Personalized training scheme generation method and system, rehabilitation training method and device and storage medium

Abstract

The invention provides a personalized training scheme generation method and system, a rehabilitation training method and device and a storage medium, and relates to the field of medical instruments.The personalized training scheme generation method comprises the steps that a scheme generation model is pre-constructed based on multiple algorithms, after evaluation information of a user is obtained, the constructed scheme generation model is adopted to generate a personalized training scheme, and the user experience is improved. Targeted training guidance is provided for the user, and a better rehabilitation effect is achieved; meanwhile, the rehabilitation robot is automatically switched to the corresponding rehabilitation assembly according to the training scheme in the training process, the rehabilitation assembly is driven to the target position through the mechanical arm, the user is guided to conduct various rehabilitation trainings, and the user can use the rehabilitation robot conveniently.

Description

Technical Field

[0001] The present invention relates to the technical field of medical devices, and in particular, to a method and system for generating a personalized training plan, a rehabilitation training method, a device, and a storage medium. Background Art

[0002] Dysphagia is a clinical symptom in which food cannot smoothly pass from the mouth to the stomach due to various reasons. Dysphagia can not only affect the normal food intake of patients, causing general malnutrition in patients, but also may cause choking and aspiration in patients, thus triggering accidents such as lung infections in patients, and even threatening the life safety of patients.

[0003] Most of the therapies in traditional technologies require manual operation, and the physical therapy methods for dysphagia are relatively single, unable to provide personalized training plans for patients based on biofeedback, and unable to conveniently and effectively treat dysphagia. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for generating a personalized training plan, a rehabilitation training method, a device, and a storage medium, so as to overcome the defects that most of the therapies in traditional technologies require manual operation, and the physical therapy methods for dysphagia are relatively single, unable to provide personalized training plans for patients based on biofeedback, and unable to conveniently and effectively treat dysphagia.

[0005] In a first aspect, the present application proposes a method for generating a personalized training plan, which is applicable to a rehabilitation robot, including: Obtaining evaluation information of a user; the evaluation information includes medical evaluation parameters and physiological ability parameters, the medical evaluation parameters include dysphagia level and complication taboos, and the physiological ability parameters include a pressure reference value, a tongue pressure ability value, and an electromyogram reference value; Processing the evaluation information by using a pre-constructed plan generation model to generate a personalized training plan, so that the rehabilitation robot automatically performs rehabilitation training on the target object of the user according to the training plan.

[0006] In one embodiment, the processing the evaluation information by using a pre-constructed plan generation model to generate a personalized training plan includes: Constructing the plan generation model based on fuzzy logic and a decision tree; Performing fuzzy processing on the evaluation information based on the constructed plan generation model to obtain a plurality of fuzzy variables; and generating a personalized training plan according to the fuzzy variables by using a decision tree algorithm.

[0007] In one embodiment, the plan generation model generating a personalized training plan according to the fuzzy variables by using a decision tree algorithm includes: Construct the fuzzy membership function corresponding to the fuzzy variable; Based on the preset fuzzy rules, comprehensively evaluate the fuzzy membership functions of the fuzzy variables to obtain a preliminary fuzzy output result; Use the preliminary fuzzy result as the input of the decision tree, and determine the training data through branch judgment; wherein, the training data includes first-class data and second-class data, the first-class data includes training modes and training items, and the second-class data includes training intensity, training duration, and progressive step size; The above-mentioned training data constitutes the training plan.

[0008] In one embodiment, using the pre-constructed plan generation model to process the evaluation information to generate a personalized training plan includes: Construct the plan generation model based on the large language model; Based on the constructed plan generation model, process the evaluation information to obtain a personalized training plan; wherein, the large language model includes a base model, a tuning model, a vertical model, a distilled inference model, and an agent created by the above models.

[0009] In one embodiment, the method further includes: Obtain the training result; Analyze the training result and adjust the training plan according to the analysis result; wherein, the training result includes training parameters and swallowing indicators; the training result is obtained by the rehabilitation robot guiding the user to perform rehabilitation training according to the training plan.

[0010] In one embodiment, analyzing the training result and adjusting the training plan according to the analysis result includes: Compare the training parameters and the swallowing indicators with preset target values respectively to obtain difference information. If the difference information exceeds the preset range, adjust the training plan; Wherein, adjusting the training plan includes adjusting the content and weight of the first-class data, and adjusting the threshold range of the second-class data.

