Upper limb full cycle rehabilitation training device for stroke patients

Abstract

The application relates to a full-cycle rehabilitation training device for the upper limbs of a stroke patient and belongs to the field of rehabilitation robots. The application aims to solve the problem of low rehabilitation efficiency of traditional artificial rehabilitation training and passive rehabilitation training of robots. An upper computer is used for controlling the movement of a mechanical arm on a table top, so as to drive the movement of a patient's affected limb; an electromyography collection device is used for collecting electromyography signals generated on the surface of a muscle skin of the patient in a spasm period, and transmitting the signals to the upper computer; the upper computer is also used for processing the received electromyography signals to obtain a motion target for adjusting the motion of the mechanical arm; a force sensor is used for collecting force information applied on a rehabilitation training handle by the patient's affected limb in real time and sending the force information to the upper computer; and the upper computer is also used for continuously adjusting the motion resistance of the mechanical arm according to the force information. Different rehabilitation training modes are set according to different conditions of the patient.

Description

Technical Field

[0001] This invention relates to active and passive rehabilitation training for the upper limbs and belongs to the field of rehabilitation robots. Background Technology

[0002] Currently, in clinical rehabilitation of upper limb motor disorders, therapists primarily use manual methods or devices to provide one-on-one passive rehabilitation training for patients' upper limbs. The increasing number of patients each year, limited by the number of therapists and the high cost of manual rehabilitation, places a significant burden and pressure on families and society. Furthermore, prolonged repetitive training can easily lead to physician fatigue, potentially causing training injuries or insufficient intensity. Therefore, there is a need to research a safe and efficient rehabilitation method to compensate for the shortcomings of traditional passive training. Based on the different conditions of hemiplegic patients in clinical medicine, the disease stages can be divided into flaccid paralysis, spasticity, and recovery. Numerous studies have shown that after damage to the central nervous system, movement can improve dendritic and axonal function, promote synaptic plasticity and neural regeneration, increase interhemispheric connections, promote neural compensation in non-infarcted areas, and facilitate brain functional reorganization—that is, movement-induced neural plasticity. Existing upper limb rehabilitation robots generally rely on force sensors to collect the patient's force information for single passive rehabilitation training, which has some effect on improving muscle strength, but its effect on neuronal regeneration and functional reorganization is very limited. Therefore, both traditional manual rehabilitation training and robotic passive rehabilitation training are passive training methods, which suffer from low rehabilitation efficiency. Summary of the Invention

[0003] The purpose of this invention is to address the problem that traditional manual rehabilitation training and robotic passive rehabilitation training are both passive training methods with low rehabilitation efficiency, and to propose a full-cycle upper limb rehabilitation training device for stroke patients.

[0004] A full-cycle rehabilitation training device for upper limbs for stroke patients, the device comprising a rehabilitation robot, an electromyography (EMG) acquisition device, a host computer, and a worktable;

[0005] Rehabilitation robots include robotic arms, force sensors, and rehabilitation training handles;

[0006] One end of the robotic arm is mounted on the table, and the other end of the robotic arm is connected to the rehabilitation training handle via a force sensor.

[0007] The host computer is used to control the movement of the robotic arm on the table, thereby moving the patient's affected limb.

[0008] The electromyography (EMG) acquisition device is used to collect the EMG signals generated on the skin surface of the muscle to be tested when the patient is in the spastic phase, and transmit them to the host computer.

[0009] The host computer is also used to process the received electromyographic signals to obtain the motion target for adjusting the movement of the robotic arm;

[0010] Force sensors are used to collect force information on the patient's affected limb applied to the rehabilitation training handle in real time and send it to the host computer.

[0011] The host computer is also used to continuously adjust the motion resistance of the robotic arm based on the force information.

[0012] Preferably, the robotic arm is controlled to move on the table, specifically as follows:

[0013] The motion trajectory of the robotic arm set by the host computer is analyzed using the DH parameter method or the Cartesian space trajectory planning method to obtain the rotation angle of each joint. The rotation angle of each joint is then applied to the driver of the corresponding joint of the robotic arm, thereby driving the motor on the corresponding joint to move, realizing the movement of the corresponding joint, and realizing the movement of the robotic arm on the table.

