure in the forefoot area and a rapid decrease in pressure in the hindfoot area. For example, in the 4.1 seconds after the forward tilting trend was detected by the trunk sensor, the pressure in the left forefoot simultaneously increased from 220 kPa to 270 kPa, while the pressure in the hindfoot decreased from 185 kPa to 130 kPa, a difference exceeding 35%. Combining the results from these two sensor sources, a joint predictive condition is established, and the risk intervention phase begins.

[0038] After confirming the forward leaning tendency twice, a layered counter-support matrix is ​​immediately deployed between the hip, knee, and ankle joints based on the pre-established counter-response mechanism. The deployment of this layered counter-support matrix uses the hip joint as the primary support center, the knee joint as the secondary response fulcrum, and the ankle joint as the end-effector stability controller. The deployment strategy is as follows: upon anticipating the initial signs of forward leaning, the hip joint first maintains its current angle stability, suspends all assist release actions, and moderately strengthens the adduction counterforce to stabilize the trunk; the knee joint gradually increases joint stiffness, enhancing the output torque by 10 to 15 N·m within 300 milliseconds to delay the knee folding tendency; the ankle joint, based on the current ground reaction force direction, actively adjusts the ankle flexion and extension angles, fine-tuning the foot support direction without disrupting the body's center of gravity, to share the load pressure brought by the forward shift of the center of gravity. This three-layered support adjustment strategy ensures that before the body is about to collapse due to forward leaning, a graded counter-support structure is collaboratively constructed by each joint, actively offsetting or buffering the premature release of driving forces and ensuring body stability.

[0039] To improve deployment efficiency and the naturalness of adjustment rhythm, after deploying the reverse support layered matrix, the support response of the three joints is continuously fine-tuned based on the recovery trend of trunk posture angle and the degree of rebalancing of plantar pressure, forming a dynamic closed-loop adjustment process. Specifically, if the trunk forward tilt angle has decreased from 3.7 degrees to below 2 degrees and the anterior-posterior plantar pressure difference has stabilized below 20 kPa one second after deployment, the counterforce output of the knee joint is gradually withdrawn, and the assist adjustment under normal response delay is restored; if the forward tilt persists, the reverse support strategy is maintained until the next balance is established. In a 30-walk test of a subject, the reverse support layered matrix was deployed 11 times, and in eight of these instances, the problem of forward center of gravity protrusion caused by premature release of the knee joint was effectively avoided. The intervention delay control of this mechanism is within 500 milliseconds, minimizing the impact of drastic changes in trunk angle on the wearer, and achieving stable suppression of slight forward tilting tendencies in multiple experiments.

[0040] S5, based on the reverse support layered matrix, generates a continuous support rhythm on the time axis consisting of alternating short-cycle inhibition segments and micro-amplitude assist segments, enabling the hip, knee and ankle joints to dynamically adjust the assist output along the rhythm path, forming a flexible buffer gait when the movement intention is not yet clear, and dispersing the risk of falling through multiple micro-balancing processes. Based on the deployed reverse support layered matrix, alternating short-cycle inhibition segments and micro-amplitude assist segments are generated on the time axis to construct a rhythmic support strategy. This allows the three-joint assist to form a flexible, cushioned gait when the movement intention is not yet clear, effectively dispersing the risk of falls. The specific implementation steps are as follows:

[0041] Within the first stable control cycle after the reverse support layered matrix is ​​triggered—that is, when the tendency of the trunk to lean forward is initially suppressed and the plantar pressure distribution becomes symmetrical again—the generation process of the rhythmic path is immediately initiated. This process is not driven by EEG motor intention signals, but rather by the joint stability in the support recovery state. In specific execution, the joint angle change rate and output torque change rate of the hip, knee, and ankle joints are set as the original variables for rhythm regulation. The initial cycle time is set to 800 milliseconds, and the rhythm is divided into two alternating phases: a short-cycle inhibition phase and a micro-amplitude assistance phase. In the short-cycle inhibition phase, the assistance of the three joints is maintained at the current level, the joint angle rate is limited to less than ten degrees per second, and the torque output is limited to within 30% of the maximum load-bearing value, forming a relatively static and stable support. In the micro-amplitude assistance phase, the three joints slightly adjust their output according to the current trunk posture and plantar pressure, with the assistance change amplitude not exceeding five Newton-meters, forming a gently guided support displacement. The existence of this alternating rhythm ensures that the support action is neither completely frozen nor blindly responded to before a clear movement initiation signal is obtained.

