An intelligent orthopedic rehabilitation training monitoring system based on multi-modal sensing

Abstract

The present application relates to the technical field of monitoring alarm, in particular to a kind of intelligent orthopedics rehabilitation training monitoring system based on multi-modal sensing, system includes: basic data synchronous acquisition module, dynamic stiffness differential analysis module, defensive shrinkage discrimination module, load boundary dynamic reconstruction module.In the present application, by synchronously collecting joint angle and torque data to construct time series, the instantaneous dynamic stiffness change trend is analyzed in combination with torque and angle difference value, the Savitzky-Golay filter is used to capture the stiffness second derivative mutation point by second-order differential processing of dynamic stiffness, identify the critical moment when the patient joint appears defensive shrinkage pain, dynamically reconstruct the maximum activity angle upper limit based on critical angle and safety buffer difference, control command is directly fed back to the driving end to execute reverse brake, realize the leap from single position monitoring to multidimensional mechanical state perception, ensure that activity boundary is automatically adjusted according to real-time physiological feedback of patient, prevent secondary injury caused by forcibly breaking through pain threshold.

Description

Technical Field

[0001] This invention relates to the field of monitoring and alarm technology, and in particular to an intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing. Background Technology

[0002] The field of monitoring and alarm technology involves the use of physical sensing components to collect specific state parameters of a target object in real time and to drive an actuator to issue an alarm signal when the parameters exceed a preset physical range. Specifically, a traditional intelligent orthopedic rehabilitation training monitoring system refers to a physical system mainly composed of inertial measurement sensors, flexible straps, data transmission cables, and a host computer display terminal. The inertial measurement sensors are mechanically fixed to the patient's joints via flexible straps. The internal accelerometer and gyroscope components collect physical signals of limb movement. These signals are input to the receiving port of the host computer display terminal via data transmission cables. The host computer display terminal then drives a connected buzzer hardware to issue a continuous audio alarm based on the received signal values.

[0003] Existing technologies mainly rely on inertial sensors combined with flexible straps for mechanical fixation monitoring. However, the flexible straps are prone to loosening or displacement during repeated limb movements, causing the collected physical signals to deviate from the actual joint movement state. Furthermore, judging abnormalities solely based on preset physical ranges is too mechanical and cannot detect physiological resistance responses triggered by muscle spasms or pain during rehabilitation training. Relying solely on single-dimensional positional signals cannot accurately identify sudden changes in the mechanical properties of the joint. Consequently, the system cannot respond in time when the patient has already generated defensive contractions before exceeding the preset range, making it difficult to ensure the safety of training for damaged joints under dynamic changes. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose an intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing includes:

[0006] The basic data synchronous acquisition module acquires joint angle values ​​and joint torque values, pairs the joint angle values ​​and joint torque values ​​to construct a joint angle-torque time series, and transmits the joint angle-torque time series to the dynamic stiffness differential analysis module.

[0007] The dynamic stiffness differential analysis module calculates the torque difference and angle difference based on the joint angle-torque time series, divides the torque difference by the angle difference to obtain the instantaneous dynamic stiffness value, calls the Savitzky-Gore filter to perform second-order differentiation on the instantaneous dynamic stiffness value to obtain the second-order derivative value of stiffness and the joint angle value at the current moment, and transmits the second-order derivative value of stiffness and the joint angle value at the current moment to the defensive contraction discrimination module;

[0008] The defensive contraction discrimination module obtains the stiffness mutation safety threshold. When the value of the second derivative of stiffness is greater than the stiffness mutation safety threshold, the current joint angle value is extracted as the pain critical angle value, and the pain critical angle value is transmitted to the load boundary dynamic reconstruction module.

[0009] The load boundary dynamic reconstruction module calculates the difference between the pain threshold angle value and the safety buffer angle value to obtain the maximum activity angle limit, writes the maximum activity angle limit into the servo controller, and sends a reverse braking command to the motor driver.

[0010] As a further aspect of the present invention, the joint angle-torque time series includes the joint angle value and the joint torque value at the same sampling time. The second derivative value of stiffness specifically refers to the second derivative calculation result characterizing the degree of acceleration of joint stiffness change. The pain critical angle value is specifically the joint angle value at the current time when stiffness mutation characteristics are identified. The upper limit of the maximum range of motion specifically refers to the safety boundary value used to update the stroke parameters of the servo controller. The reverse braking command includes a motor smooth deceleration signal and a position locking signal.