[0014] Preferably, the DH parameter method is used to analyze the motion trajectory of the robotic arm and obtain the rotation angles of each joint. The specific process is as follows:

[0015] First, the homogeneous transformation matrix for the rotation and translation of the i-th link relative to the (i-1)-th link is established using the DH parameter method.

[0016]

[0017] In the formula, i = 1, 2, 3, ..., 6; Rot is the rotation matrix, Trans is the translation matrix; Rot(x, α) i-1 ) represents the rotation of the coordinate system about the x-axis by α. i-1 Angle, making z i-1 axis and z i Axis parallel; Trans(x,a) i-1 ) indicates a translation along the x-axis. i-1 , make z i-1 The z-axis is collinear with the z-axis; Rot(z,θ) i ) indicates that the coordinate system rotates about the z-axis by θ. i Angle, making x i-1 axis and x i Axis parallel; Trans(z,d) i ) indicates a translation d along the z-axis i , making x i-1 axis and x i The axes are collinear; c represents cosine, s represents sinine; a i For along x i The z-axis is the distance between the z-axis of two adjacent link coordinate systems. i To circle x i The axis, the rotation angle between the z-axis of two adjacent link coordinate systems, d i For along z iThe x-axis is the distance between the x-axis of two adjacent link coordinate systems, θ. i For along z i The x-axis is the rotation angle between the x-axis of two adjacent link coordinate systems.

[0018] Substituting the DH parameter into Equation 1, we get Equation 1 as follows:

[0019]

[0020]

[0021]

[0022] Will and Multiplying these matrices yields the forward motion transformation matrix of the robotic arm (1-1).

[0023]

[0024] Perform an inverse transformation on Equation 2 to obtain the inverse transformation matrix. Using the principle that elements in the same row and column positions on both sides of the equation in the inverse transformation matrix are equal, establish a system of equations about θ1, θ2, θ3, θ4, θ5, and θ6, and solve the system of equations for θ1, θ2, θ3, θ4, θ5, and θ6.

[0025] Preferably, the host computer processes the received electromyographic signals, specifically as follows:

[0026] The electromyographic signals are preprocessed, feature extracted and action classified in sequence, and the obtained action targets are used to adjust the action of the robotic arm (1-1).

[0027] Preferably, the specific process of preprocessing is as follows:

[0028] The electromyographic signal is sequentially processed to remove power frequency interference noise, high frequency noise, and limb movement artifacts, resulting in a preprocessed signal.

[0029] Preferably, the specific process of feature extraction is as follows:

[0030] Three feature values ​​were extracted from the preprocessed signal using the time-domain method: integrated electromyography (IEMG) value, root mean square (RMS) value, and waveform length (WL).

[0031]

[0032]

[0033]

[0034] In the formula, x(t) is the preprocessed signal, t is the time sequence number of the sample point, t1 is the start time of the preprocessed signal, t2 is the end time of the preprocessed signal, and N is the number of sample points of the preprocessed signal in the window.

[0035] Preferably, the specific process of action classification is as follows:

[0036] The three feature values ​​are input into the trained BP neural network to classify actions and obtain the action target.

[0037] Preferably, the force sensor is a six-dimensional force sensor or a torque sensor.

[0038] Preferably, the motion resistance of the robotic arm (1-1) is continuously adjusted based on the force information, specifically as follows:

[0039] The error in the force information is eliminated by the compensation algorithm to obtain the compensated force. It is then determined whether the compensated force is greater than the preset threshold. If it is, the end of the robotic arm (1-1) is controlled to move slowly toward the direction of the force applied to the patient. If not, the robotic arm (1-1) is controlled to stop moving.

