Lower extremity intelligent feedback training system and method

Abstract

This invention discloses a lower limb intelligent feedback training system and method, belonging to the field of rehabilitation training technology. The system includes a multi-source information acquisition module, a real-time intention and state recognition module, an adaptive decision-making and control module, a multimodal feedback execution module, and a training data management and individual model update module. The multi-source information acquisition module collects force, motion, and physiological information during lower limb training; the real-time intention and state recognition module fuses and analyzes the multi-source information to identify the user's motion intention and assess their motion state; the adaptive decision-making and control module generates training parameter adjustment instructions based on the motion intention, motion state, training goals, and individual ability parameters; the multimodal feedback execution module provides feedback to the user based on the training parameters; and the training data management and individual model update module updates the individual ability parameters. Through this structure, adaptive adjustment and closed-loop feedback in the lower limb training process are achieved.

Description

Technical Field

[0001] This invention relates to the field of rehabilitation training technology, and more specifically, to a lower limb intelligent feedback training system and method. Background Technology

[0002] Lower limb motor function is fundamental for standing, walking, and daily activities. Patients often experience varying degrees of lower limb motor dysfunction after stroke, spinal cord injury, orthopedic surgery, or neurodegenerative diseases. For these individuals, clinical rehabilitation training typically promotes neurological function reconstruction and muscle strength recovery through repetitive lower limb exercises. However, traditional rehabilitation methods rely heavily on therapist guidance or experience-based judgment, making quantitative assessment difficult. Furthermore, training intensity and rhythm are significantly influenced by human factors, making it challenging to maintain long-term stability and consistency.

[0003] With the development of sensor and computing technologies, lower limb training systems based on wearable devices, rehabilitation robots, or intelligent training devices have emerged. These systems can typically collect joint angles, gait parameters, or simple biomechanical information, and control the training process through preset programs, reducing reliance on manual intervention and improving training repeatability to some extent. However, most existing systems focus on collecting single or limited types of motion information, lacking comprehensive utilization of multi-dimensional information such as plantar force, joint force, and muscle electrical activity, resulting in an incomplete portrayal of the user's true movement state and abilities.

[0004] Furthermore, existing lower limb training systems often employ fixed parameters or simple threshold-based adjustment strategies in their training control and feedback methods. This makes it difficult to accurately distinguish between a user's active movement intentions and passive following state during continuous training, and also hinders the timely reflection of changes in user fatigue levels and movement stability. Without effective individual ability modeling and dynamic update mechanisms, training parameters often fail to adaptively adjust to changes in user capabilities, easily leading to problems such as excessively high training intensity affecting safety, or insufficient training stimulation affecting rehabilitation outcomes. Therefore, how to achieve multi-source information fusion, real-time recognition of movement intentions and states, and adaptive closed-loop adjustment of training parameters during lower limb training remains a pressing technical challenge that needs to be addressed in current technologies.

[0005] Therefore, there is an urgent need for a lower limb intelligent feedback training system and method to solve these problems. Summary of the Invention

[0006] The purpose of this invention is to solve the technical problems mentioned in the background section and to provide a lower limb intelligent feedback training system and method.

[0007] The objective of this invention is achieved through the following technical solution: One aspect of the present invention provides a lower limb intelligent feedback training system, comprising: The multi-source information acquisition module is used to collect the user's force information, motion information and physiological information in real time during lower limb training, and to perform time synchronization processing on the collected multi-source information; The real-time intent and state recognition module is communicatively connected to the multi-source information acquisition module. It is used to perform feature extraction and multimodal fusion analysis on the synchronized multi-source information, and to recognize the user's motion intent and evaluate the user's motion state in real time during continuous training. The adaptive decision and control module is communicatively connected to the real-time intention and state recognition module. It is used to take the recognized movement intention and movement state as input, and combine them with preset training objectives and individual ability parameters to dynamically generate parameter adjustment instructions for lower limb training. The multimodal feedback execution module is communicatively connected to the adaptive decision and control module and is used to output mechanical feedback, visual feedback and / or auditory feedback to the user according to the parameter adjustment instructions, so as to guide or correct the user's lower limb training movements. The training data management and individual model update module is used to record training process data and update the individual ability parameters based on the training results, thereby forming a closed-loop lower limb intelligent feedback training system.