[0011] As a further aspect of the present invention, the specific function of the basic data synchronization acquisition module is as follows:

[0012] The timestamp alignment submodule acquires the original encoder signal and the original torque sensor signal, extracts the timestamp information of the original encoder signal and the original torque sensor signal, performs linear interpolation alignment processing, removes invalid data frames with time deviations exceeding the preset synchronization window, and generates synchronization joint data frames.

[0013] The sequence splicing construction submodule extracts the effective angle components and effective torque components from the synchronized joint data frames, combines the effective angle components and effective torque components into a multi-dimensional vector format according to the time order, and continuously stores the multi-dimensional vector format to establish a joint angle-torque time series.

[0014] As a further aspect of the present invention, the specific function of the dynamic stiffness differential analytical module is as follows:

[0015] The differential ratio calculation submodule reads the state data of two adjacent sampling points in the joint angle-torque time series, calculates the torque change and angle change at adjacent times respectively, and performs a division operation when the angle change is detected to be non-zero, generating an instantaneous dynamic stiffness value.

[0016] The filtering and differentiation processing submodule constructs a Savitsky-Gorye convolution kernel including a data window of preset length, inputs the continuously generated instantaneous dynamic stiffness values ​​into the data window of preset length for polynomial least squares fitting, and performs second-order differentiation on the fitted curve equation to generate the second-order derivative value of stiffness.

[0017] As a further aspect of the present invention, the second-order differential calculation process of the instantaneous dynamic stiffness value by the filtering differential processing submodule includes:

[0018] Obtain multiple instantaneous dynamic stiffness values ​​at the current time and adjacent time points. Based on a preset convolution window length and polynomial order, calculate the second derivative value of the stiffness using the following formula:

[0019] ;

[0020] in, represent The value of the second derivative of the stiffness calculated at time t. An integer value representing the width of the filter's half-window. These represent the second-order differential convolution coefficients corresponding to the Savitzky-Gorley filter. Representative at The instantaneous dynamic stiffness value at time t. This represents the normalization factor value.

[0021] As a further aspect of the present invention, the specific function of the defensive contraction discrimination module is as follows:

[0022] The threshold adaptive setting submodule collects the patient's baseline muscle tone data in a non-training state, calculates the standard deviation of the baseline muscle tone data and multiplies it by a preset safety factor, and combines it with the maximum tolerance stiffness change rate in historical training to generate a stiffness mutation safety threshold.

[0023] The mutation feature recognition submodule monitors the input stiffness second derivative value in real time. When the value exceeds the stiffness mutation safety threshold for three consecutive sampling cycles, it determines that a defensive contraction event has occurred and triggers an angle locking signal to extract the pain critical angle value.

[0024] As a further aspect of the present invention, the process by which the mutation feature recognition submodule extracts the pain threshold angle value includes:

[0025] The timestamp of the defensive contraction event is identified, the joint angle-torque time series is traced back to locate the original data frame corresponding to the timestamp, the angle position information in the original data frame is extracted, the extracted angle position information is denoised and smoothed to generate the pain critical angle value.

[0026] As a further aspect of the present invention, the specific function of the load boundary dynamic reconfiguration module is as follows:

[0027] The buffer margin calculation submodule obtains the current joint motion angular velocity value, dynamically adjusts the safety margin based on the joint motion angular velocity value, and adds the basic fixed buffer value to the dynamic velocity compensation value to generate a safety buffer angle value.

[0028] The braking strategy execution submodule uses the pain threshold angle value minus the safety buffer angle value to determine the current motion stroke endpoint position, converts the current motion stroke endpoint position into a pulse control signal for the servo motor, generates the maximum activity angle upper limit and sends it to the driver.

[0029] As a further aspect of the present invention, the calculation process by which the buffer margin calculation submodule generates the safe buffer angle value includes:

[0030] The safe buffer angle value is calculated using the following formula, based on the real-time monitored joint angular velocity modulus and the preset neural response delay time:

[0031] ;

[0032] in, This represents the calculated safe buffer angle value. This represents the preset minimum mechanical safety distance value. The magnitude of the joint angular velocity at the current moment. This represents the total delay time between the system's braking response and neural reaction. Represents the stiffness sensitivity coefficient. Represents the current acceleration. This represents the maximum permissible acceleration of the system.