[0040] Preferably, the compensated force is obtained through the following process:

[0041] Establish a mathematical model for a six-dimensional force sensor:

[0042] V1″=A·F1″Formula 6,

[0043] In the formula, F1″ is a given 6-channel input matrix of a 6-dimensional force sensor. F i '=[F xi F yi F zi M xi M yi M zi ] T , i = 1, 2, ..., 6, F xi For F i The force F in the x-axis direction yi For F i The force F in the y-axis direction of ′ zi For F i The force M along the z-axis in the middle of the triangle is... xi For F i The torque about the x-axis, M yi For F i The torque M about the y-axis in ′ zi For F i The torque about the z-axis in the middle. V i '=[V 1i V2i V 3i V 4i V 5i V 6i ] T "V1" represents the given 6-channel output matrix of a 6-dimensional force sensor, T represents matrix transpose, and V 1i For V i The force in the x-axis direction, V 2i For V i The force in the y-axis direction, V 3i For V i The force V along the z-axis 4i For V i The torque V about the x-axis 5i For V i The torque about the y-axis, V 6i For V i The torque about the z-axis, where A is a 6×6 coefficient matrix.

[0044] get:

[0045] A = V1·F1 -1 Formula 7,

[0046] Let A represent the static calibration matrix C:

[0047] C = A -1 Formula 8,

[0048] The compensated force F2 is expressed as:

[0049] Formula 9: F2 = C·V2

[0050] In the formula, V2 represents the force information exerted by the patient's affected limb on the rehabilitation training handle, output by the force sensor.

[0051] The beneficial effects of this invention are:

[0052] The rehabilitation modes of the upper limb rehabilitation robot are set according to the different conditions of the patients, such as... Figure 1As shown. During the flaccid paralysis stage, the patient experiences low muscle strength and tone in the upper limbs, resulting in loss of motor function. Autonomous navigation based on robot trajectory planning guides passive training of the affected limb, using repetitive movements to help strengthen muscles and alleviate symptoms such as joint stiffness. During the spastic stage, muscle spasms cause difficulty in movement and increased muscle tone. Decoding the patient's upper limb movement intentions based on surface electromyography (EMG) signals provides assisted movement with the patient's conscious participation, forming active rehabilitation training. During the recovery phase, the patient's muscle strength and tone gradually return to normal levels, allowing for largely free movement. Force information from the affected limb is detected using a six-dimensional force / torque sensor, and the resistance is continuously adjusted through real-time calculated end-effector interaction, forming resistance rehabilitation training.

[0053] The principle of this application is to design different training programs based on the patient's upper limb condition. It aims to integrate passive and active rehabilitation training by combining robot trajectory planning, surface electromyography signal motion intent decoding, and force sensor analysis, all while ensuring safe interaction. This approach fully adapts to the characteristics of different patient conditions and offers higher rehabilitation efficiency compared to existing passive rehabilitation training methods. The advantage of this application lies in designing customized rehabilitation programs for stroke patients with different conditions, stimulating their enthusiasm for rehabilitation training, promoting in-situ regeneration and functional reorganization of motor neurons, thereby improving rehabilitation efficacy. Attached Figure Description

[0054] Figure 1 This is a full-cycle rehabilitation training model for the upper limbs;

[0055] Figure 2 It is a full-cycle rehabilitation training device for the upper limbs;

[0056] Figure 3 Let be the link coordinate system of the robotic arm;

[0057] Figure 4 A flowchart for passive rehabilitation training based on robot trajectory planning;

[0058] Figure 5 A flowchart of active rehabilitation training based on upper limb movement intention recognition using sEMG (surface electromyography) signals;

[0059] Figure 6 This is a flowchart of an active rehabilitation training program based on force sensors. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0062] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0063] Example 1:

[0064] Combination Figure 1 and Figure 2 This embodiment describes a full-cycle upper limb rehabilitation training device for stroke patients, comprising a rehabilitation robot 1, an electromyography (EMG) acquisition device 4, a host computer 5, and a worktable.

[0065] The rehabilitation robot 1 includes a robotic arm 1-1, a force sensor 1-2, and a rehabilitation training handle 1-3;

[0066] One end of the robotic arm 1-1 is mounted on the table, and the other end of the robotic arm 1-1 is connected to the rehabilitation training handle 1-3 through the force sensor 1-2.