[0042] A continuity mechanism for the rhythmic support path is established and smoothly spliced ​​along the time axis to ensure a buffered transition between rhythmic segments and avoid abrupt changes. In practice, at the end of each cycle, the system reassesses the postural stability indicators and plantar pressure balance of the previous cycle, and determines the time ratio of the inhibition and assistance segments in the next cycle accordingly. For example, if the ankle joint angle change in the previous cycle is stable within one degree and the knee joint output torque fluctuates within 200 milliseconds without exceeding three Newton-meters, it indicates that the system is in a stable period. At this time, the proportion of the assistance segment in the next cycle can be increased to 60%, and the inhibition segment shortened to 40%; conversely, the current ratio is maintained or even the proportion of the inhibition segment is appropriately increased. In a continuous walking experiment with a subject, after identifying high-risk chains and deploying a reverse support layered matrix, the system continuously executed nine rhythmic cycles with an average cycle length of 750 milliseconds. In five of these cycles, the proportion of the assistance segment did not exceed 50%, demonstrating that the system can autonomously adjust its rhythmic response based on a stable state.

[0043] Based on the formation of the rhythmic path, dynamic coordinated adjustment is achieved by adjusting the output parameters of the three major joints in real time within this path. The hip joint, as the main support axis of the upper body, is responsible for the stable output of support force during the rhythmic process. Its output torque variation should be controlled within five Newton-meters, with the response primarily reflected in joint stiffness adjustment rather than angle variation. The knee joint, as the core of support transfer, appropriately reduces its counter-stiffness during the assist phase, allowing its output torque to slightly increase by three to five Newton-meters, guiding the thigh's forward extension tendency without forming a complete swing. The ankle joint, as the foot contact adjustment terminal, focuses on fine-tuning the support surface. During the assist phase, it appropriately enhances the plantar flexion response to help balance the distribution of ground reaction force, while maintaining a zero-angle-difference lock-in state during the inhibition phase. In one experiment, before the subject received a clear signal to initiate gait, the ankle joint's output torque fluctuated within two Newton-meters during the rhythm, demonstrating the system's high level of assist-softening capability.

[0044] The system continuously and dynamically executes a support adjustment process around this rhythmic path, forming a continuous buffering mechanism of multiple micro-balancing behaviors, ultimately achieving temporal dispersion and spatial unloading of fall risk. In each rhythmic cycle, the system records the redistribution of current support force and judges whether the risk continues to decrease through cross-cycle comparisons. If, over multiple cycles, the trunk posture angle is found to be stably maintained below three degrees of forward tilt, the anterior-posterior pressure difference of the foot is below 20 kPa, and there is no obvious tendency for misalignment between the three joints, the system continues to execute the rhythmic path. If a forward tilt tendency or a sudden change in the support center of gravity reappears in any cycle, the support strategy is immediately strengthened by extending the inhibition period or increasing the hip joint output torque level. This gait buffering mechanism, with temporal rhythm as its framework, coordinated assistance and micro-adjustment as its means, and the dispersion of fall risk as its goal, forms an important defense system for controlling exoskeleton behavior before the movement intention is clear. In a full-process intervention experiment, the subject successfully completed six consecutive micro-balancing processes before completing the intention recognition stage, avoiding postural imbalance caused by accidental triggering, demonstrating the high adaptability and stability of the rhythmic buffered gait in real rehabilitation scenarios.

