Adjustable postoperative rehabilitation trolley

Abstract

The invention discloses an adjustable postoperative rehabilitation trolley, and relates to the field of intelligent rehabilitation. According to the scheme, the vehicle comprises a vehicle frame, a rolling wheel frame is fixed to the bottom of the vehicle frame, and rolling wheels are rotationally connected into the rolling wheel frame; the resistance adjusting mechanism is used for controlling the rotating resistance of the rolling wheels through hydraulic pressure; an oil pump; the acquisition module is used for acquiring limb feature data of a user; the analysis module is used for filtering and normalizing the original data of the sensor module to generate an input feature vector; the intention recognition module judges the rehabilitation intention of the user through a trained intention recognition model based on the input feature vector, and the rehabilitation intention comprises active, fatigue, rejection and abnormal postures; and the control module is used for generating real-time rehabilitation strength by combining a fuzzy control rule and a PID dynamic compensation algorithm according to an output result of the intention recognition module. The training strength is intelligently adjusted, and the rehabilitation training effect is affected.

Description

Technical Field

[0001] The present invention relates to the field of intelligent rehabilitation, and particularly to an adjustable rehabilitation cart for postoperative use. Background Art

[0002] Rehabilitation training is an essential link after surgery. A complete rehabilitation training can restore the patient's limbs to a healthy state, so rehabilitation training is particularly important.

[0003] A rehabilitation cart is a device to assist patients in rehabilitation training. For example, a fall-proof walking aid for lower limb rehabilitation training and a rehabilitation training method disclosed in the authorized announcement number CN110353952B include a frame designed with a wider upper part and a narrower lower part; universal wheels located at the bottom of the frame; a collision prevention frame located at the bottom of the frame; a seat cushion located in the internal space of the frame; elastic ropes connecting the seat cushion and the frame, and the seat cushion and the elastic force adjusting device; a tray rack located at the top of the frame; the front end of the tray rack and the frame connecting rod can both be opened.

[0004] This patent enables patients to push the cart body for rehabilitation training. However, since the training intensity of patients varies in each stage and every day, there is a doctor's guidance during training in the hospital. If training outdoors by themselves, patients cannot control the training intensity well. Currently, the rehabilitation cart can only provide a supporting effect but cannot intelligently adjust the training intensity, which affects the rehabilitation training effect. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to propose an adjustable rehabilitation cart for postoperative use to achieve intelligent adjustment of the training intensity and affect the rehabilitation training effect.

[0006] To achieve the above technical purpose, the present invention provides an adjustable rehabilitation cart for postoperative use:

[0007] It includes: a frame with a roller frame fixed at the bottom, and a roller rotatably connected inside the roller frame; a resistance adjustment mechanism for controlling the rotational resistance of the roller using hydraulic pressure; an oil pump for providing hydraulic pressure to the resistance adjustment mechanism and controlling the hydraulic pressure value; a collection module for collecting user limb characteristic data; an analysis module for filtering and normalizing the original data of the sensor module to generate an input feature vector; an intention recognition module for judging the user's rehabilitation intention based on the input feature vector through a trained intention recognition model, where the rehabilitation intention includes active, fatigued, resistant, and abnormal postures; a control module for generating real-time rehabilitation intensity according to the output result of the intention recognition module, combined with fuzzy control rules and a PID dynamic compensation algorithm.

[0008] Preferably, it further includes:

[0009] A safety control module, which is used to determine whether there is a risk based on limb feature data. If there is a risk, an alarm instruction is generated.

[0010] An alarm module, which generates an alarm based on the alarm instruction.

[0011] Preferably, the method for the safety control module to determine whether there is a risk includes:

[0012] Preset a safety threshold and a continuous time threshold, and continuously collect i motion angle values and compare them with the preset safety threshold.

[0013] If i consecutive motion angle values are greater than the safety threshold and the continuous time is greater than the continuous time threshold, it is determined that there is a risk of falling.

[0014] If i consecutive motion angle values are greater than the safety threshold and the continuous time is less than the continuous time threshold, it is determined that there is no risk of falling.

[0015] If i consecutive motion angle values are less than the safety threshold and the continuous time is greater than the continuous time threshold, it is determined that there is no risk of falling.

