Upper limb rehabilitation training system closed loop design method and system
The upper limb rehabilitation training system, which uses multimodal information acquisition and closed-loop control, can assess the patient's condition in real time and dynamically adjust the training strategy. This solves the problem of insufficient personalization in existing equipment, improves the safety and efficacy of rehabilitation training, and provides personalized and safe rehabilitation solutions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- MAIZU INTELLIGENT TECH (SHANGHAI) CO LTD
- Filing Date
- 2026-02-14
- Publication Date
- 2026-06-05
AI Technical Summary
Existing automated rehabilitation training equipment lacks accurate assessment and dynamic adjustment of patients' real-time physiological status and training effects, resulting in insufficient personalization of training programs and an inability to intelligently intervene based on changes in patients' immediate abilities, thus affecting the safety and efficacy of rehabilitation training.
A closed-loop control method involving multimodal information acquisition, real-time assessment, and decision-making is adopted. By collecting physiological electrical signals, kinematic information, dynamic information, and physiological state information of patients during the training process in real time, training strategy adjustment instructions are dynamically generated to form a closed-loop control process, thereby realizing personalized rehabilitation training.
It enables dynamic adjustment of training programs based on patients' real-time performance, identification and timely intervention of abnormal movement patterns, improved safety and efficacy of rehabilitation training, enhanced patients' immersion and motivation to participate, and provided objective quantitative indicators of rehabilitation effects.
Smart Images
Figure CN122157955A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of medical rehabilitation equipment technology, specifically, it relates to a closed-loop design method and system for an upper limb rehabilitation training system. Background Technology
[0002] Currently, rehabilitation training is a key means of restoring function for upper limb motor dysfunction caused by stroke, spinal cord injury, musculoskeletal diseases, etc. Existing automated rehabilitation training equipment, such as exoskeleton robots and rehabilitation robotic arms, can provide repetitive training, but most of them adopt preset and fixed training modes. They lack accurate assessment and dynamic adjustment of the patient's real-time physiological state and training effect. Their open-loop control mode results in insufficient personalization of training programs and cannot make intelligent interventions based on the patient's immediate changes in ability, such as fatigue, attention, muscle strength growth, and abnormal situations during training, such as compensation and spasticity. This affects the safety and final efficacy of rehabilitation training. Therefore, we propose a closed-loop design method for upper limb rehabilitation training systems. Summary of the Invention
[0003] To solve the above-mentioned technical problems, the basic concept of the technical solution adopted by the present invention is as follows: A closed-loop design method for an upper limb rehabilitation training system includes the following steps: S1: Real-time acquisition of multimodal information during patient training; S2: Based on multimodal information, calculate the evaluation indicators of patient status and training effect in real time; S3: Based on evaluation metrics, dynamically generate training strategy adjustment instructions through preset decision rules; S4: Execute the adjustment instruction, change the training parameters and provide feedback to the patient, then return to step S1 to form a closed-loop control process.
[0004] In a preferred embodiment of the present invention, the multimodal information in S1 includes physiological electrical signals, kinematic information, dynamic information, and physiological state information reflecting fatigue and concentration.
[0005] In a preferred embodiment of the present invention, the evaluation index in S2 includes at least two of the following: motor function index, muscle activation index, abnormal movement pattern recognition index, and comprehensive fatigue index.
[0006] In a preferred embodiment of the present invention, the training strategy adjustment instructions in S3 include adjusting the assistive force parameters provided by the rehabilitation robot, adjusting the task difficulty of the virtual training environment, adjusting the feedback mode, and triggering safety intervention.
[0007] This invention also provides a closed-loop design system for an upper limb rehabilitation training system, comprising: The multimodal perception module is used to collect multi-dimensional data from patients during training. The signal processing and feature extraction module is used to process the multi-dimensional data and extract features. The intelligent evaluation and decision-making module is used to calculate evaluation indicators and generate adjustment instructions in real time based on the characteristics. The training execution and interaction module is used to execute the adjustment instructions and provide the patient with a physical training and human-computer interaction environment; The closed-loop control and data management module is used to coordinate the work of each module, manage data flow, and store training data.