[0067] The host computer 5 is used to control the movement of the robotic arm 1-1 on the table, thereby driving the patient's affected limb to move;

[0068] Electromyography (EMG) acquisition device 4 is used to acquire EMG signals generated on the skin surface of the muscle to be tested when the patient is in the spastic phase, and transmit them to the host computer 5.

[0069] The host computer 5 is also used to process the received electromyographic signals to obtain the motion target for adjusting the motion of the robotic arm 1-1;

[0070] Force sensors 1-2 are used to collect force information on the patient's affected limb applied to the rehabilitation training handle 1-3 in real time and send it to the host computer 5;

[0071] The host computer 5 is also used to continuously adjust the motion resistance of the robotic arm 1-1 based on the force information.

[0072] like Figure 2As shown, the electromyography (EMG) signal acquisition device includes a signal acquisition box, electrode wires, and electrode pads; the patient can perform two-dimensional movements on the platform by holding or binding the rehabilitation training handle with the robotic arm; the EMG signal acquisition device collects the sEMG signal data of the affected limb and transmits it to the host computer for processing and displays it in real time on the monitor 6.

[0073] This embodiment further defines the preferred method for controlling the movement of the robotic arm 1-1 on the table:

[0074] The DH parameter method is used to analyze the motion trajectory of the robotic arm 1-1 set by the host computer 5, obtain the rotation angle of each joint, and apply the rotation angle of each joint to the driver of the corresponding joint of the robotic arm 1-1, thereby driving the motor on the corresponding joint to move, realizing the movement of the corresponding joint, and realizing the movement of the robotic arm 1-1 on the table.

[0075] Combination Figure 3 This explains how the motion trajectory is analyzed to determine the rotation angles of each joint:

[0076] First, the homogeneous transformation matrix for the rotation and translation of the i-th link relative to the (i-1)-th link is established using the DH parameter method.

[0077]

[0078] In the formula, i = 1, 2, 3, ..., 6; Rot is the rotation matrix, Trans is the translation matrix; Rot(x, α) i-1 ) represents the rotation of the coordinate system about the x-axis by α. i-1 Angle, making z i-1 axis and z i Axis parallel; Trans(x,a) i-1 ) indicates a translation along the x-axis. i-1 , make z i-1 The z-axis is collinear with the z-axis; Rot(z,θ) i ) indicates that the coordinate system rotates about the z-axis by θ. i Angle, making x i-1 axis and x i Axis parallel; Trans(z,d) i ) indicates along z Axis translation d i , making x i-1 axis and x i The axes are collinear; c represents cosine, s represents sinine; a i For along x i The z-axis is the distance between the z-axis of two adjacent link coordinate systems. i To circle x i The axis, the rotation angle between the z-axis of two adjacent link coordinate systems, d i For along zi The x-axis is the distance between the x-axis of two adjacent link coordinate systems, θ. i For along z i The x-axis is the rotation angle between the x-axis of two adjacent link coordinate systems.

[0079] Substituting the DH parameter into Equation 1, we get Equation 1 as follows:

[0080]

[0081]

[0082]

[0083] Will and Multiplying these matrices yields the forward motion transformation matrix of the robotic arm (1-1).

[0084]

[0085] Perform an inverse transformation on Equation 2 to obtain the inverse transformation matrix. Using the principle that elements in the same row and column positions on both sides of the equation in the inverse transformation matrix are equal, establish a system of equations about θ1, θ2, θ3, θ4, θ5, and θ6, and solve the system of equations for θ1, θ2, θ3, θ4, θ5, and θ6.

[0086] Figure 4 In the task decision, various information such as the starting and ending velocities, accelerations, and motion paths of the robotic arm are specified. Several time points are interpolated in Cartesian space for the motion path. At each time point, the kinematic model of the robotic arm is established using the DH parameter method. The model is inverted to obtain the joint angles. The joint angles and the starting and ending velocities and accelerations of the robotic arm are sent to the host computer to drive the motors on the corresponding joints. This ensures that the robotic arm has six corresponding joint angles at each time point in the Cartesian plan, thus enabling the robotic arm to complete continuous movements.