[0008] Furthermore, the force information includes at least plantar pressure distribution information and joint force or joint torque information; the motion information includes at least lower limb joint angle, angular velocity and angular acceleration information; and the physiological information includes at least muscle electrical activity information related to lower limb movement.

[0009] Furthermore, the multi-source information acquisition module collects various types of information at a uniform sampling frequency and aligns data from different sources using timestamps to construct multi-source state data reflecting the lower limb movement state at the same moment.

[0010] Furthermore, the real-time intent and state recognition module extracts features from multi-source state data. The extracted features include at least lower limb force distribution features, joint motion coordination features, motion smoothness features, and muscle activation intensity features.

[0011] Furthermore, the real-time intent and state recognition module classifies and identifies the user's movement intent within the current training cycle based on multimodal fusion features, and simultaneously outputs a movement state assessment result reflecting the user's current movement stability, coordination, and fatigue level.

[0012] Furthermore, the adaptive decision-making and control module constructs a comprehensive control state based on the motion intention, motion state, preset training target, and individual ability parameters, which serves as the decision-making basis for dynamically adjusting training parameters.

[0013] Furthermore, the adaptive decision-making and control module dynamically adjusts lower limb training parameters, including at least training intensity, assistance level, damping, or rhythm control, based on the comprehensive control state to adapt to changes in the user's abilities.

[0014] Furthermore, the multimodal feedback execution module includes one or more of a force feedback unit, a visual guidance unit, and an auditory cue unit, used to provide the user with real-time feedback information corresponding to the training parameters.

[0015] Furthermore, the training data management and individual model update module adaptively updates the individual ability parameters based on the user's changes in motion state and training completion status during the training process.

[0016] Furthermore, the updated individual ability parameters are used to adjust the parameters of the adaptive decision and control module during subsequent training to achieve long-term personalized optimization of the lower limb training process.

[0017] The present invention also provides a lower limb intelligent feedback training method, comprising the following steps: S1. During lower limb training, the user's force information, motion information and physiological information are collected in real time, and the collected multi-source information is processed in time synchronization to form multi-source state data reflecting the lower limb motion state at the same time. S2. Perform feature extraction and multimodal fusion analysis on the multi-source state data to identify the user's motion intention and evaluate the user's motion state during continuous training. S3. The identified movement intentions and movement states are comprehensively analyzed with the preset training goals and individual ability parameters to generate training parameter adjustment instructions for regulating the lower limb training process. S4. Based on the training parameter adjustment instructions, output mechanical feedback, visual feedback and / or auditory feedback to the user to guide or correct the user's lower limb training movements; S5. Record the multi-source information generated during training, the results of motion intention recognition, the adjustment of training parameters and feedback results, and update the individual ability parameters according to the training completion status. S6. Apply the updated individual ability parameters to subsequent training processes to achieve adaptive adjustment and closed-loop feedback in the lower limb training process.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention achieves intelligent management of the entire lower limb training process by constructing a closed-loop training structure that includes multi-source information acquisition, real-time intent and state recognition, adaptive decision-making and control, and multimodal feedback execution. During training, the system simultaneously collects multi-dimensional data such as plantar pressure, joint motion information, and muscle electrical activity. Through time synchronization and fusion analysis, it forms a unified description of the lower limb movement state, thus avoiding the incomplete information problem caused by relying on a single kinematic parameter. This improves the accuracy of perceiving the user's actual movement ability and real-time state, providing a reliable data foundation for subsequent training adjustments.

[0019] 2. This invention introduces a real-time motion intention recognition and motion state assessment mechanism based on multi-source perception, enabling the system to distinguish between the user's active motion behavior and passive following behavior during continuous training, and dynamically assess motion stability, coordination, and fatigue trends. Based on this, the adaptive decision-making and control module can combine preset training goals and individual ability parameters to adjust training intensity, assistance level, and damping or rhythm control in real time, thereby effectively avoiding problems of excessive training intensity or insufficient stimulation, and improving the safety and relevance of the training process.