[0008] In a preferred embodiment of the present invention, the multimodal sensing module includes an electromyography signal acquisition unit, a kinematics acquisition unit, a dynamics acquisition unit, a physiological state acquisition unit, and a synchronization and data aggregation unit; The electromyography (EMG) signal acquisition unit consists of a multi-channel surface EMG sensor array and an analog-to-digital converter; the kinematics acquisition unit consists of an infrared camera from an optical motion capture system and a high-precision joint encoder from the rehabilitation robot body; the dynamics acquisition unit consists of a flexible pressure sensor installed on the robot's end effector and used to measure grip force; the physiological state acquisition unit consists of an optical heart rate and blood oxygen sensor and a skin conductance sensor; and the synchronization and data aggregation unit is a master microcontroller responsible for providing a unified clock signal to all sensors and collecting the digital signals from each unit in real time, packaging them into data frames with timestamps.
[0009] In a preferred embodiment of the present invention, the signal processing and feature extraction module includes a preprocessing unit, a feature calculation unit, and a buffer unit; The preprocessing unit is a digital filter bank; the feature calculation unit is a high-performance embedded processor that runs feature extraction algorithms; and the caching unit is used to temporarily store recent historical data for analysis that requires a time window.
[0010] In a preferred embodiment of the present invention, the intelligent assessment and decision-making module includes a status assessment unit, a decision-making unit, and a safety monitoring and intervention unit; The status assessment unit loads a pre-trained lightweight model to classify muscle states in real time, such as normal, fatigued, or spastic. The decision-making unit includes a strategy rule base and an adaptive algorithm. The strategy rule base contains adjustment strategy mapping tables under different rehabilitation goals, while the adaptive algorithm can implement simple PID control. The safety monitoring and intervention unit is an independent high-priority process that continuously monitors safety-related parameters. Once a hard safety threshold is exceeded, it immediately bypasses the decision-making process and directly sends an emergency stop or relaxation command to the execution module.
[0011] In a preferred embodiment of the present invention, the training execution and interaction module includes a physical training execution unit and a virtual human-computer interaction unit; The physical training execution unit consists of the rehabilitation robot body, servo motors and their drivers, and a safety clutch; the virtual human-computer interaction unit consists of a VR head-mounted display, stereo headphones, and a haptic feedback handle.
[0012] In a preferred embodiment of the present invention, the closed-loop control and data management module includes a real-time control scheduling unit, a data bus and communication interface unit, a database and storage unit, and a human-machine configuration and management interface unit. The real-time control and scheduling unit is an industrial control computer running a real-time operating system, responsible for precisely managing the timing of the entire closed loop; the data bus and communication interface unit includes EtherCAT and CAN buses for high-speed hard real-time control of the robot, and USB for medium-speed data transmission; the database and storage unit is a local solid-state drive or a cloud database for structured storage of all raw data, feature data, evaluation results, decision logs, and patient information; the human-machine configuration and management interface unit is a graphical user interface for therapists, used to set training prescriptions, view real-time data streams, replay historical training videos, and analyze long-term trend reports.
[0013] Compared with the prior art, the present invention has the following advantages: This invention, through a closed-loop design, enables the system to dynamically adjust the training program based on the patient's real-time performance, achieving personalized rehabilitation for different individuals, and can identify abnormal movement patterns in real time and intervene immediately to prevent secondary injuries.
[0014] This invention allows for dynamic difficulty adjustment based on assessment results and rich multimodal feedback, ensuring that training remains at an appropriate level of challenge, enhancing patient immersion and motivation. Furthermore, the full-process data recording provides objective and continuous quantitative indicators for rehabilitation effects, facilitating therapists to remotely monitor training progress and adjust macro-rehabilitation plans.
[0015] The specific embodiments of the present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description
[0016] In the attached diagram: Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system block diagram of the present invention.
[0017] In the diagram: 1. Multimodal perception module; 2. Signal processing and feature extraction module; 3. Intelligent evaluation and decision-making module; 4. Training execution and interaction module; 5. Closed-loop control and data management module. Detailed Implementation
[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments will be clearly and completely described below with reference to the accompanying drawings. The following embodiments are used to illustrate the present invention.
[0019] like Figures 1 to 2 As shown, the present invention provides a technical solution, a closed-loop design method for an upper limb rehabilitation training system, comprising the following steps: S1: Real-time acquisition of multimodal information during patient training; S2: Based on multimodal information, calculate the evaluation indicators of patient status and training effect in real time; S3: Based on evaluation metrics, dynamically generate training strategy adjustment instructions through preset decision rules; S4: Execute the adjustment instruction, change the training parameters and provide feedback to the patient, then return to step S1 to form a closed-loop control process.