[0045] This invention, based on the continuous EEG fluctuation characteristics during the intention transition period, combines the support response of the three lower limb joints to construct a traceable transition recognition zone. Furthermore, by extracting the drive misalignment points and setting joint response delays, it effectively avoids premature drive before support is completed. It predicts forward leaning trends by combining trunk posture angles and plantar pressure changes, and dynamically deploys a reverse support layered matrix between the three joints to ensure the temporal matching and compensation of support forces between joints when abnormal drive trends occur. Finally, through rhythmic inhibition and micro-assistance alternation mechanisms, it guides joint output to form a flexible, buffered gait. Even when the EEG motor intention is not fully clear, it can actively unload the risk of falls through multiple micro-balancing processes, thereby significantly improving gait stability, neural involvement, and patient safety during rehabilitation training. This scheme constructs a dynamic buffer interface between neural intention recognition and mechanical assistance control, providing an effective path for the safe adaptation of intelligent exoskeleton control in rehabilitation scenarios.

[0046] 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. A method for adaptive adjustment of an exoskeleton for hip, knee, and ankle orthopedic rehabilitation guided by electroencephalogram (EEG) signals, characterized in that, Includes the following steps: S1 extracts continuous temporal fluctuation signals from the EEG signal during the transition from standing intention to stepping intention. Combined with the support response signals of the hip, knee and ankle joints, a transition recognition band unfolding along the time axis is constructed to mark unstable segments during gait initiation. S2, extract the temporal segments in the transition recognition zone where the EEG rhythm and the lower limb driving force response gradually decouple, splice the extracted segments into a continuous high-risk chain according to the order of appearance, and mark the driving misalignment points in the chain where the hip joint support is unstable but the knee joint driving has been released in advance; S3 sets response delay intervals for the hip, knee, and ankle joints based on the drive misalignment point, so that the joint drive release time lags behind the movement intention rise interval, so as to achieve support force priority over gait development and form adjustment buffer space. S4, within the response delay interval, uses a prediction mechanism driven by changes in trunk posture angle and plantar pressure to detect the initial tendency of the body to lean forward. It dynamically deploys a reverse support layered matrix between the hip, knee and ankle joints to counteract the premature release of the driving force output. S5 generates a continuous support rhythm on the time axis based on the reverse support layered matrix, consisting of alternating short-cycle inhibition segments and micro-amplitude assist segments. This allows the hip, knee, and ankle joints to dynamically adjust the assist output along the rhythmic path, forming a flexible buffer gait when the movement intention is not yet clear, and dispersing the risk of falls through multiple micro-balancing processes.

2. The method for adaptive adjustment of hip, knee, and ankle orthopedic rehabilitation exoskeleton guided by electroencephalogram (EEG) signals according to claim 1, characterized in that, Step S1 includes: Continuous temporal fluctuation signals during the transition from the intention to stand to the intention to walk are collected from electroencephalogram (EEG) signals to form data segments; By combining the support response data of the hip, knee and ankle joints, the micro-vibrations and asymmetrical force distribution in the loosening phase of the pre-stepping movement are extracted to form a temporal mixed data segment. The temporal mixed data segment is located on the time axis and extended forward and backward to construct a transition recognition zone that combines EEG and joint response; In the transition recognition zone, asynchronous segments where the EEG frequency has increased but the joint support has not decreased are marked, extracted as unstable segments, and recognition label blocks are established.

3. The method for adaptive adjustment of hip, knee, and ankle orthopedic rehabilitation exoskeleton guided by electroencephalogram (EEG) signals according to claim 2, characterized in that, Step S2 includes: In the transition recognition zone, time-series segments where the frequency of EEG rhythms changes abruptly are selected, and asynchronous disconnection segments between the driving force responses of the hip, knee and ankle joints are extracted. The extracted multiple time-series disjoint segments are spliced ​​together in chronological order to construct a continuous high-risk signal chain; Mark drive misalignment points in the high-risk signal chain where the hip joint support is not yet stable but the knee or ankle joint has prematurely released the driving force; By establishing an index relationship between high-risk signal chains and driving misalignment sites, and archiving EEG fluctuation patterns and joint support curves, a micro-segment waveform archive for gait regulation is formed.

4. The method for adaptive adjustment of hip, knee, and ankle orthopedic rehabilitation exoskeleton guided by electroencephalogram (EEG) signals according to claim 3, characterized in that, The determination of the driving misalignment point is based on the fact that the stability index of the hip joint support angle has not dropped to the support end threshold, and the knee or ankle joint has experienced a rapid decrease in output torque or a joint rotation trend.