[0020] 3. This invention, through training data management and individual model update mechanisms, continuously feeds back the motor performance during training to individual ability parameters, enabling the system to gradually adapt to changes in user ability as the number of training sessions increases, achieving long-term personalized optimization. By providing users with real-time feedback in multiple forms, including mechanical, visual, and auditory feedback through a multimodal feedback execution module, it not only enhances the interactivity and comprehensibility of the training process but also improves user participation and training compliance, thereby contributing to the continuous improvement of lower limb function and the stable enhancement of rehabilitation effects. Attached Figure Description

[0021] Figure 1 This is a system block diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the following description is provided in conjunction with embodiments and appendices. Figures 1-2 The present invention will be further described in detail below. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0023] In this embodiment, the lower limb intelligent feedback training system is applied to rehabilitation training scenarios to provide active-assisted training for users with lower limb motor dysfunction. The system as a whole includes a multi-source information acquisition module, a real-time intention and state recognition module, an adaptive decision-making and control module, a multimodal feedback execution module, and a training data management and individual model update module. The modules are connected via wired or wireless means to form a complete closed-loop training structure.

[0024] Before training begins, the user wears or is equipped with a lower limb training device that is compatible with the system. The multi-source information acquisition module includes various sensors for synchronously acquiring force, motion, and physiological information during lower limb training. Among these, plantar pressure distribution information is acquired through a pressure sensor array arranged in the training shoe or footpad. This pressure sensor array can employ thin-film pressure sensors or resistive pressure sensors to obtain the force changes in different areas of the foot during standing, stepping, or weight-bearing.

[0025] Joint force or torque information is collected by force sensors or torque sensors installed at the joints of the lower limb training device. These sensors are used to reflect the mechanical action on the joints during active or passive movement by the user.

[0026] The lower limb joint angle, angular velocity, and angular acceleration information are collected by an inertial measurement unit. The inertial measurement unit is fixedly installed in the thigh, calf, or near the joint. The inertial measurement unit includes at least an accelerometer and a gyroscope to acquire the motion posture changes of each joint of the lower limb in real time.

[0027] Muscle electrical activity information related to lower limb movement is collected by surface electromyography (EMG) sensors, which are attached to the surface of key muscles such as the quadriceps, hamstrings, or calf muscles, to collect information on muscle activation during training.

[0028] The multi-source information acquisition module uniformly samples the data from the various sensors mentioned above and synchronizes the data from different sources using timestamps to form multi-source state data reflecting the lower limb movement state at the same moment. This data is then sent to the real-time intent and state recognition module.

[0029] The real-time intent and state recognition module is deployed in the control terminal of the training device or an external computing device to process the received multi-source state data. This module first extracts features from plantar pressure distribution, joint motion information, and electromyographic signals. The extracted features include the trend of plantar force distribution changes, joint motion coordination, movement smoothness, and muscle activation strength. Subsequently, it fuses and analyzes features from different sources to identify the user's movement intent during continuous training, such as determining whether the user is currently in an active exertion phase, a passive following phase, or a support and swing phase, and simultaneously assesses the user's movement stability, coordination, and fatigue trend.

[0030] The adaptive decision-making and control module receives motion intention and motion state information output by the real-time intention and state recognition module, and makes a comprehensive judgment based on pre-set training target parameters and stored individual ability parameters. The training target parameters may include the type of training movement, target range of motion, training rhythm, or training intensity level; the individual ability parameters describe the user's current lower limb motor ability level. Based on the above information, the adaptive decision-making and control module dynamically generates training parameter adjustment instructions to adjust training intensity, level of assistance, and damping or rhythm control, thereby matching the training process to the user's real-time ability.

[0031] The multimodal feedback execution module provides feedback to the user based on the training parameter adjustment instructions output by the adaptive decision and control module. The multimodal feedback execution module includes one or more of a force feedback unit, a visual guidance unit, and an auditory cueing unit. Specifically, the force feedback unit can apply auxiliary or damping forces to the lower limb training device via a motor, pneumatic, or hydraulic actuator to guide the user in completing the target movement; the visual guidance unit can present training movement prompts, movement trajectories, or completion information to the user via a display screen or projection; and the auditory cueing unit can provide rhythmic prompts or movement correction prompts to the user via voice or audible alerts.