[0020] Furthermore, the multimodal information in S1 includes physiological electrical signals, kinematic information, dynamic information, and physiological state information reflecting fatigue and concentration. Specifically, physiological electrical signals: electromyographic signals of the target muscle group are collected through surface electromyography (sEMG) sensors; kinematic information: the angle, angular velocity, motion trajectory, and end-effector position of the patient's upper limb joints are collected through inertial measurement units (IMUs), optical motion capture systems, or robotic joint encoders; dynamic information: the interaction force and joint torque between the patient and the training equipment are collected through force / torque sensors; physiological state information: the patient's heart rate variability, skin conductance, and other signals reflecting fatigue and concentration are collected through heart rate sensors, skin conductance sensors, etc.
[0021] Furthermore, the evaluation indicators in S2 include motor function indicators and abnormal movement patterns; Specifically, motor function assessment indicators include: joint range of motion, movement smoothness (Jerk value), trajectory error, and task completion time; abnormal state identification includes: identifying muscle spasms or compensatory movement patterns based on electromyographic and force signals.
[0022] Furthermore, the training strategy adjustment instructions in S3 include adjusting the assistive force parameters provided by the rehabilitation robot, adjusting the task difficulty of the virtual training environment, adjusting the feedback mode, and triggering safety interventions. Specifically, the adjustment of assistive force parameters includes: adjusting the magnitude and direction control parameters of the assistive force provided by the rehabilitation robot; adjusting the task difficulty, including the task objectives, path complexity, and interference items in the virtual reality training environment; adjusting the feedback mode, including the content, intensity, and form of visual, auditory, or tactile feedback; and safety intervention instructions, which trigger a safety pause or reduce the load when spasticity or high-risk compensation is detected.
[0023] This invention also provides a closed-loop design system for an upper limb rehabilitation training system, comprising: The multimodal perception module is used to collect multi-dimensional data from patients during training. The signal processing and feature extraction module is used to process multi-dimensional data and extract features; The intelligent assessment and decision-making module is used to calculate assessment indicators and generate adjustment instructions in real time based on characteristics; The training execution and interaction module is used to execute adjustment instructions and provide patients with a physical training and human-computer interaction environment; The closed-loop control and data management module is used to coordinate the work of each module, manage data flow, and store training data.
[0024] Furthermore, the multimodal sensing module includes an electromyography signal acquisition unit, a kinematics acquisition unit, a dynamics acquisition unit, a physiological state acquisition unit, and a synchronization and data aggregation unit. The electromyography (EMG) signal acquisition unit consists of a multi-channel surface EMG sensor array and an analog-to-digital converter; the kinematics acquisition unit consists of an infrared camera from an optical motion capture system and a high-precision joint encoder from the rehabilitation robot body; the dynamics acquisition unit consists of a flexible pressure sensor installed on the robot's end effector and used to measure grip force; the physiological state acquisition unit consists of an optical heart rate and blood oxygen sensor and a skin conductance sensor; and the synchronization and data aggregation unit is a main control microcontroller responsible for providing a unified clock signal to all sensors and collecting the digital signals from each unit in real time, packaging them into data frames with timestamps. Specifically, a multi-channel surface electromyography (EMG) sensor array is attached to the skin surface of the target muscle group (such as the deltoid, biceps, and triceps); the IMU is strapped to the upper arm, forearm, and back of the hand; the force sensor is integrated into the robot handle; and the physiological sensor is worn on the wrist or finger. After the system is powered on, the synchronization unit issues a sampling command, and each sensor synchronously collects the raw signal at a preset frequency. After preliminary conditioning and ADC conversion, the data packet with timestamp is sent to the signal processing and feature extraction module in real time via USB protocol.
[0025] Furthermore, the signal processing and feature extraction module includes a preprocessing unit, a feature calculation unit, and a buffer unit. The preprocessing unit is a digital filter bank; the feature calculation unit is a high-performance embedded processor that runs feature extraction algorithms; and the caching unit is used to temporarily store recent historical data for analysis that requires a time window. Specifically, the electromyography (sEMG) signal processing involves rectifying and filtering the raw sEMG signal, then calculating time-domain, frequency-domain, and time-frequency-domain features within a sliding time window; motion and dynamics processing involves performing posture calculations on IMU data to obtain joint angles, directly reading encoder and force sensor data, and calculating velocity, acceleration, and interaction force / torque; physiological signal processing involves calculating time-domain / frequency-domain indices of heart rate variability (HRV), average level of skin conductance signals, or event-related fluctuations; and finally, the output is an integrated "feature vector" every 100-300ms, containing all calculated evaluation indicators, which is then sent to the intelligent evaluation and decision-making module.