[0033] As a further aspect of the present invention, the specific process by which the braking strategy execution submodule sends the reverse braking command includes:

[0034] After determining the maximum active angle limit, a motion interruption signal with the highest priority is immediately generated. Based on the current motor speed, an S-shaped smooth deceleration curve is planned, and the motor is controlled to move in the opposite direction along the S-shaped smooth deceleration curve to a safe area. When the speed drops to zero, the position loop locking function is activated, and a reverse braking command is generated.

[0035] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0036] In this invention, a time series is constructed by synchronously collecting joint angle and torque data. The instantaneous dynamic stiffness change trend is analyzed by combining the torque and angle difference. The Savitzky-Gore filter is used to perform second-order differential processing on the dynamic stiffness to accurately capture the abrupt change point of the second derivative of stiffness. This identifies the critical moment when the patient's joint experiences defensive contraction pain. Based on the difference between the critical angle and the safety buffer, the upper limit of the maximum range of motion is dynamically reconstructed. The control command is directly fed back to the drive end to execute reverse braking. This realizes the leap from single-position monitoring to multi-dimensional mechanical state perception, ensuring that the protective gear can automatically adjust the activity boundary according to the patient's real-time physiological feedback, effectively preventing secondary damage caused by forcibly exceeding the pain threshold. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the main module interaction principle of the intelligent orthopedic rehabilitation training monitoring system of the present invention;

[0038] Figure 2 This is a flowchart illustrating the basic data synchronization and time series construction process of this invention.

[0039] Figure 3 This is a flowchart of the dynamic stiffness differential analysis and second derivative generation process of this invention;

[0040] Figure 4 This is a flowchart of the defensive contraction discrimination and pain threshold angle extraction of the present invention;

[0041] Figure 5 This is a flowchart of the load boundary dynamic reconstruction and reverse braking control of the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the software-based technical solution is described in detail below with reference to system architecture diagrams and embodiments. It should be understood that the specific embodiments described herein are only for explaining the technical solutions of this invention and do not constitute a limitation on the scope of protection.

[0043] In the description of this invention, the system architecture relationships or data processing flows indicated by terms such as "layer," "module," "interface," "data flow," "client," and "server" are all defined based on the architecture diagram or flowchart corresponding to the embodiments. This way of describing is only used to clearly illustrate the logical relationships between the elements in the technical solution, and not to limit the physical deployment form. The term "multiple" includes two or more technical units, including but not limited to multiple data nodes, processing threads, service instances, or functional components and other scalable elements. The specific number is determined according to the actual business scenario and needs to be specifically specified.

[0044] Please see Figure 1 and Figure 2 This invention provides a technical solution: an intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing, comprising:

[0045] The basic data synchronous acquisition module obtains joint angle values ​​and joint torque values, pairs the joint angle values ​​and joint torque values ​​to construct a joint angle-torque time series, and transmits the joint angle-torque time series to the dynamic stiffness differential analysis module.

[0046] The joint angle-torque time series includes the joint angle values ​​and joint torque values ​​at the same sampling time.

[0047] The specific functions of the basic data synchronization acquisition module are as follows:

[0048] The timestamp alignment submodule acquires the original encoder signal and the original torque sensor signal, extracts the timestamp information of the original encoder signal and the original torque sensor signal, performs linear interpolation alignment processing, removes invalid data frames with time deviations exceeding the preset synchronization window, and generates synchronized joint data frames.

[0049] The sequence splicing construction submodule extracts the effective angle components and effective torque components from the synchronous joint data frames, combines the effective angle components and effective torque components into a multi-dimensional vector format according to the time order, and continuously stores the multi-dimensional vector format to establish the joint angle-torque time series.

[0050] The basic data synchronous acquisition module obtains joint angle values ​​and joint torque values, pairs the joint angle values ​​and joint torque values ​​to construct a joint angle-torque time series, and transmits the joint angle-torque time series to the dynamic stiffness differential analysis module.

[0051] The joint angle-torque time series includes the joint angle values ​​and joint torque values ​​at the same sampling time.

[0052] The specific functions of the basic data synchronization acquisition module are as follows:

[0053] The timestamp alignment submodule reads the position register values ​​of the servo motor encoder and the torque register values ​​of the end-effector six-dimensional torque sensor at a sampling frequency of 1000Hz via the EtherCAT real-time industrial Ethernet bus. The system performs this process in each sampling cycle. Each entry records the timestamp of the original encoder signal arriving at the main control unit. The timestamp of the original torque sensor signal arriving at the main control unit When a difference in timestamps is detected between the two, i.e. At that time, using encoder timestamps Use this as the baseline timeline. Extract... Two adjacent torque sensor sampling points before and after the time point and ,in Perform linear interpolation operations to calculate... Alignment torque value at time 1 The calculation formula is Set the preset synchronization window to 3ms. If the absolute value of the difference is greater than 3ms, the data frame is determined to be invalid due to excessive transmission delay and is directly discarded without entering the subsequent queue. For the retained data, Joint angle at any moment With the calculated Bind and generate synchronized joint data frames.