[0032] The training data management and individual model update module records and stores multi-source sensor data, motion intent recognition results, training parameter adjustments, and feedback execution results generated during training. At the end of each training cycle, this module updates the individual ability parameters based on changes in the user's motion state and training completion status, and applies the updated individual ability parameters to subsequent training processes. This allows the system to gradually adapt to changes in user capabilities, achieving long-term personalized optimization.

[0033] Through the above implementation methods, the lower limb intelligent feedback training system can perceive the user's lower limb movement status in real time during the training process, dynamically understand the user's movement intention, and achieve safe and effective lower limb intelligent training through adaptive decision-making and multimodal feedback.

[0034] Example 2: In this embodiment, the lower limb intelligent feedback training system is used to perform real-time perception analysis, decision-making, and feedback control of the user's lower limb movements during rehabilitation or functional training, thereby achieving intelligent and individualized adjustment of the training process. The system as a whole includes a multi-source information acquisition module, a real-time intention and state recognition module, an adaptive decision-making and control module, a multimodal feedback execution module, and a training data management and individual model update module. These modules form a closed-loop working structure through data communication.

[0035] During training, the multi-source information acquisition module uses a preset sampling frequency. Real-time data collection of the user's lower limb movements, with a sampling frequency of... This indicates the number of times lower limb motion information is sampled per unit time. The collected information includes force information, motion information, and physiological information. Force information includes at least plantar pressure distribution information and joint force or joint torque information; motion information includes at least lower limb joint angles, joint angular velocities, and joint angular acceleration information; and physiological information includes at least muscle electrical activity information related to lower limb motion.

[0036] At any sampling time The multi-source information acquisition module performs time synchronization processing on various types of information and constructs a unified original state vector: ; in, Indicates the first Each sampling time This indicates the time at which each pressure sampling point on the sole of the foot is located. The pressure distribution vector; Indicates the time of each joint in the lower limb The joint torque vector; This represents the joint angle vectors of each joint in the lower limbs; Represents the joint angular velocity vector; Represents the joint angular acceleration vector; This represents the signal vector of muscle electrical activity related to the muscle groups of the lower limbs.

[0037] The aforementioned original state vector serves as the input to the real-time intent and state recognition module. This module first performs feature extraction processing on the original state vector using feature extraction operators. The state vector is mapped to the feature vector: ; in, For at any time The extracted multi-source feature vector includes at least the features of plantar force distribution variation, joint motion coordination, motion smoothness, and muscle activation intensity.

[0038] Subsequently, the real-time intent and state recognition module performs multimodal fusion processing on features from different sources, dividing the feature vectors into... The class feature vectors are weighted and fused according to preset weights to obtain the fused feature vector: ; in, Indicates the first Class feature vectors, This represents the corresponding feature weight coefficient. This represents the number of feature categories, and each weight coefficient satisfies the weighted normalization condition to ensure the stability of the fused features.

[0039] After obtaining the fused feature vector, the real-time intent and state recognition module identifies the user's motion intent within the current training period based on the fused feature vector. The motion intent is represented as follows: ; in, Indicates at time The identified movement intention category is used to characterize whether the user is currently in the active exertion, passive following, support, or swinging phase. This represents a motion intent mapping model.

[0040] Meanwhile, the real-time intent and state recognition module calculates motion state parameters based on the same fused feature vector. ; in, This indicates the user's current motion status assessment result, which is used to comprehensively reflect motion stability, movement coordination, and fatigue level; This represents a motion state assessment model.

[0041] Identified motion intentions and motion state This information serves as input to the adaptive decision-making and control module. The adaptive decision-making and control module then compares this information with the preset training objective. and individual ability parameters Combine them to construct a comprehensive control state vector: ; in, This represents the preset training target parameters, used to limit the type of training task, target amplitude, or rhythm requirements; Indicates at time Individual ability parameters are used to describe the user's current lower limb motor ability level.

[0042] Based on the integrated control state vector, the adaptive decision and control module dynamically generates training parameter vectors: ; in, This represents the set of training parameters within the current control period, including at least training intensity parameters, assistance level parameters, and damping or rhythmic control parameters. This represents the adaptive decision function.