[0087] Formula 2 can also be transformed into various inverse transformation matrices, each yielding a set of θ1, θ2, θ3, θ4, θ5, and θ6. Below is an example of obtaining θ1, θ2, θ3, θ4, θ5, and θ6 using one inverse transformation matrix:

[0088] If the inverse transformation matrix is: Formula 10, will and After substituting the inverse transformation matrix, both sides of the equals sign are 4x4 matrices. Setting the elements in the second row and fourth column of both sides to be equal, we get:

[0089] -p x si +p y c i =d2+d3 formula 11,

[0090] In the formula, p x =d5c1s 234 +d4s1+d6(c5s1-s5c1c 234 )+a2c1c2+a3c1c 234 p y =d5s1s 234 -d4c1-d6(c1c5+s5s1c 234 )+a2c2s1+a3s1c 23 c i =cosθ,s i =sinθ i s ijk =sin(θ) i +θ j +θ k ), c ijk =cos(θ) i +θ j +θ k ),

[0091] For p x and p y Performing trigonometric transformations respectively, we obtain:

[0092] Formula 12, where, φ=arctan(p y ,p x ),

[0093] Substituting formula 12 into formula 11, we get... In the formula,

[0094] Similarly, using the elements with θ2, θ3, θ4, θ5, and θ6 on both sides of the equal sign in Formula 10, a system of equations can be constructed to deduce θ2, θ3, θ4, θ5, and θ6. During passive training, after the host computer issues a motion command for the robotic arm's end effector, it can obtain multiple combinations of six joint angles based on the calculations in Formulas 1 and 2. All of these combinations can move the robotic arm's end effector to the same pose. Therefore, the host computer selects the optimal combination of joint angles from these combinations to ensure the robotic arm's end effector moves to that pose at the fastest speed.

[0095] This embodiment further defines the preferred method for the host computer to process the received electromyographic signals:

[0096] The electromyographic signals are preprocessed, feature extracted and action classified in sequence, and the obtained action targets are used to adjust the action of the robotic arm (1-1).

[0097] During the spastic phase, an active rehabilitation training model based on surface electromyography (sEMG) signal motor intention recognition was designed.

[0098] First, the patient's biceps, triceps, posterior deltoid, and pectoralis major muscles were wiped with alcohol to remove surface oil and dead skin cells. Electrode pads were then attached to the skin surface of the muscles to be tested, and the electrode wires were connected to the electromyography (EMG) acquisition device and the electrode pads. Because sEMG signals are weak, they are highly susceptible to environmental noise interference, mainly including noise from impurities on the subject's skin surface, power frequency interference noise, and motion artifact noise. Preprocessing of the raw signal begins with filtering power frequency interference using a notch filter to remove interference while preserving sEMG information to the greatest extent possible. A Butterworth filter was then used for 10-150Hz bandpass filtering to remove high-frequency noise and limb motion artifacts, with the following basic parameters set (stopband frequency: 49-51Hz, stopband attenuation: rs = 50dB, passband attenuation: rp = 1dB). Next, an overlapping sliding window method is used to process the signal. The average absolute value of the window is represented by the energy value, and a threshold is set to determine the start and end of the action. If the energy value is greater than the threshold more than N times, the action is considered to have started; otherwise, the action is considered to have ended. To ensure the accuracy of feature extraction and the real-time performance of the system, the window size is set to 100ms, the sliding step size is 50ms, and the threshold is set to 1.5 times the energy value.

[0099] This embodiment further defines the preferred method of preprocessing:

[0100] The electromyographic signals were sequentially processed to remove power frequency interference noise, high frequency noise, and limb movement artifacts, resulting in preprocessed signals.

[0101] This embodiment further defines the preferred method for feature extraction:

[0102] Three feature values ​​were extracted from the preprocessed signal using the time-domain method: integrated electromyography (IEMG) value, root mean square (RMS) value, and waveform length (WL).