[0016] If i consecutive motion angle values are less than the safety threshold and the continuous time is less than the continuous time threshold, it is not determined that there is no risk of falling.

[0017] Preferably, the limb feature data includes a pressure value, muscle activity, motion angle, and angular velocity.

[0018] Preferably, the method for collecting the muscle activity is: preset a sliding window; collect N electromyographic signals of the user's limbs in the sliding window, where N is an integer greater than 1, and convert them into muscle activity through root mean square calculation.

[0019] Preferably, the training method for the intention recognition model is:

[0020] The intention recognition module adopts a lightweight neural network classification model, and its input feature vector is limb feature data; the output of the neural network is the probability distribution of intention labels, and the label corresponding to the maximum probability is selected as the intention determination result.

[0021] Preferably, the fuzzy control rules of the control module include: when the user's intention is fatigue, the output force coefficient satisfies a preset force coefficient threshold one; when the user's intention is resistance, the output force coefficient satisfies a preset force coefficient threshold two; the force coefficient is calculated by weighted calculation of the membership function.

[0022] Preferably, the frame includes: a chassis distributed on both sides of the first connecting rod, and the chassis is slidably connected to the first connecting rod; a vertical rod fixed to the surface of the chassis; a second connecting rod with one end slidably connected to the vertical rod and the other end slidably connected to a support rod; and a handrail fixed to the end of the support rod.

[0023] Preferably, the resistance adjusting mechanism includes: a resistance disk coaxially connected and fixed to the roller; a disk clamp fixed to the roller frame; a liquid guide pipe fixed to the surface of the disk clamp for supplying oil to the disk clamp; a resistance block slidably and sealingly connected to the disk clamp for abutting against the resistance disk to control the rotational frictional force of the resistance disk; and a reed with one end fixed to the inner wall of the disk clamp and the other end abutting against the resistance block.

[0024] Preferably, a connecting block is fixed to the outer surface of the disk clamp, and the connecting block is fixedly connected to the disk clamp by bolts.

[0025] From the above technical solutions, it can be seen that the present application has the following beneficial effects:

[0026] By collecting the limb feature data of the user and based on the input feature vector, the intention recognition model that has been trained is used to judge the user's rehabilitation intention. The rehabilitation intention includes active, fatigued, resistant, and abnormal postures. According to the output result of the intention recognition module, combined with the fuzzy control rule and the PID dynamic compensation algorithm, a real-time rehabilitation force is generated to achieve intelligent adjustment of the training force and affect the rehabilitation training effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0028] Figure 1 It is a schematic diagram of the overall structure of an adjustable rehabilitation trolley for postoperative use provided by the present invention;

[0029] Figure 2 It is a schematic diagram of the overall structure of the resistance adjusting mechanism of an adjustable rehabilitation trolley for postoperative use provided by the present invention;

[0030] Figure 3 It is a schematic cross-sectional view of the resistance adjusting mechanism of an adjustable rehabilitation trolley for postoperative use provided by the present invention;

[0031] Figure 4 It is a schematic diagram of the intention recognition module structure of an adjustable rehabilitation trolley for postoperative use provided by the present invention;

[0032] Figure 5 Schematic structural diagram of the safety control module of an adjustable postoperative rehabilitation trolley provided by the present invention.

[0033] Description of the drawings: 1. Frame; 11. Underframe; 12. First connecting rod; 13. Vertical rod; 14. Second connecting rod; 15. Support rod; 16. Handrail frame; 2. Roller frame; 21. Roller; 3. Resistance adjustment mechanism; 31. Resistance disc; 32. Disc clamp; 321. Connecting block; 33. Liquid guide pipe; 34. Resistance block; 35. Reed; 4. Oil pump. Detailed implementation manners

[0034] The following description is essentially exemplary only and is not intended to limit the present disclosure, its application, and uses. It should be understood that in all these drawings, the same or similar reference numerals indicate the same or similar parts and features. Each drawing only schematically shows the concept and principle of the embodiments of the present disclosure, and does not necessarily show the specific dimensions and their ratios of the embodiments of the present disclosure. Specific parts in a specific drawing may be exaggerated to illustrate the relevant details or structures of the embodiments of the present disclosure.