[0043] The training parameter vector is further used as input to the multimodal feedback execution module, which generates corresponding feedback signals based on the training parameters. ; in, Indicates at time The set of output feedback signals includes at least one or more of the following: mechanical feedback signals, visual guidance signals, and auditory cue signals; This represents the feedback mapping function. The feedback signal guides or corrects the user's lower limb movements, thereby influencing the movement state in the next control cycle.

[0044] The training data management and individual model update module records the raw data, recognition results, training parameters, and feedback results throughout the entire training process, and updates the individual ability parameters based on the motion state assessment results. The update rules are as follows: ; in, This indicates the updated individual ability parameters. The ability update learning rate controls the magnitude of individual ability parameter updates. The updated individual ability parameters continue to serve as input to the adaptive decision-making and control module in subsequent training cycles, enabling long-term personalized adaptive optimization of training parameters and thus constructing a complete lower limb intelligent feedback training system.

[0045] Example 3: This example discloses a lower limb intelligent feedback training method. The method is based on the aforementioned lower limb intelligent feedback training system and is used to guide users to complete a lower limb training process with real-time feedback and adaptive adjustment capabilities.

[0046] At the start of training, the system enters an initialization state, where training goals and initial individual ability parameters are preset by the rehabilitation personnel or the system. Subsequently, during lower limb training, the system uses a multi-source information acquisition module to perceive the user's lower limb movements in real time. Specifically, plantar pressure is collected through a pressure sensor array installed in the training shoe or footpad; lower limb joint motion information is collected through inertial measurement units fixed to the thigh and calf; and muscle electrical activity information related to lower limb movement is collected through surface electromyography (EMG) sensors attached to the target muscle groups. The system uniformly samples data from different sensors and performs time synchronization processing using timestamps, thereby obtaining multi-source state data reflecting the lower limb movement state at the same moment.

[0047] After acquiring multi-source state data, the system performs feature processing and fusion analysis on the collected data. The system extracts features reflecting force distribution and center of gravity shift from plantar pressure data, features reflecting joint range of motion and coordination from joint motion data, and features reflecting muscle activation levels from electromyography (EMG) signals. Based on these fused features, the system identifies the user's movement intentions during continuous training, such as determining whether the user is currently in an active exertion, passive following, or support and swing phase. Simultaneously, it comprehensively evaluates the user's movement stability, coordination, and fatigue trends, generating corresponding movement state assessment results.

[0048] Subsequently, the system comprehensively analyzes the identified movement intentions and states with preset training objectives and current individual ability parameters. Based on the analysis results, the system dynamically generates training parameter adjustment instructions to regulate the training process, controlling training intensity, assistance level, and rhythm or damping-related parameters to match the training load with the user's current ability level.

[0049] After generating training parameter adjustment instructions, the system outputs corresponding feedback information to the user based on these instructions. Specifically, when the system determines that the user needs assistance, it applies auxiliary force to the user through the motor or actuator in the lower limb training device; when it detects movement deviations or rhythm abnormalities, it provides visual guidance information to the user through the display terminal or reminds the user to adjust the movement rhythm through voice prompts. Through the synergistic effect of multiple feedback methods, the system guides the user to complete lower limb movements that meet the training objectives.

[0050] During training, the system continuously records multi-source sensor data, motion intention recognition results, training parameter adjustments, and feedback execution status. At the end of each training cycle, the system updates individual ability parameters based on changes in the user's movement state and training completion, storing the updated parameters for future adjustments during training. By repeatedly performing this training process, the system gradually adapts to changes in user ability with increasing training sessions, achieving adaptive adjustment and closed-loop feedback in lower limb training.

[0051] The method described in this embodiment enables continuous, stable, and personalized training control of the lower limb intelligent feedback training process while ensuring safety. It is applicable to various application scenarios such as hospital rehabilitation training, community rehabilitation, and home rehabilitation.

[0052] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, but the present invention is not limited to these embodiments. Equivalent modifications made by those skilled in the art without departing from the principles of the present invention should fall within the protection scope of the present invention.