[0103]

[0104]

[0105]

[0106] In the formula, x(t) is the preprocessed signal, t is the time sequence number of the sample point, t1 is the start time of the preprocessed signal, t2 is the end time of the preprocessed signal, and N is the number of sample points of the preprocessed signal in the window.

[0107] This embodiment further defines the preferred method for action classification:

[0108] The three feature values ​​are input into the trained BP neural network to classify actions and obtain the action target.

[0109] The multidimensional vector formed by these three feature values ​​is used as the input to the classifier, and a backpropagation (BP) neural network is chosen as the classifier. The BP neural network classifier has many advantages; it can adaptively learn and adjust, thus adapting to different classification problems and data types. Furthermore, the BP neural network classifier can handle nonlinear problems and has high classification accuracy, especially when dealing with large amounts of data. It performs well in classifying non-stationary signals and can be used for action classification of surface electromyography (EMG) signals. This method also has strong fault tolerance and robustness; even if the data contains a certain degree of noise or interference, the classifier's performance will not be significantly affected. Thus, as... Figure 5 As shown, the active rehabilitation training design based on surface electromyography signals has been completed.

[0110] This embodiment further specifies that force sensors 1-2 are six-dimensional force sensors or torque sensors.

[0111] This embodiment further specifies an optimal method for processing the force information output by the force sensor to obtain the true force, and for continuously adjusting the motion resistance of the robotic arm (1-1) based on the true force:

[0112] The error in the force information is eliminated by the compensation algorithm to obtain the compensated force. It is then determined whether the compensated force is greater than the preset threshold. If it is, the end of the robotic arm (1-1) is controlled to move slowly toward the direction of the force applied to the patient. If not, the robotic arm (1-1) is controlled to stop moving.

[0113] During the recovery period, an active rehabilitation training mode based on a six-dimensional force / torque sensor was designed. To more accurately acquire the human-machine interaction force at the end effector, a compensation algorithm was used to eliminate external interference such as the sensor's own system errors and the influence of gravity on the end effector handle, thereby reducing the impact on the experiment. The force sensor transmits the collected raw data to the host computer in real time via UDP protocol. After the host computer performs number conversion and external force compensation, the compensated force is obtained. To ensure the robot's compliance, a preset threshold judgment time of 1 second was set. The robot will only execute the corresponding command if the applied force remains in a certain state for more than 1 second. Thus, as... Figure 6As shown, the active rehabilitation training mode based on a six-dimensional force sensor has been designed. This embodiment is equivalent to weight-bearing training. The end effector of the robotic arm monitors the magnitude of the force applied by the patient, and then sets an appropriate resistance according to this force, allowing the patient to carry out weight-bearing training within a safe range.

[0114] This embodiment further limits the error in removing force information and obtains the preferred method for compensating the force:

[0115] The calibration matrix of the six-dimensional force sensor 1-2 is divided into a static calibration matrix and a dynamic calibration matrix. Since the movement of the robotic arm 1-1 is uniform and slow, it is considered static. Therefore, during the static calibration of the six-dimensional force sensor 1-2, the robotic arm 1-1 is moved to six different postures and then stopped, with the postures not collinear. The 6×6 coefficient matrix A is solved, and the force information of the patient's affected limb applied to the rehabilitation training handle is further obtained from the force sensor output. The specific process is as follows:

[0116] Establish a mathematical model for a six-dimensional force sensor:

[0117] V1″=A·F1″ Formula 6,

[0118] In the formula, F1″ is a given 6-channel input matrix of a 6-dimensional force sensor. F i '=[F xi F yi F zi M xi M yi M zi ] T , i = 1, 2, ..., 6, F xi For F i The force F in the x-axis direction yi For F i The force F in the y-axis direction of ′ zi For F i The force M along the z-axis in the middle of the triangle is... xi For F i The torque about the x-axis, M yi For F i The torque M about the y-axis in ′ zi For F i The torque about the z-axis in the middle. V i '=[V 1i V 2i V 3i V 4i V 5i V 6i ] T "V1" represents the given 6-channel output matrix of a 6-dimensional force sensor, T represents matrix transpose, and V 1i For Vi The force in the x-axis direction, V 2i For V i The force in the y-axis direction, V 3i For V i The force V along the z-axis 4i For V i The torque V about the x-axis 5i For V i The torque about the y-axis, V 6i For V i The torque about the z-axis, where A is a 6×6 coefficient matrix.