[0026] Furthermore, the intelligent assessment and decision-making module includes a status assessment unit, a decision-making unit, and a safety monitoring and intervention unit. The status assessment unit loads a pre-trained lightweight model to classify muscle states in real time, such as normal, fatigued, or spastic. The decision-making unit includes a strategy rule base and an adaptive algorithm. The strategy rule base contains adjustment strategy mapping tables under different rehabilitation goals, while the adaptive algorithm can implement simple PID control. The safety monitoring and intervention unit is an independent high-priority process that continuously monitors safety-related parameters. Once a hard safety threshold is exceeded, it immediately bypasses the decision-making process and directly sends an emergency stop or relaxation command to the execution module. Specifically, in the evaluation phase: after receiving the feature vectors, they are simultaneously fed into both the rule model and the machine learning model. The rule model quickly provides an interpretable preliminary judgment, while the machine learning model provides a comprehensive state probability. The results of the two are fused to form a final comprehensive evaluation report. The decision engine then queries the policy rule base based on the evaluation report and the objectives of the current training phase.
[0027] Furthermore, the training execution and interaction module includes a physical training execution unit and a virtual human-computer interaction unit; The physical training execution unit consists of the rehabilitation robot body, servo motors and their drivers, and a safety clutch; the virtual human-computer interaction unit consists of a VR head-mounted display, stereo headphones, and a haptic feedback handle. Specifically, in physical execution: upon receiving instructions from the decision module to "adjust the auxiliary force to how many Newtons" or "switch to impedance control mode and set the stiffness to YY N / m", the robot driver adjusts the motor current in real time to change the robot's mechanical behavior; in virtual interaction: upon receiving instructions to "adjust the task difficulty".
[0028] Furthermore, the closed-loop control and data management module includes a real-time control scheduling unit, a data bus and communication interface unit, a database and storage unit, and a human-machine configuration and management interface unit; The real-time control and scheduling unit is an industrial control computer running a real-time operating system, responsible for precisely managing the timing of the entire closed loop; the data bus and communication interface unit includes EtherCAT and CAN buses for high-speed hard real-time control of the robot, and USB for medium-speed data transmission; the database and storage unit is a local solid-state drive or a cloud-connected database for structured storage of all raw data, feature data, evaluation results, decision logs, and patient information; the human-machine configuration and management interface unit is a graphical user interface for therapists, used to set training prescriptions, view real-time data streams, replay historical training videos, and analyze long-term trend reports. Specifically, the system operates on a fixed cycle. At the start of each cycle, the scheduling unit triggers the perception module to collect data, ensuring processing and evaluation are completed within 5ms, decision-making within 8ms, and new instructions are sent to the execution module before the 10ms cycle ends. Communication between all modules is conducted through a central data bus, with data packets bearing strict timestamps and priority tags, and security-related data receiving the highest priority. After each training session, the system automatically generates a report containing key indicator curves and a summary. All data is stored encrypted, and therapists can use the management interface to longitudinally compare patient data from multiple training sessions, assess rehabilitation progress, and manually update the macro-level training goals for the next stage.
[0029] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A closed-loop design method for an upper limb rehabilitation training system, characterized in that, Includes the following steps: S1: Real-time acquisition of multimodal information during patient training; S2: Based on multimodal information, calculate the evaluation indicators of patient status and training effect in real time; S3: Based on evaluation metrics, dynamically generate training strategy adjustment instructions through preset decision rules; S4: Execute the adjustment instruction, change the training parameters and provide feedback to the patient, then return to step S1 to form a closed-loop control process.
2. The closed-loop design method for an upper limb rehabilitation training system according to claim 1, characterized in that, The multimodal information in S1 includes physiological electrical signals, kinematic information, dynamic information, and physiological state information reflecting fatigue and concentration.
3. The closed-loop design method for an upper limb rehabilitation training system according to claim 1, characterized in that, The evaluation indicators in S2 include at least two of the following: motor function indicators, muscle activation indicators, abnormal movement pattern recognition indicators, and comprehensive fatigue indicators.