[0053] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A lower limb intelligent feedback training system, characterized in that, include: The multi-source information acquisition module is used to collect the user's force information, motion information and physiological information in real time during lower limb training, and to perform time synchronization processing on the collected multi-source information; The real-time intent and state recognition module is communicatively connected to the multi-source information acquisition module. It is used to perform feature extraction and multimodal fusion analysis on the synchronized multi-source information, and to recognize the user's motion intent and evaluate the user's motion state in real time during continuous training. The adaptive decision and control module is communicatively connected to the real-time intention and state recognition module. It is used to take the recognized movement intention and movement state as input, and combine them with preset training objectives and individual ability parameters to dynamically generate parameter adjustment instructions for lower limb training. The multimodal feedback execution module is communicatively connected to the adaptive decision and control module and is used to output mechanical feedback, visual feedback and / or auditory feedback to the user according to the parameter adjustment instructions, so as to guide or correct the user's lower limb training movements. The training data management and individual model update module is used to record training process data and update the individual ability parameters based on the training results, thereby forming a closed-loop lower limb intelligent feedback training system.

2. The lower limb intelligent feedback training system according to claim 1, characterized in that, The force information includes at least plantar pressure distribution information and joint force or joint torque information; the motion information includes at least lower limb joint angle, angular velocity and angular acceleration information; and the physiological information includes at least muscle electrical activity information related to lower limb movement.

3. The lower limb intelligent feedback training system according to claim 1, characterized in that, The multi-source information acquisition module collects various types of information at a uniform sampling frequency and aligns data from different sources using timestamps to construct multi-source state data reflecting the lower limb movement state at the same moment.

4. The lower limb intelligent feedback training system according to claim 1, characterized in that, The real-time intent and state recognition module extracts features from multi-source state data. The extracted features include at least lower limb force distribution features, joint motion coordination features, motion smoothness features, and muscle activation intensity features.

5. The lower limb intelligent feedback training system according to claim 4, characterized in that, The real-time intent and state recognition module classifies and identifies the user's movement intent within the current training cycle based on multimodal fusion features, and simultaneously outputs movement state assessment results reflecting the user's current movement stability, coordination, and fatigue level.

6. The lower limb intelligent feedback training system according to claim 1, characterized in that, The adaptive decision-making and control module constructs a comprehensive control state based on the motion intention, motion state, preset training goal, and individual ability parameters, which serves as the decision-making basis for dynamically adjusting training parameters.

7. The lower limb intelligent feedback training system according to claim 6, characterized in that, The adaptive decision and control module dynamically adjusts lower limb training parameters, including at least training intensity, level of assistance, damping, or rhythm control, based on the comprehensive control state to adapt to changes in the user's abilities.

8. The lower limb intelligent feedback training system according to claim 1, characterized in that, The multimodal feedback execution module includes one or more of a force feedback unit, a visual guidance unit, and an auditory cue unit, and is used to provide the user with real-time feedback information corresponding to the training parameters.

9. The lower limb intelligent feedback training system according to claim 1, characterized in that, The training data management and individual model update module adaptively updates the individual ability parameters based on the user's changes in movement state and training completion status during the training process. The updated individual ability parameters are used for parameter adjustment in the adaptive decision and control module during subsequent training processes to achieve long-term personalized optimization of the lower limb training process.

10. A lower limb intelligent feedback training method, characterized in that, Includes the following steps: S1. During lower limb training, the user's force information, motion information and physiological information are collected in real time, and the collected multi-source information is processed in time synchronization to form multi-source state data reflecting the lower limb motion state at the same time. S2. Perform feature extraction and multimodal fusion analysis on the multi-source state data to identify the user's motion intention and evaluate the user's motion state during continuous training. S3. The identified movement intentions and movement states are comprehensively analyzed with the preset training goals and individual ability parameters to generate training parameter adjustment instructions for regulating the lower limb training process. S4. Based on the training parameter adjustment instructions, output mechanical feedback, visual feedback and / or auditory feedback to the user to guide or correct the user's lower limb training movements; S5. Record the multi-source information generated during training, the results of motion intention recognition, the adjustment of training parameters and feedback results, and update the individual ability parameters according to the training completion status. S6. Apply the updated individual ability parameters to subsequent training processes to achieve adaptive adjustment and closed-loop feedback in the lower limb training process.

Images