[0035] Example 1, refer to Figure 1 、 Figure 2 and Figure 3 As shown in, an adjustable postoperative rehabilitation trolley includes a frame 1. In this embodiment, the frame 1 is made of aluminum alloy material, and a roller frame 2 is installed at the bottom. The roller 21 is rotatably connected to the roller frame 2 through a bearing, and the surface of the roller 21 is coated with an anti-slip rubber layer;

[0036] Specifically, the frame 1 adopts a modular design. The frame 1 includes at least two underframes 11. The two underframes 11 are slidably connected through a first connecting rod 12, and the first connecting rod 12 and the two underframes 11 are fixed through locking bolts, which facilitates the adjustment of the distance between the two underframes 11. For example, the adjustment range is set to 0 - 50 cm, so as to adapt to different user body types;

[0037] The vertical rod 13 is fixed to the surface of the underframe 11 by means of buckles or welding, etc. The second connecting rod 14 and the vertical rod 13 are connected through a telescopic sleeve to achieve adjustable height. After the second connecting rod 14 and the vertical rod 13 adjust the height, they are fixed by means of screws or quick-release buckles. For example, the set range is 80 - 120 cm, which is suitable for people of different heights to use; The end of the support rod 15 is slidably connected with a handrail frame 16, and is also fixed by a locking bolt.

[0038] Exemplarily, the user can quickly adjust the frame size according to height, such as a height of 150 - 190 cm, or the rehabilitation stage, such as sitting / standing training; in some embodiments, a fixed chair can be installed on the chassis 11 to adapt to sitting rehabilitation training; the first connecting rod 12 can be replaced with an electric push rod to achieve automatic adjustment; an anti-slip silicone sleeve can be added to the outer surface of the armrest frame 16;

[0039] Furthermore, the resistance adjustment mechanism 3 is coaxially connected to the roller 21, and the rotation resistance of the roller is controlled by hydraulic pressure; the resistance adjustment mechanism 3 includes a resistance disk 31, a disk clamp 32, a liquid guide pipe 33, a resistance block 34, and a reed 35. The resistance disk 31 is coaxially fixed to the roller 21, and an annular friction groove is formed on the surface. The disk clamp 32 is fixed to the roller frame 2 through a connecting block 321. A hydraulic cavity is provided inside the disk clamp 32. The liquid guide pipe 33 is fixed to the disk clamp 32. One end of the resistance block 34 passes through the disk clamp 32 and is slidably and sealingly connected to the disk clamp 32; the reed 35 is fixed inside the disk clamp 32, and the reed 35 is used to provide the elastic force required for the resistance block 34 to return to its original position;

[0040] Exemplarily, the liquid guide pipe 33 introduces the hydraulic oil output by the oil pump 4 into the hydraulic cavity of the disk clamp 32. The resistance block 34 slides along the inner wall of the disk clamp 32 under the hydraulic push and contacts the friction groove of the resistance disk 31, thereby changing the rotational friction force of the disk clamp 32, changing the rotational friction force of the roller 21, thereby changing the force required to push the frame, and changing the rehabilitation training intensity;

[0041] Specifically, the friction force calculation formula: F = μ·P·S, where μ is the friction coefficient (0.1 - 0.3), P is the oil pressure (MPa), and S is the contact area of the resistance block (cm 2 ); when the oil pressure decreases, the reed 35 provides a restoring force to retract the resistance block 34.

[0042] The oil pump 4 is fixed on the frame 1. The oil pump 4 uses a micro servo hydraulic pump and conveys hydraulic oil to the resistance adjustment mechanism 3 through the liquid guide pipe 33. For example, the pressure adjustment range is 0.1 - 5 MPa.

[0043] Embodiment 2, based on Embodiment 1, refer to Figure 4 As shown, an adjustable rehabilitation cart for postoperative use includes an acquisition module, an analysis module, an intention recognition module, and a control module, where each module and the oil pump 4 are connected by wire and / or wirelessly;

[0044] It is worth mentioning that the power supply for each module and the oil pump 4 is provided by a storage battery, and the storage battery is fixed on the frame 1.