[0054] The aforementioned EtherCAT real-time industrial Ethernet bus refers to a high-performance real-time Ethernet fieldbus communication protocol used in factory automation and process control. This protocol uses a "processing on the fly" processing method to enable data frames to be read and processed as they pass through slave devices, thereby greatly improving data transmission efficiency and system synchronization. It is suitable for multi-axis motion control and data acquisition scenarios with strict time synchronization requirements.

[0055] The sequence splicing construction submodule allocates a double-ended queue buffer in memory with a capacity of 5000 data points. It extracts angle components from the synchronization joint data frames. (Unit: degrees) and torque components (Unit: Newton-meter). According to The data is encapsulated into a multidimensional vector in ascending order. .Will Push it to the end of the queue, when the queue length reaches the preset analysis window length (e.g., When ), output all vector sequences in the current queue to construct a joint angle-torque time series.

[0056] Table 1 shows a comparison of the data before and after the basic data synchronous acquisition module performs timestamp alignment processing on the original signals during a knee flexion and extension training exercise.

[0057] Table 1. Comparison of Original Signal and Synchronized Data:

[0058] ;

[0059] Table 1 lists the specific values ​​of the basic data synchronous acquisition module when processing data from frames 1001 to 1004. Frame 1003 was discarded because the torque sensor timestamp lagged by more than 3ms. Frame 1002 obtained a torque value of 10.58Nm corresponding to time 1002.0ms through interpolation, ensuring that the timing of subsequent dynamic stiffness calculations strictly corresponded.

[0060] Please see Figure 1 and Figure 3 The dynamic stiffness differential analysis module calculates the torque difference and angle difference based on the joint angle-torque time series. It divides the torque difference by the angle difference to obtain the instantaneous dynamic stiffness value. It calls the Savitzky-Gore filter to perform second-order differentiation on the instantaneous dynamic stiffness value to obtain the second derivative value of stiffness and the joint angle value at the current moment. It then transmits the second derivative value of stiffness and the joint angle value at the current moment to the defensive contraction discrimination module.

[0061] The numerical value of the second derivative of stiffness specifically refers to the second-order differential calculation result that characterizes the degree of acceleration of joint stiffness change;

[0062] The specific functions of the dynamic stiffness differential analysis module are as follows:

[0063] The differential ratio calculation submodule reads the state data of two adjacent sampling points in the joint angle-torque time series, calculates the torque change and angle change at adjacent times respectively, and performs a division operation when the angle change is detected to generate the instantaneous dynamic stiffness value.

[0064] The filtering and differentiation processing submodule constructs a Savitsky-Gorye convolution kernel including a data window of preset length, inputs the continuously generated instantaneous dynamic stiffness values ​​into the data window of preset length for polynomial least squares fitting, and performs second-order differentiation on the fitted curve equation to generate the second-order derivative value of stiffness.

[0065] The second-order differential calculation process for the instantaneous dynamic stiffness value in the filtering differential processing submodule includes:

[0066] Obtain multiple instantaneous dynamic stiffness values ​​for the current time and adjacent time steps. Based on the preset convolution window length and polynomial order, calculate the second derivative of the stiffness using the following formula:

[0067] ;

[0068] in, represent The value of the second derivative of stiffness calculated at time ... An integer value representing the width of the filter's half-window. These represent the second-order differential convolution coefficients corresponding to the Savitzky-Gorley filter. Representative at The instantaneous dynamic stiffness value at time t. This represents the normalization factor value.

[0069] The specific functions of the dynamic stiffness differential analysis module are as follows:

[0070] The differential ratio calculation submodule reads the current moment from the joint angle-torque time series. Compared with the previous sampling time The status data are respectively and The change in torque at adjacent time points is obtained by performing a subtraction operation. With angle change Set the minimum angle change resolution threshold to 0.01 degrees. If so, the stiffness value from the previous moment is used as the current value; if Then the division operation is performed. Generate instantaneous dynamic stiffness values.