[0011] In one embodiment, analyzing the training result and adjusting the training plan according to the analysis result further includes: Periodically obtain the training result to obtain a training data set; Generate a development curve or heat map of swallowing ability according to the training data set; Predict the training effect according to the development curve or heat map, and adjust the training plan according to the training effect.

[0012] Second aspect, the present application proposes a system for generating a personalized training plan, the system comprising: An acquisition module, configured to acquire evaluation information of a user; the evaluation information includes medical evaluation parameters and physiological ability parameters, the medical evaluation parameters include a dysphagia level and complication taboos, and the physiological ability parameters include a pressure reference value, a tongue pressure ability value, and an electromyogram reference value; A plan generation module, configured to process the evaluation information by using a pre-constructed plan generation model to generate a personalized training plan, so that the rehabilitation robot automatically performs rehabilitation training on a target object of the user according to the training plan. Third aspect, the present application further provides a rehabilitation training method, applicable to a rehabilitation robot, the rehabilitation robot comprising a support member, a robotic arm, a rehabilitation component, and a first sensor component disposed on the support member; the method comprises: According to a detection signal acquired by the first sensor component, confirming whether the support member reaches a target position; the detection signal includes an electromyogram value; When the support member reaches the target position, executing a training plan, so that the robotic arm drives the rehabilitation component to move to the target position along a preset trajectory to guide the user to perform rehabilitation training on a target object; wherein, the training plan is pre-generated according to the personalized training plan generation method according to any one of claims 1-7.

[0013] In one embodiment, the detection signal further includes a pressure value; The confirming whether the support member reaches the target position according to the detection signal acquired by the first sensor component includes: Calculating a pressure distribution of the user's lower jaw on the support member according to the pressure value; When the pressure distribution and the electromyogram value reach a preset reference threshold, it is considered that the support member reaches the target position. In one embodiment, the rehabilitation robot includes an interaction module, a storage module, and a communication module; training information is stored in the storage module or a cloud server; the training information includes a training plan, evaluation information, training data, and training results; The user is communicatively connected to the communication module through a terminal, and views the training information in the storage module or the cloud server through the interaction module.

[0014] In one embodiment, the rehabilitation component includes a tongue suction module, a tongue pressure resistance module, and an air pulse module; The guiding the user to perform rehabilitation training on a target object includes: The tongue suction module fixes the user's tongue and pulls the tongue to perform rehabilitation exercises through the movement of the mechanical arm; Or, the tongue pressure resistance module provides a plurality of detection points at different positions for the tongue to contact; Or, the air pulse module blows air to a designated position in the user's mouth; Alternatively, the user is guided to place the lower jaw on the supporting member and press down in different directions.

[0015] In one embodiment, the detection signal further includes a pressure value; and the method further includes: Calculating the pressure distribution of the user's lower jaw on the supporting member according to the pressure value; Whether the user's jaw is positioned correctly is determined based on the pressure distribution.

[0016] In one of the embodiments, the rehabilitation robot further includes a photobiomodulation module and a hyperoxia and hypoxia treatment module; The guiding the user to perform rehabilitation training on the target object also includes: The training is performed using any module and / or auxiliary module of the rehabilitation component; wherein the auxiliary module is any one or more of the photobiomodulation module and the hyperoxia and hypoxia therapy module. In one embodiment, the preset trajectory is calculated based on the starting position and the target position of the robot arm, and the step of calculating the preset trajectory includes: Configuring a coordinate system for each joint of the robotic arm and determining motion parameters of each joint, wherein the motion parameters include joint length, joint torsion angle, joint offset and joint angle; According to the motion parameters, construct a homogeneous transformation matrix of each joint, and multiply the homogeneous transformation matrices in sequence to obtain a total transformation matrix, wherein the total transformation matrix is ​​used to represent the position and posture of the target coordinate system relative to the starting coordinate system; According to the total transformation matrix, the mapping relationship between the motion parameters and the posture is determined through forward kinematics and inverse kinematics; Based on the mapping relationship, an interpolation algorithm is used to generate a series of intermediate postures between the starting posture and the target posture; The starting posture, the target posture and each intermediate posture constitute the preset trajectory.