[0045] Specifically, the acquisition module is used to collect the user's limb feature data; the acquisition module of this embodiment includes a pressure sensor, an electromyographic sensor (EMG), a motion angle sensor and a gyroscope. The pressure sensor is embedded in the gripping part of the armrest 16. The pressure sensor is used to detect the user's hand pressure value, and the pressure value is marked as F. For example, the range is set to 0-500N; the electromyographic sensor (EMG) is attached to the user's limb muscle group to collect muscle electrical signals (sampling rate 1kHz), and the muscle activity is obtained based on the muscle electrical signal formula calculation. The specific calculation method is shown below, and the muscle activity is marked as A muscle ; The motion angle sensor is attached to the user's joint or fixed to the user's joint by means of a strap, and is used to detect the joint motion angle value, and the motion angle value is marked as θ; the gyroscope is integrated inside the support rod 15, and the gyroscope is used to detect the angular velocity, and the angular velocity is marked as ω, for example, the range is set to ±2000° / s;

[0046] It is worth mentioning that the electromyography sensor (EMG) and the motion angle sensor are powered and transmit data through a retractable cable, such as a spiral cable, or can be powered by a wearable power supply, which is not specifically limited here.

[0047] Specifically, the analysis module is used to filter and normalize the raw data of the sensor module. The input feature vector is marked as X, X = [F, θ, ω, A muscle ].

[0048] The intention recognition module judges the user's rehabilitation intention based on the input feature vector through the trained intention recognition model. Rehabilitation intentions include activity, fatigue, resistance and abnormal posture. It uses a lightweight convolutional neural network (CNN) to output the intention label after inputting the feature vector.

[0049] The control module generates real-time rehabilitation force according to the output result of the intention recognition module, combined with fuzzy control rules and PID dynamic compensation algorithm. Specifically, the fuzzy-PID hybrid algorithm is called according to the rehabilitation intention label to generate a target hydraulic pressure instruction and send it to the oil pump 4.

[0050] Exemplarily, when the user pushes the cart, the system recognizes the rehabilitation intention in real time and adjusts the roller resistance, for example, the resistance is reduced by 30% in a fatigue state; abnormal posture, such as a roll angle >15° triggers emergency braking, and the response time is <0.2 seconds.

[0051] In some embodiments, the oil pump 4 can be replaced by an electromagnetic brake, and the roller resistance can be controlled by electric current; the electromyography sensor can be replaced by a surface electromyography patch.

[0052] Specifically, the limb feature data includes pressure value, muscle activity, movement angle, and angular velocity. The pressure value reflects the magnitude of the force exerted by the user's upper limb on the frame 1. If a large force is exerted, it indicates insufficient lower limb support force, and the upper limb needs to provide a large force to support the body. Muscle activity reflects the vitality of the body. Movement angle and angular velocity determine whether the work is flexible and standard.

[0053] Furthermore, the method for collecting muscle activity is as follows:

[0054] Preset a sliding window with a window length T = 100 ms and a sliding step ΔT = 50 ms. Collect N electromyogram signals of the user's limbs within the sliding window, where N is an integer greater than 1. That is, the number of sampling points N within each window = T × sampling rate = 100. After root mean square calculation, it is converted into muscle activity. The calculation formula is:

[0055]

[0056] Among them,

[0057]

[0058] E i is the original value of the nth electromyogram signal within the window, where n = 1, 2, 3,..., N, is the window mean;

[0059] Normalization processing:

[0060]

[0061] Among them, A min A max is calibrated according to the user's static relaxation state and maximum activity; for the movement angle value filtering, Kalman filtering is used to eliminate jitter noise, and the Kalman gain K is dynamically adjusted according to the IMU noise covariance. Exemplarily, the processing delay of the electromyogram signal < 5 ms, and the measurement error of the movement angle value < 1°. It supports dynamic updating of the muscle activity baseline value to adapt to the electromyogram characteristics of different users.