[0071] The filtering and differentiation processing submodule calls the Savitzky-Golay filter algorithm library to perform smoothing and second-order differentiation processing on the instantaneous dynamic stiffness numerical sequence. It sets the filter half-window width. (i.e., window length is) The polynomial fitting order is 2. This process not only suppresses noise but also directly calculates the second derivative of the stiffness curve, reflecting the accelerated change characteristics of stiffness. This step is performed according to the following formula:

[0072] ;

[0073] in, represent The value of the second derivative of stiffness calculated at time t, in units of t. , used to quantify the acceleration of stiffness changes; The integer value representing the width of the filter's half-window is taken as 2 here, which means that the calculation needs to cover 2 data points before and after the current point. The standard convolution coefficient sequence representing the second-order differential convolution coefficients of the Savitzky-Gore filter, for a window length of 5 and using second-order polynomial fitting, is as follows: ; Representative at The instantaneous dynamic stiffness value at a given moment; This represents the normalization factor value. For the above parameter settings, its value is given by the formula... The decision is made because the sampling interval is considered to be a unit step size of 1 and is based on the standard coefficient table. .

[0074] The Savitzky-Gore filter mentioned above refers to a polynomial smoothing algorithm based on the least squares principle. This algorithm uses a moving window to perform polynomial fitting on the data in the time domain, thereby removing signal noise while effectively preserving the waveform characteristics of the signal, such as peaks, troughs, and width. It can also directly calculate the first or higher-order derivatives of the signal through convolution operations.

[0075] Actual calculation example: capturing a specific moment during rehabilitation training. and its adjacent instantaneous dynamic stiffness numerical sequences (unit: ):

[0076]

[0077] Substitute the above values ​​into the formula to calculate the numerator:

[0078]

[0079] The final calculation yielded the numerical value of the second derivative of stiffness. .

[0080] Table 2. Data table for the calculation process of the second derivative of stiffness:

[0081] ;

[0082] Table 2 shows the calculation results for consecutive time points. This indicates that at the current moment, the joint stiffness is in an accelerated increase phase, with its rate of change increasing at a rate of 1.5 units. This value will be used later for comparison with the safety threshold.

[0083] Please see Figure 1 and Figure 4 The defensive contraction discrimination module obtains the stiffness mutation safety threshold. When the value of the second derivative of stiffness is greater than the stiffness mutation safety threshold, the current joint angle value is extracted as the pain critical angle value, and the pain critical angle value is transmitted to the load boundary dynamic reconstruction module.

[0084] The pain threshold angle value is specifically the joint angle value at the current moment when the stiffness abrupt change characteristic is identified.

[0085] The specific functions of the defensive contraction discrimination module are as follows:

[0086] The threshold adaptive setting submodule collects the patient's baseline muscle tone data in a non-training state, calculates the standard deviation of the baseline muscle tone data and multiplies it by a preset safety factor, and combines it with the maximum tolerance stiffness change rate in historical training to generate a stiffness mutation safety threshold.

[0087] The mutation feature recognition submodule monitors the input stiffness second derivative value in real time. When the value exceeds the stiffness mutation safety threshold for three consecutive sampling cycles, it determines that a defensive contraction event has occurred and triggers an angle locking signal to extract the pain critical angle value.

[0088] The process of extracting the pain threshold angle value by the mutation feature recognition submodule includes:

[0089] The timestamp of the defensive contraction event is identified, and the joint angle-torque time series is traced back to locate the original data frame corresponding to the timestamp. The angle position information in the original data frame is extracted, and the extracted angle position information is denoised and smoothed to generate the pain critical angle value.

[0090] The specific functions of the defensive contraction discrimination module are as follows:

[0091] The threshold adaptive setting submodule, in the patient's resting state before training begins, drives the motor to perform three reciprocating passive movements of the affected limb within a pain-free range. Baseline muscle tone data (mapped to the second derivative of stiffness) is collected during this process to obtain a sample set. Calculate the average of the set. and standard deviation Simultaneously, historical training records were retrieved from the patient's electronic medical record database, and the rate of change of maximum tolerable stiffness when no pain was triggered in the previous treatment course was extracted. Set a safety factor. (This coefficient is selected based on the normal distribution) (In accordance with principles and with added redundancy), historical weighting coefficients are set. The threshold calculation rule is as follows: .

[0092] Actual calculation threshold setting process: After data collection and calculation, the mean value of the second derivative of stiffness under resting state is... Standard deviation Maximum rate of change of tolerance stiffness in historical records Substitute the values:

[0093] .

[0094] Therefore, the safety threshold for sudden stiffness change is set as follows: .