5. The method for adaptive adjustment of exoskeleton for hip, knee, and ankle orthopedic rehabilitation guided by electroencephalogram (EEG) signals according to claim 3, characterized in that, Step S3 includes: Extract continuous signal segments of fixed duration before and after the driving misalignment point, and analyze the support force change curves and angle change curves of the hip joint, knee joint and ankle joint; Based on the time difference between the stable support angle range and the early release range of the driving force, response delay ranges are set for the hip, knee and ankle joints respectively; The response delay range is individually adjusted based on the gait characteristics of different individuals, and the delay value is fine-tuned in real time according to the changes in the intensity of EEG intention. Before the drive preparation action is about to be triggered, check whether the current intent timing is in a high-risk chain or drive misalignment segment, and activate the response delay control mechanism.

6. The method for adaptive adjustment of exoskeleton for hip, knee, and ankle orthopedic rehabilitation guided by electroencephalogram (EEG) signals according to claim 5, characterized in that, The response delay interval is set based on the stability of the hip joint support angle and the starting moment of the decrease in driving force of the knee or ankle joint. It is calculated by the time difference between the two and dynamically adjusted according to the continuous upward trend of EEG frequency.

7. The method for adaptive adjustment of hip, knee, and ankle orthopedic rehabilitation exoskeleton guided by electroencephalogram (EEG) signals according to claim 5, characterized in that, Step S4 includes: Within the joint response delay interval, acquire changes in trunk posture angle and determine whether the forward tilt angle and angular acceleration reach the warning threshold. After identifying the initial signs of forward leaning, the plantar pressure distribution was collected and the pressure difference between the front and rear areas was analyzed to jointly determine the forward shift of the center of gravity. After meeting the conditions for predicting forward tilting, a reverse support layered matrix is ​​dynamically deployed between the hip, knee and ankle joints to construct an anti-countermeasure structure. After deployment, the three-joint support response is continuously fine-tuned based on the recovery of posture angle and the rebalancing of plantar pressure to form a dynamic closed-loop adjustment process.

8. The method for adaptive adjustment of exoskeleton for hip, knee, and ankle orthopedic rehabilitation guided by electroencephalogram (EEG) signals according to claim 7, characterized in that, During the deployment of the reverse support layered matrix, the hip joint maintains angular stability and enhances the adduction direction resistance, the knee joint increases joint stiffness and enhances output torque, and the ankle joint adjusts the flexion and extension angles to share the load pressure generated by the forward shift of the center of gravity.

9. The method for adaptive adjustment of hip, knee, and ankle orthopedic rehabilitation exoskeleton guided by electroencephalogram (EEG) signals according to claim 7, characterized in that, Step S5 includes: Based on the reverse support layered matrix, a rhythmic path consisting of alternating short-cycle suppression segments and micro-amplitude assist segments is generated on the time axis, and the angle change rate and output torque change rate of the three joints are set as the adjustment basis. After each rhythmic cycle ends, the ratio of the inhibition and assistance phases in the next cycle is adjusted based on the postural stability index and the plantar pressure balance. The output parameters of the hip, knee and ankle joints in each segment are adjusted in real time according to the rhythm path to achieve dynamic synergistic assistance adjustment; By comparing across cycles to determine the downward trend of risk, and by extending the proportion of the inhibition segment or increasing the output torque level of the hip joint when the center of gravity of the support changes abruptly, the support response can be strengthened.

10. The method for adaptive adjustment of hip, knee, and ankle orthopedic rehabilitation exoskeleton guided by electroencephalogram (EEG) signals according to claim 9, characterized in that, In the micro-assist phase, the hip joint maintains a stable angle and adjusts the output torque to maintain upper body support, the knee joint reduces its counter-stiffness and guides the thigh forward extension, and the ankle joint enhances its plantar flexion response to balance the distribution of ground reaction force. In the short-cycle inhibition phase, the three joints maintain a zero-angle difference locked state.

Images