[0062] Specifically, the training method of the intention recognition model is as follows:

[0063] The intention recognition module adopts a lightweight neural network classification model, and its input feature vector is the limb feature data; the output of the neural network is the probability distribution of the intention labels, and the label corresponding to the maximum probability is selected as the intention determination result;

[0064] The specific training method includes: dataset construction, collecting limb feature data of g groups of postoperative rehabilitation patients. The limb feature data are respectively pressure value, muscle activity, movement angle value and angular velocity, where g is an integer greater than or equal to 1, and labeling rehabilitation intention labels, which are respectively active, fatigued, resistant and abnormal posture;

[0065] Data augmentation, adding Gaussian noise, marked as σ, for example, σ = 0.05, and time translation setting, for example, ±10ms, to improve generalization. The specific data of σ and time translation setting are determined by those skilled in the art according to actual needs and are not specifically limited here;

[0066] Neural network structure, the input layer includes 4 nodes, corresponding to the normalized features

[0067] The hidden layer includes:

[0068] The first layer has 16 nodes, and the activation function is ReLU (f(x) = max(0, x));

[0069] The second layer has 8 nodes, and the activation function is ReLU;

[0070] The output layer has 4 nodes, and the activation function is Softmax (output probability distribution):

[0071]

[0072] Among them, W j is the weight vector, and b j is the bias term;

[0073] Training parameters, the loss function uses cross-entropy loss

[0074]

[0075] Among them, M is the batch size, for example, set to 32, and y ij is the one-hot encoding of the true label.

[0076] The optimizer is the Adam optimizer, the initial learning rate η = 0.001, the decay rate β1 = 0.9, and β2 = 0.999;

[0077] The training process is to iterate 50 epochs, and the early stopping method (patience = 5) is used to prevent overfitting;

[0078] Until the accuracy of the test set reaches 92.5%, and the confusion matrix shows that the discrimination between "fatigued" and "resistant" reaches the preset accuracy to complete the training. The preset accuracy is determined by the art according to the actual situation and is not specifically limited here.

[0079] Furthermore, the fuzzy control rules of the control module include: when the rehabilitation intention is fatigue, the output force coefficient meets the first preset force coefficient threshold; when the user intention is resistance, the output force coefficient meets the second preset force coefficient threshold; when the rehabilitation intention is active, the output force coefficient meets the third preset force coefficient threshold; when the rehabilitation intention is abnormal posture, the output force coefficient meets the fourth preset force coefficient threshold; the force coefficient is calculated by weighted calculation of the membership function; where the first preset force coefficient threshold, the second preset force coefficient threshold, the third preset force coefficient threshold, and the fourth preset force coefficient threshold are determined by those skilled in the art according to the actual situation and are not specifically limited herein. For example, when it is fatigue and resistance, the rehabilitation resistance is reduced; when it is active, the rehabilitation resistance is increased; when the rehabilitation intention is abnormal posture, the rehabilitation training is stopped and the rehabilitation resistance is reduced to approach 0;

[0080] Exemplarily, the following data are experimental data, and the specific data are obtained by those skilled in the art according to the actual situation. For example, when the rehabilitation intention is fatigue, the range of the force coefficient k is 0.5 ≤ k ≤ 0.8, then

[0081] when the intention is resistance, the range of the force coefficient k is: 1.2 ≤ k ≤ 1.5, then

[0082] when the intention is active, the range of the force coefficient k is: 0.8 ≤ k ≤ 1.2, then μ3 = 1 - μ1 - μ2;

[0083] Force coefficient synthesis:

[0084]

[0085] where k1 = 0.65, k2 = 1.35, k3 = 1.0 are the mid-values of each interval;

[0086] Then the PID parameter is dynamically adjusted: the proportional gain K p is negatively correlated with k, K p = 2.0 / k; the integral gain K i is fixed at 0.1 to eliminate the steady-state error; the derivative gain K d is positively correlated with the angular velocity ω, K d = 0.05 * |ω|;

[0087] The final rehabilitation force is marked as F KF Final rehabilitation force calculation:

[0088]

[0089] where, F maxPreset by the doctor, for example, the default is 200N, and sat is the limiting function.

[0090] For example, the step response test shows that the system adjustment time < 0.5 seconds and the overshoot < 5%; when the user suddenly gets fatigued, the rehabilitation force drops to the safe range of rehabilitation training within 0.3 seconds, avoiding muscle strains caused by excessive rehabilitation training.