[0095] The mutation feature recognition submodule establishes a shift register of length 3 to store the latest stiffness second derivative value in real time. The logic judgment unit compares the value in the register with the stiffness change safety threshold in each sampling cycle. If three consecutive values ​​in the register... All strictly greater than If this occurs, a defensive contraction event is determined to have taken place. At this point, the current timestamp is immediately locked. .

[0096] The aforementioned shift register refers to a sequence of storage cells used in digital circuits or computer software. Under the control of clock pulses, it can move the stored data bit by bit in a specified direction (left or right shift). It is typically used for serial-to-parallel data conversion, delay processing, or as a temporary buffer queue for historical data.

[0097] The mutation feature identification submodule is based on The corresponding original data frame is indexed in the joint angle-torque time series buffer. The joint angle values ​​in that frame are then read. Because the raw angle data may contain high-frequency jitter from the encoder, a mean filter with a window length of 5 is applied. The average of the angle values ​​at two points before and after the pain threshold is recorded as the critical angle value. For example, in Table 2 above, time... The value of 1.50 is less than the threshold of 1.6, so it is not triggered; while at time... The value of 2.29 is greater than the threshold of 1.6. If the subsequent... , The time values ​​are 2.50 and 2.80, both greater than 1.6, therefore... The trigger is confirmed at all times, and the corresponding angle is retrieved by backtracking, assuming that it has been smoothed. .

[0098] Please see Figure 1 and Figure 5 The load boundary dynamic reconstruction module calculates the difference between the pain threshold angle value and the safety buffer angle value to obtain the maximum activity angle limit, writes the maximum activity angle limit to the servo controller, and sends a reverse braking command to the motor driver.

[0099] The maximum active angle limit specifically refers to the safety boundary value used to update the servo controller's travel parameters. Reverse braking commands include motor smooth deceleration signals and position lock signals.

[0100] The specific functions of the load boundary dynamic reconfiguration module are as follows:

[0101] The buffer margin calculation submodule obtains the current joint motion angular velocity value, dynamically adjusts the safety margin based on the joint motion angular velocity value, and adds the basic fixed buffer value to the dynamic velocity compensation value to generate the safety buffer angle value.

[0102] The calculation process for generating the safe buffer angle value in the buffer margin calculation submodule includes:

[0103] The safe buffer angle value is calculated using the following formula, based on the real-time monitored joint angular velocity modulus and the preset neural response delay time:

[0104] ;

[0105] in, This represents the calculated safe buffer angle value. This represents the preset minimum mechanical safety distance value. The magnitude of the joint angular velocity at the current moment. This represents the total delay time between the system's braking response and neural reaction. Represents the stiffness sensitivity coefficient. Represents the current acceleration. Represents the maximum permissible acceleration of the system;

[0106] The braking strategy execution submodule uses the pain threshold angle value minus the safety buffer angle value to determine the current motion stroke end position, converts the current motion stroke end position into a pulse control signal for the servo motor, generates the maximum activity angle upper limit and sends it to the driver.

[0107] The specific process of the braking strategy execution submodule sending the reverse braking command includes:

[0108] After determining the maximum upper limit of the maximum active angle, the highest priority motion interruption signal is immediately generated. Based on the current motor speed, an S-shaped smooth deceleration curve is planned, and the motor is controlled to move in the opposite direction along the S-shaped smooth deceleration curve to the safe area. When the speed drops to zero, the position loop locking function is activated, and a reverse braking command is generated.

[0109] The specific functions of the load boundary dynamic reconfiguration module are as follows:

[0110] The buffer margin calculation submodule obtains the current joint motion angular velocity modulus through the differential encoder position signal. The current acceleration is obtained by second derivative. The preset neural response delay time is read from the system configuration file. (Set to 150ms, or 0.15s, covering the total lag in the transmission of electromyographic signals to the robotic arm actuator) and minimum mechanical safety distance. (Set at 2.0 degrees). Calculate the safety buffer angle value using the following formula:

[0111] ;

[0112] in, This represents the calculated safe buffer angle value, in degrees. This represents the preset minimum mechanical safety distance value, taken as 2.0 degrees; The modulus of the joint angular velocity at the current moment, in units of ; The total delay time between the system's braking response and neural reaction is taken as 0.15s; This represents the stiffness sensitivity coefficient, with a value of 0.5, which is used to adjust the correction weight of the acceleration term on the buffer distance. Represents the current acceleration, in units of ; This represents the maximum permissible acceleration of the system; the system hardware limit is 200. The advantage of this formula lies in the introduction of an acceleration percentage term. Dynamic nonlinear compensation is applied to the buffer distance caused by speed, which significantly increases the buffer margin when the joint is in a state of high acceleration, preventing insufficient braking distance due to excessive inertia.