[0119] Since the six poses are not collinear, meaning F1 is an invertible matrix, there exists a unique solution:

[0120] A = V1·F1 -1 Formula 7,

[0121] Let A represent the static calibration matrix C:

[0122] C = A -1 Formula 8,

[0123] The compensated force F2 is expressed as:

[0124] Formula 9: F2 = C·V2

[0125] In the formula, V2 represents the force information exerted by the patient's affected limb on the rehabilitation training handle, output by the force sensor.

[0126] The calibration matrix of a six-dimensional force sensor is divided into a static calibration matrix and a dynamic calibration matrix. Static calibration means applying a load to the six-dimensional force sensor or torque sensor under static or low-speed conditions to perform calculations. Since the rehabilitation movement process of the upper limb rehabilitation robot is uniform and slow, it can be regarded as static, so this embodiment uses static calibration.

[0127] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A full-cycle rehabilitation training device for the upper limbs of stroke patients, characterized in that, The device includes a rehabilitation robot (1), an electromyography (EMG) acquisition device (4), a host computer (5), and a worktable; The rehabilitation robot (1) includes a robotic arm (1-1), a force sensor (1-2), and a rehabilitation training handle (1-3); One end of the robotic arm (1-1) is mounted on the table, and the other end of the robotic arm (1-1) is connected to the rehabilitation training handle (1-3) through a force sensor (1-2); The host computer (5) is used to control the movement of the robotic arm (1-1) on the table, thereby driving the patient's affected limb to move; The electromyography acquisition device (4) is used to collect electromyographic signals generated on the skin surface of the muscle to be tested when the patient is in the spastic phase and transmit them to the host computer (5). The host computer (5) is also used to process the received electromyographic signals to obtain the motion target for adjusting the motion of the robotic arm (1-1); Force sensors (1-2) are used to collect force information on the patient's affected limb applied to the rehabilitation training handle (1-3) in real time and send it to the host computer (5); The host computer (5) is also used to continuously adjust the motion resistance of the robotic arm (1-1) based on the force information. The force sensors (1-2) are six-dimensional force sensors or torque sensors; The error in the force information is eliminated by a compensation algorithm to obtain the compensated force; The specific process of obtaining the compensated force is as follows: Establish a mathematical model for a six-dimensional force sensor: V1”=AF1 formula 6, In the formula, F1″ is a given 6-channel input matrix of a 6-dimensional force sensor. F i '=[F xi F yi F zi M xi M yi M zi ] T , i = 1, 2, ..., 6, F xi For F i The force F in the x-axis direction yi For F i The force F in the y-axis direction of ′ zi For F i The force M along the z-axis in the middle of the triangle is... xi For F i The torque M about the x-axis in ′ yi For F i The torque M about the y-axis in ′ zi For F i The torque about the z-axis in the middle. V i '=[V 1i V 2i V 3i V 4i V 5i V 6i ] T "V1" represents the given 6-channel output matrix of a 6-dimensional force sensor, T represents matrix transpose, and V 1i For V i The force in the x-axis direction, V 2i For V i The force in the y-axis direction, V 3i For V i The force V along the z-axis 4i For V i The torque V about the x-axis 5i For V i The torque about the y-axis, V 6i For V i The torque about the z-axis, where A is a 6×6 coefficient matrix. get: A = V1 F1 -1 Formula 7, Let A represent the static calibration matrix C: C = A -1 Formula 8, The compensated force F2 is expressed as: F2=CV2 formula 9, In the formula, V2 represents the force information exerted by the patient's affected limb on the rehabilitation training handle, output by the force sensor.