[0017] In one embodiment, the method further comprises: Quintic spline interpolation is used to calculate a variation curve of the joint angle of each joint; the variation curve is used to achieve smooth acceleration or deceleration of the robotic arm.

[0018] In one of the embodiments, a second sensor assembly is provided on the mechanical arm; The second sensor assembly is configured to obtain protection data of the robotic arm, where the protection data includes force parameters, rotation angle parameters of each joint on the robotic arm, and current parameters of the motor driving the robotic arm; Execute a protection strategy based on the protection data.

[0019] In one embodiment, the executing a protection strategy based on the protection data includes: When the force parameter is greater than a preset force threshold, the robotic arm retracts to the starting position; When the rotation angle parameter is greater than a preset rotation threshold, the joint returns to the initial angle; When the current parameter is greater than a preset current threshold, enter a stall protection mode to reduce or stop the current supply.

[0020] In a fourth aspect, the present application also provides a computer device. The computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the method steps of the rehabilitation training method in the first aspect are implemented.

[0021] In a fifth aspect, the present application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, the method steps of the rehabilitation training method in the first aspect are implemented.

[0022] The above-mentioned method for generating a personalized training plan, system, rehabilitation training method, device, and storage medium have at least the following advantages: The present application pre-constructs a plan generation model and obtains the evaluation information of the user. The evaluation information includes medical evaluation parameters and physiological ability parameters. The medical evaluation parameters include the swallowing disorder level and complication taboos, and the physiological ability parameters include the pressure reference value, tongue pressure ability value, and electromyogram reference value. After obtaining the above evaluation information, the constructed plan generation model is used to generate a personalized training plan for the user, providing targeted training guidance for the user to achieve a better rehabilitation effect; at the same time, the rehabilitation robot in the present application automatically switches to the corresponding rehabilitation component according to the training plan during the training process, and drives the rehabilitation component to the target position by the robotic arm to guide the user to perform various rehabilitation trainings, which is convenient for the user to use. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is an application environment diagram of the rehabilitation training method in an embodiment; Figure 2 It is a schematic structural diagram of a rehabilitation robot in an embodiment; Figure 3Flow diagram of the method for generating a personalized training plan in an embodiment; Figure 4 Flow diagram of the steps for generating a training plan in an embodiment; Figure 5 Flow diagram of the rehabilitation training method in an embodiment; Figure 6 Flow diagram of the steps for calculating a preset trajectory in an embodiment; Figure 7 Structural block diagram of the system for generating a personalized training plan in an embodiment; Figure 8 Structural block diagram of the rehabilitation training system in an embodiment; Figure 9 Internal structure diagram of a computer device in an embodiment.

[0024] Reference numerals: 100, support main body; 110, supporting member; 140, positioning member; 300, robotic arm; 340, turntable; 400, rehabilitation component; 500, display screen. Detailed implementation manners

[0025] The following uses specific specific examples to illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0026] For the purpose of illustration, some exemplary embodiments of the present invention are described. It should be understood that the present invention can be implemented in other ways not specifically shown in the drawings.

[0027] Please refer to Figure 1 , the rehabilitation training method provided by the embodiments of the present application can be applied to an application environment as Figure 1 shown. Among them, the terminal 102 communicates with the rehabilitation robot 104 through the network, so that the user can view the training information involved during training on the terminal 102. The training information includes training plans, evaluation information, training data, and training results. Exemplarily, an application program (app) associated with the rehabilitation robot is installed on the terminal 102, and the user can view the training information in the app.

[0028] Further, the above training information can also be stored in the cloud server 106. At this time, the rehabilitation robot 104 also communicates with the cloud server 106 through the network, and the user views the training information stored in the cloud server 106 through the terminal 102.

[0029] Among them, the terminal 102 can be but is not limited to various personal computers, laptop computers, smart phones, tablet computers, and portable wearable devices. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, etc. The user can select any one of the above terminals 102 to be wired or wirelessly connected to the rehabilitation robot 104.

[0030] Please refer to Figure 2 , in one embodiment, a rehabilitation robot is provided. The rehabilitation robot includes: a support main body 100, a supporting member 110, a robotic arm 300, a rehabilitation component 400, and a first sensor component.