[0113] Actual calculation example: Assuming that when the defensive contraction decision is triggered, the system detects: current angular velocity Current acceleration Substitute the previously defined parameters into the calculation:

[0114]

[0115] The results indicate that in order to safely counteract the current motion inertia and the risk of neural delay, the system needs to reserve a buffer space of 5.75 degrees.

[0116] The braking strategy execution submodule reads the pain threshold angle value obtained from the previous steps. The maximum active angle is determined by performing a subtraction operation: The submodule will The pulse count is converted to the corresponding pulse count for the servo motor (assuming an encoder resolution of 10,000 pulses / degree, then it corresponds to 597,500 pulses), and this value is written to the driver's soft limit register. Simultaneously, a reverse braking command is generated. Based on the current speed of 20 degrees / s, the system plans a smooth S-shaped deceleration curve with a duration of 200 ms, with the target position being... Upon receiving the command, the servo drive controls the motor to move in the opposite direction along the curve, forcing the affected limb back to a safe area, and upon reaching... The speed is reduced to 0, and then the position loop PID locking function is activated to maintain the motor shaft's anti-disturbance capability at this position, thus completing the protective braking.

[0117] The aforementioned S-shaped smooth deceleration curve refers to a speed planning curve commonly used in motion control. Its rate of change of speed (i.e. acceleration) exhibits a continuous S-shaped change during the start-up and stopping phases. Compared to the trapezoidal speed curve, the S-shaped curve can significantly reduce mechanical shock and residual vibration of flexible loads during motor acceleration and deceleration by limiting the derivative of acceleration (jerk / Jerk).

[0118] The above embodiments illustrate preferred embodiments of the present invention. Any equivalent adjustments to the technical solution based on software engineering methods are within the scope of protection, including but not limited to: implementing algorithm logic using different programming languages, refactoring functional modules into services, adjusting data interaction protocols, and optimizing resource scheduling strategies. Any implementation scheme derived from reasonable modifications to the data processing flow, service call chain, or system architecture layer without departing from the core technology of the present invention should be considered within the scope of protection defined by the claims of the present invention.

Claims

1. An intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing, characterized in that, The system includes: The basic data synchronous acquisition module acquires joint angle values ​​and joint torque values, pairs the joint angle values ​​and joint torque values ​​to construct a joint angle-torque time series, and transmits the joint angle-torque time series to the dynamic stiffness differential analysis module. The dynamic stiffness differential analysis module calculates the torque difference and angle difference based on the joint angle-torque time series, divides the torque difference by the angle difference to obtain the instantaneous dynamic stiffness value, calls the Savitzky-Gore filter to perform second-order differentiation on the instantaneous dynamic stiffness value to obtain the second-order derivative value of stiffness and the joint angle value at the current moment, and transmits the second-order derivative value of stiffness and the joint angle value at the current moment to the defensive contraction discrimination module; The defensive contraction discrimination module obtains the stiffness mutation safety threshold. When the value of the second derivative of stiffness is greater than the stiffness mutation safety threshold, the current joint angle value is extracted as the pain critical angle value, and the pain critical angle value is transmitted to the load boundary dynamic reconstruction module. The load boundary dynamic reconstruction module calculates the difference between the pain threshold angle value and the safety buffer angle value to obtain the maximum activity angle limit, writes the maximum activity angle limit into the servo controller, and sends a reverse braking command to the motor driver.

2. The intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing according to claim 1, characterized in that, The joint angle-torque time series includes the joint angle value and the joint torque value at the same sampling time. The second derivative value of stiffness specifically refers to the second derivative calculation result that characterizes the degree of acceleration of joint stiffness change. The pain critical angle value is specifically the joint angle value at the current time when stiffness mutation characteristics are identified. The upper limit of the maximum range of motion specifically refers to the safety boundary value used to update the stroke parameters of the servo controller. The reverse braking command includes a motor smooth deceleration signal and a position lock signal.