2. The upper limb full-cycle rehabilitation training device for stroke patients according to claim 1, characterized in that, Control the movement of the robotic arm (1-1) on the table, specifically as follows: The DH parameter method is used to analyze the motion trajectory of the robotic arm (1-1) set by the host computer (5), obtain the rotation angle of each joint, and apply the rotation angle of each joint to the driver of the corresponding joint of the robotic arm (1-1), thereby driving the motor on the corresponding joint to move, realizing the movement of the corresponding joint, and realizing the movement of the robotic arm (1-1) on the table.

3. The upper limb full-cycle rehabilitation training device for stroke patients according to claim 2, characterized in that, The motion trajectory of the robotic arm (1-1) is analyzed using the DH parameter method to obtain the rotation angles of each joint. The specific process is as follows: First, the homogeneous transformation matrix T for the rotation and translation of the i-th link relative to the (i-1)-th link is established using the DH parameter method. i i-1 : In the formula, i = 1, 2, 3, ..., 6; Rot is the rotation matrix, Trans is the translation matrix; Rot(x, α) i-1 ) represents the rotation of the coordinate system about the x-axis by α. i-1 Angle, making z i-1 axis and z i Axis parallel; Trans(x,a) i-1 ) indicates a translation along the x-axis. i-1 , make z i-1 The z-axis is collinear with the z-axis; Rot(z,θ) i ) indicates that the coordinate system rotates about the z-axis by θ. i Angle, making x i-1 axis and x i Axis parallel; Trans(z,d) i ) indicates a translation d along the z-axis i , making x i-1 axis and x i The axes are collinear; c represents cosine, s represents sinine; a i For along x i The z-axis is the distance between the z-axis of two adjacent link coordinate systems. i To circle x i The axis, the rotation angle between the z-axis of two adjacent link coordinate systems, d i For along z i The x-axis is the distance between the x-axis of two adjacent link coordinate systems, θ. i For along z i The x-axis is the rotation angle between the x-axis of two adjacent link coordinate systems. Substituting the DH parameter into Equation 1, we get Equation 1 as follows: T1 0 , and Multiplying these matrices yields the forward motion transformation matrix of the robotic arm (1-1). Perform an inverse transformation on Equation 2 to obtain the inverse transformation matrix. Using the principle that elements in the same row and column positions on both sides of the equation in the inverse transformation matrix are equal, establish a system of equations about θ1, θ2, θ3, θ4, θ5, and θ6, and solve the system of equations for θ1, θ2, θ3, θ4, θ5, and θ6.

4. The upper limb full-cycle rehabilitation training device for stroke patients according to claim 1, characterized in that, The host computer (5) processes the received electromyographic signals, specifically as follows: The electromyographic signals are preprocessed, feature extracted and action classified in sequence, and the obtained action targets are used to adjust the action of the robotic arm (1-1).

5. The upper limb full-cycle rehabilitation training device for stroke patients according to claim 4, characterized in that, The specific process of preprocessing is as follows: The electromyographic signal is sequentially processed to remove power frequency interference noise, high frequency noise, and limb movement artifacts, resulting in a preprocessed signal.

6. The upper limb full-cycle rehabilitation training device for stroke patients according to claim 5, characterized in that, The specific process of feature extraction is as follows: Three feature values ​​were extracted from the preprocessed signal using the time-domain method: integrated electromyography (IEMG) value, root mean square (RMS) value, and waveform length (WL). In the formula, x(t) is the preprocessed signal, t is the time sequence number of the sample point, t1 is the start time of the preprocessed signal, t2 is the end time of the preprocessed signal, and N is the number of sample points of the preprocessed signal in the window.

7. The upper limb full-cycle rehabilitation training device for stroke patients according to claim 6, characterized in that, The specific process of action classification is as follows: The three feature values ​​are input into the trained BP neural network to classify actions and obtain the action target.

8. The upper limb full-cycle rehabilitation training device for stroke patients according to claim 1, characterized in that, Based on the force information, the motion resistance of the robotic arm (1-1) is continuously adjusted. The specific process is as follows: Determine whether the compensated force is greater than the preset threshold. If it is, control the end of the robotic arm (1-1) to move slowly toward the direction of the force applied to the patient. If not, control the robotic arm (1-1) to stop moving.

Images