[0091] Example 3, refer to Figure 5 As shown, on the basis of Example 1, an adjustable postoperative rehabilitation trolley further includes a safety control module and an alarm module, wherein each module among the safety control module, the alarm module, the acquisition module, the analysis module, the intention recognition module and the control module, as well as the oil pump 4, are connected by wire and / or wirelessly;

[0092] The safety control module is used to determine whether there is a risk based on the limb feature data. Specifically,

[0093] Preset a safety threshold and a continuous time threshold, and continuously collect i motion angle values and compare them with the preset safety threshold. i is an integer greater than 1. The preset safety threshold and the continuous time threshold are determined by those skilled in the art according to the actual situation and are not specifically limited herein;

[0094] If i consecutive motion angle values are greater than the safety threshold and the continuous time is greater than the continuous time threshold, it is determined that there is a risk of falling;

[0095] If i consecutive motion angle values are greater than the safety threshold and the continuous time is less than the continuous time threshold, it is determined that there is no risk of falling;

[0096] If i consecutive motion angle values are less than the safety threshold and the continuous time is greater than the continuous time threshold, it is determined that there is no risk of falling;

[0097] If i consecutive motion angle values are less than the safety threshold and the continuous time is less than the continuous time threshold, it is not determined that there is no risk of falling.

[0098] Exemplarily, set the continuous time threshold to 2 seconds. If the motion angle value θ exceeds the safety threshold for 2 seconds continuously, for example, the preset safety threshold is set to the roll angle > 15°, it is determined that there is a risk of falling.

[0099] Furthermore, based on the muscle activity A muscle judgment, if the muscle activity A muscle is continuously lower than the baseline value by 20% for 5 seconds, it is determined that the muscle is fatigued.

[0100] If it is determined that there is a risk of falling or muscle fatigue, a shutdown instruction is sent to the control module for emergency braking. The control module receives the shutdown instruction and controls the oil pump 4. The oil pump 4 releases pressure to 0 MPa, and the reed 35 releases elastic force to the resistance block 34, causing the resistance block 34 to disengage from the resistance disk 31;

[0101] At the same time, the safety control module generates an alarm instruction and sends the alarm instruction to the alarm module. After receiving the alarm instruction, the alarm module triggers an audible and visual alarm. For example, the buzzer frequency is 2 kHz, and the LED red light flashes. The audible and visual alarm is fixed on the vehicle frame 1.

[0102] Exemplarily, in the simulation test, the average time from abnormal detection to full braking is 0.18 seconds; effectively avoiding secondary injuries caused by the user's out-of-control situation.

[0103] In some embodiments, a GPS positioning module can be added to send the alarm location to medical staff; the braking method can be replaced with mechanical braking.

[0104] Embodiment 4, an adjustable rehabilitation cart for postoperative use, on the basis of Embodiment 1, the training method of the intention recognition model can also be:

[0105] Pre-collect k groups of historical limb feature data and the corresponding hydraulic pressure values. That is, simulate the limb feature data scenario in the experimental environment, and use different hydraulic pressure values as the scenarios of this limb feature data for rehabilitation training. Observe the training effect of the patient, and correspond the best hydraulic pressure value to this group of limb feature data scenarios. In this way, by simulating k groups of different limb feature data scenarios, k groups of historical limb feature data and the corresponding hydraulic pressure values can be obtained.

[0106] Use the historical limb feature data and the corresponding hydraulic pressure values as a sample set. The sample set is divided into a training set and a test set. Use the historical limb feature data in the training set as the input of the intention recognition model, and use the hydraulic pressure value in the training set as the output of the intention recognition model. Train the intention recognition model to output an intention recognition model that meets the preset accuracy. The intention recognition model is one of the naive Bayes model or the support vector machine model;

[0107] In this embodiment, the hydraulic pressure value is directly obtained through model training and directly used as the adjustment parameter of the oil pump 4, which is more convenient.

[0108] The exemplary embodiments of the solution proposed by the present disclosure have been described in detail above with reference to the preferred embodiments. However, those skilled in the art can understand that, without departing from the concept of the present disclosure, various modifications and variations can be made to the above specific embodiments, and various combinations of the technical features and structures proposed by the present disclosure can be made without exceeding the protection scope of the present disclosure. The protection scope of the present disclosure is determined by the appended claims.