3. The intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing according to claim 1, characterized in that, The specific functions of the basic data synchronous acquisition module are as follows: The timestamp alignment submodule acquires the original encoder signal and the original torque sensor signal, extracts the timestamp information of the original encoder signal and the original torque sensor signal, performs linear interpolation alignment processing, removes invalid data frames with time deviations exceeding the preset synchronization window, and generates synchronization joint data frames. The sequence splicing construction submodule extracts the effective angle components and effective torque components from the synchronized joint data frames, combines the effective angle components and effective torque components into a multi-dimensional vector format according to the time order, and continuously stores the multi-dimensional vector format to establish a joint angle-torque time series.

4. The intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing according to claim 1, characterized in that, The specific function of the dynamic stiffness differential analytical module is as follows: The differential ratio calculation submodule reads the state data of two adjacent sampling points in the joint angle-torque time series, calculates the torque change and angle change at adjacent times respectively, and performs a division operation when the angle change is detected to be non-zero, generating an instantaneous dynamic stiffness value. The filtering and differentiation processing submodule constructs a Savitsky-Gorye convolution kernel including a data window of preset length, inputs the continuously generated instantaneous dynamic stiffness values ​​into the data window of preset length for polynomial least squares fitting, and performs second-order differentiation on the fitted curve equation to generate the second-order derivative value of stiffness.

5. The intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing according to claim 4, characterized in that, The second-order differential calculation process of the instantaneous dynamic stiffness value by the filtering differential processing submodule includes: Obtain multiple instantaneous dynamic stiffness values ​​at the current time and adjacent time points. Based on a preset convolution window length and polynomial order, calculate the second derivative value of the stiffness using the following formula: ； in, represent The value of the second derivative of the stiffness calculated at time t. An integer value representing the width of the filter's half-window. These represent the second-order differential convolution coefficients corresponding to the Savitzky-Gorley filter. Representative at The instantaneous dynamic stiffness value at time t. This represents the normalization factor value.

6. The intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing according to claim 1, characterized in that, The specific function of the defensive contraction discrimination module is as follows: The threshold adaptive setting submodule collects the patient's baseline muscle tone data in a non-training state, calculates the standard deviation of the baseline muscle tone data and multiplies it by a preset safety factor, and combines it with the maximum tolerance stiffness change rate in historical training to generate a stiffness mutation safety threshold. The mutation feature recognition submodule monitors the input stiffness second derivative value in real time. When the value exceeds the stiffness mutation safety threshold for three consecutive sampling cycles, it determines that a defensive contraction event has occurred and triggers an angle locking signal to extract the pain critical angle value.

7. The intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing according to claim 6, characterized in that, The process by which the mutation feature recognition submodule extracts the pain threshold angle value includes: The timestamp of the defensive contraction event is identified, the joint angle-torque time series is traced back to locate the original data frame corresponding to the timestamp, the angle position information in the original data frame is extracted, the extracted angle position information is denoised and smoothed to generate the pain critical angle value.

8. The intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing according to claim 1, characterized in that, The specific functional implementation of the load boundary dynamic reconfiguration module is as follows: The buffer margin calculation submodule obtains the current joint motion angular velocity value, dynamically adjusts the safety margin based on the joint motion angular velocity value, and adds the basic fixed buffer value to the dynamic velocity compensation value to generate a safety buffer angle value. The braking strategy execution submodule uses the pain threshold angle value minus the safety buffer angle value to determine the current motion stroke endpoint position, converts the current motion stroke endpoint position into a pulse control signal for the servo motor, generates the maximum activity angle upper limit and sends it to the driver.

9. The intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing according to claim 8, characterized in that, The calculation process by which the buffer margin calculation submodule generates the safe buffer angle value includes: The safe buffer angle value is calculated using the following formula, based on the real-time monitored joint angular velocity modulus and the preset neural response delay time: ； in, This represents the calculated safe buffer angle value. This represents the preset minimum mechanical safety distance value. The magnitude of the joint angular velocity at the current moment. This represents the total delay time between the system's braking response and neural reaction. Represents the stiffness sensitivity coefficient. Represents the current acceleration. This represents the maximum permissible acceleration of the system.

10. The intelligent orthopedic rehabilitation training monitoring system based on multimodal sensing according to claim 8, characterized in that, The specific process by which the braking strategy execution submodule sends the reverse braking command includes: After determining the maximum active angle limit, a motion interruption signal with the highest priority is immediately generated. Based on the current motor speed, an S-shaped smooth deceleration curve is planned, and the motor is controlled to move in the opposite direction along the S-shaped smooth deceleration curve to a safe area. When the speed drops to zero, the position loop locking function is activated, and a reverse braking command is generated.

Images