4. The closed-loop design method for an upper limb rehabilitation training system according to claim 1, characterized in that, The training strategy adjustment instructions in S3 include adjusting the assistive force parameters provided by the rehabilitation robot, adjusting the task difficulty of the virtual training environment, adjusting the feedback mode, and triggering safety intervention.
5. A closed-loop design system for an upper limb rehabilitation training system, characterized in that, The method for closed-loop design of an upper limb rehabilitation training system according to any one of claims 1-4, wherein the closed-loop design system for the upper limb rehabilitation training system comprises: The multimodal perception module (1) is used to collect multidimensional data from patients during training. The signal processing and feature extraction module (2) is used to process the multi-dimensional data and extract features; The intelligent evaluation and decision-making module (3) is used to calculate evaluation indicators and generate adjustment instructions in real time based on the features. The training execution and interaction module (4) is used to execute the adjustment instructions and provide the patient with a physical training and human-computer interaction environment; The closed-loop control and data management module (5) is used to coordinate the work of each module, manage data flow and store training data.
6. The closed-loop design system for an upper limb rehabilitation training system according to claim 5, characterized in that, The multimodal sensing module (1) includes an electromyography signal acquisition unit, a kinematics acquisition unit, a dynamics acquisition unit, a physiological state acquisition unit, and a synchronization and data collection unit; The electromyography (EMG) signal acquisition unit consists of a multi-channel surface EMG sensor array and an analog-to-digital converter; the kinematics acquisition unit consists of an infrared camera from an optical motion capture system and a high-precision joint encoder from the rehabilitation robot body; the dynamics acquisition unit consists of a flexible pressure sensor installed on the robot's end effector and used to measure grip force; the physiological state acquisition unit consists of an optical heart rate and blood oxygen sensor and a skin conductance sensor; and the synchronization and data aggregation unit is a master microcontroller responsible for providing a unified clock signal to all sensors and collecting the digital signals from each unit in real time, packaging them into data frames with timestamps.
7. The closed-loop design system for an upper limb rehabilitation training system according to claim 5, characterized in that, The signal processing and feature extraction module (2) includes a preprocessing unit, a feature calculation unit, and a caching unit; The preprocessing unit is a digital filter bank; the feature calculation unit is a high-performance embedded processor that runs feature extraction algorithms; and the caching unit is used to temporarily store recent historical data for analysis that requires a time window.
8. The closed-loop design system for an upper limb rehabilitation training system according to claim 5, characterized in that, The intelligent assessment and decision-making module (3) includes a status assessment unit, a decision-making unit, and a safety monitoring and intervention unit; The status assessment unit loads a pre-trained lightweight model to classify muscle states in real time, such as normal, fatigued, or spastic. The decision-making unit includes a strategy rule base and an adaptive algorithm. The strategy rule base contains adjustment strategy mapping tables under different rehabilitation goals, while the adaptive algorithm can implement simple PID control. The safety monitoring and intervention unit is an independent high-priority process that continuously monitors safety-related parameters. Once a hard safety threshold is exceeded, it immediately bypasses the decision-making process and directly sends an emergency stop or relaxation command to the execution module.
9. The closed-loop design system for an upper limb rehabilitation training system according to claim 5, characterized in that, The training execution and interaction module (4) includes a physical training execution unit and a virtual human-computer interaction unit; The physical training execution unit consists of the rehabilitation robot body, servo motors and their drivers, and a safety clutch; the virtual human-computer interaction unit consists of a VR head-mounted display, stereo headphones, and a haptic feedback handle.
10. A closed-loop design system for an upper limb rehabilitation training system according to claim 5, characterized in that, The closed-loop control and data management module (5) includes a real-time control scheduling unit, a data bus and communication interface unit, a database and storage unit, and a human-machine configuration and management interface unit; The real-time control and scheduling unit is an industrial control computer running a real-time operating system, responsible for precisely managing the timing of the entire closed loop; the data bus and communication interface unit includes EtherCAT and CAN buses for high-speed hard real-time control of the robot, and USB for medium-speed data transmission; the database and storage unit is a local solid-state drive or a cloud database for structured storage of all raw data, feature data, evaluation results, decision logs, and patient information; the human-machine configuration and management interface unit is a graphical user interface for therapists, used to set training prescriptions, view real-time data streams, replay historical training videos, and analyze long-term trend reports.