Claims

1. An adjustable rehabilitation trolley for postoperative use, characterized in that, Comprising: A frame (1) with a roller frame (2) fixed to the bottom. A roller (21) is rotatably connected inside the roller frame (2); A resistance adjustment mechanism (3) for using hydraulic pressure to control the rotational resistance of the roller (21); An oil pump (4) for providing hydraulic pressure to the resistance adjustment mechanism (3) and controlling the hydraulic pressure value; A collection module for collecting user limb feature data; An analysis module for filtering and normalizing the original data of the sensor module to generate input feature vectors; An intention recognition module for judging the user's rehabilitation intention based on the input feature vectors through a trained intention recognition model. The rehabilitation intention includes active, fatigued, resistant, and abnormal postures; A control module for generating real-time rehabilitation force according to the output result of the intention recognition module, combining fuzzy control rules and a PID dynamic compensation algorithm.

2. The adjustable postoperative rehabilitation trolley according to claim 1, wherein, It further comprises: A safety control module for determining whether there is a risk based on limb feature data. If there is a risk, an alarm instruction is generated; An alarm module for generating an alarm based on the alarm instruction.

3. The adjustable rehabilitation trolley for postoperative use according to claim 2, wherein, The method for the safety control module to determine whether there is a risk includes: Presetting a safety threshold and a continuous time threshold, and continuously collecting i movement angle values for comparison with the preset safety threshold; If the continuous i movement angle values are greater than the safety threshold and the continuous time is greater than the continuous time threshold, it is determined that there is a risk of falling; If the continuous i movement angle values are greater than the safety threshold and the continuous time is less than the continuous time threshold, it is determined that there is no risk of falling; If the continuous i movement angle values are less than the safety threshold and the continuous time is greater than the continuous time threshold, it is determined that there is no risk of falling; If the continuous i movement angle values are less than the safety threshold and the continuous time is less than the continuous time threshold, it is not determined that there is no risk of falling.

4. The adjustable rehabilitation trolley for postoperative use according to claim 1, wherein, The limb feature data includes pressure value, muscle activity, movement angle, and angular velocity.

5. The adjustable rehabilitation trolley for postoperative use according to claim 1, wherein The method for collecting the muscle activity is: Presetting a sliding window; Collecting N electromyographic signals of the user's limbs in the sliding window, where N is an integer greater than 1, and converting them into muscle activity through root mean square calculation.

6. The adjustable rehabilitation cart for postoperative use according to claim 1, wherein, The training method of the intention recognition model is: The intention recognition module adopts a lightweight neural network classification model, and its input feature vectors are limb feature data; the output of the neural network is the probability distribution of intention labels, and the label corresponding to the maximum probability is selected as the intention determination result.

7. The adjustable rehabilitation cart for postoperative use according to claim 1, characterized in that, The fuzzy control rules of the control module include: When the user's intention is fatigued, the output force coefficient satisfies a preset force coefficient threshold one; When the user's intention is resistant, the output force coefficient satisfies a preset force coefficient threshold two; The force coefficient is calculated by weighted calculation of the membership function.

8. The adjustable rehabilitation cart for postoperative use according to claim 1, wherein, The frame (1) includes: A chassis (11) distributed on both sides of the first connecting rod (12), and the chassis (11) is slidably connected to the first connecting rod (12); A vertical rod (13) fixed to the surface of the chassis (11); A second connecting rod (14) with one end slidably connected to the vertical rod (13) and the other end slidably connected to a support rod (15); An armrest frame (16) fixed to the end of the support rod (15).

9. The adjustable rehabilitation trolley for postoperative use according to claim 1, wherein, The resistance adjustment mechanism (3) includes: The resistance disk (31) is coaxially connected and fixed to the roller (21); The disk clamp (32) is fixed to the roller frame (2); The liquid guide pipe (33) is fixed to the surface of the disk clamp (32) and is used to supply oil to the disk clamp (32); The resistance block (34) is slidably and sealingly connected to the disk clamp (32) and is used to abut against the resistance disk (31) to control the rotational frictional force of the resistance disk (31); One end of the reed (35) is fixed to the inner wall of the disk clamp (32), and the other end abuts against the resistance block (34).

10. The adjustable rehabilitation cart for postoperative use according to claim 7, wherein, A connecting block (321) is fixed to the outer surface of the disk clamp (32), and the connecting block (321) is fixedly connected to the disk clamp (32) by bolts.

Images