[0031] The support main body 100 is used to support the supporting member 110 and drive the supporting member 110 to move along a set trajectory. Exemplarily, in the embodiment of the present application, a lifting platform 700 is provided on the support main body 100, and a motor capable of driving the lifting platform 700 to rise or fall is configured on the lifting platform 700. Thus, under the drive of the motor, the lifting platform 700 can drive the supporting member 110 to move up and down. In another embodiment, a horizontal track can also be provided on the support main body 100, so that the support main body 100 can move back and forth, left and right in the horizontal direction, so that the supporting member 110 can better adapt to the user's mandible.

[0032] The supporting member 110 is used to support the user's mandible.

[0033] The first sensor component is disposed on the supporting member 110 and is used to obtain a detection signal, where the detection signal includes a pressure value and an electromyogram value.

[0034] Optionally, the first sensor component includes a pressure sensor and an sEMG sensor.

[0035] The pressure sensor is disposed on the surface of the supporting member 110 and at the contact position between the supporting member 110 and the mandible, and is used to detect the pressure value of the mandible on the supporting member 110. The number of pressure sensors can be multiple, and each pressure sensor can be dispersedly disposed around the above contact position. Thus, by the pressure values of each pressure sensor, the pressure distribution of the mandible on the supporting member 110 is calculated, and the support data can be obtained more accurately, so as to determine the training state of the user.

[0036] Exemplarily, the pressure sensor includes a front sensor, a middle sensor, a rear sensor, a left sensor, and a right sensor. In the Shaker training method (tremor training method) or the mandibular retrusion resistance training, after the user's mandible is placed on the support member 110, the user presses down in different directions through the support member 110. For example, when the user presses down vertically, the front, middle, and rear sensors can display the pressure values in real time. According to the above pressure values, the pressure distribution is calculated to determine whether the pressing force reaches the set value and whether the pressing duration reaches the set value. The user can also press down along the lower left or lower right diagonal line, that is, the user can press down obliquely. The left sensor and the right sensor are respectively used as real-time biofeedback to confirm whether the user's mandible is placed correctly and whether there is displacement.

[0037] The sEMG sensor is disposed on the surface of the support member 110 and at the contact position between the support member 110 and the swallowing muscle, and is used to detect the myoelectric value of the swallowing muscle. It should be understood that in order to capture richer muscle activity information and obtain more comprehensive signal characteristics, a multi-channel sEMG sensor can be selected.

[0038] The robotic arm 300 is used to drive the rehabilitation component 400 to move to the target position along a preset trajectory after receiving a start instruction. It should be understood that the robotic arm 300 includes multiple joints, and each joint can move independently. By reasonably planning the movement of each joint axis, the joints can move in coordination and move to the target position along a preset trajectory.

[0039] Optionally, when the number of the rehabilitation components 400 is multiple, the rehabilitation robot of the present application further includes: a turntable 340.

[0040] The turntable 340 is connected to the end effector of the robotic arm 300, and each rehabilitation component 400 is disposed on the turntable 340. By rotating the turntable 340, the switching of different rehabilitation components 400 can be realized. Further, the rotation of the turntable 340 is driven by a servo motor.

[0041] Optionally, the rehabilitation robot further includes: an interaction module.

[0042] The interaction module is used to guide the user to perform rehabilitation training and to allow the user to view all the data involved in the training process.

[0043] The above-mentioned rehabilitation robot automatically executes a training program in response to a user's start instruction, drives the turntable 340 to switch to the corresponding rehabilitation component 400, and drives the robotic arm 300 to drive the rehabilitation component 400 to move to a target position along a preset trajectory. In this way, the user only needs simple training to use the rehabilitation component 400 to perform self-help rehabilitation training on the target object, greatly improving the training efficiency. It should be noted that the target object in the following embodiments is mainly described with the tongue muscle. Further, it can also be other parts associated with the tongue muscle, such as the neck.

[0044] Optionally, the rehabilitation component 400 includes: a tongue suction module, a tongue pressure resistance module, and an air pulse module.

[0045] The tongue suction module is used to fix the user's tongue and, through the movement of the robotic arm 300, traction the tongue to perform rehabilitation exercises. Exemplarily, the user places the mandible on the support member 110. After the support member 110 rises to an appropriate height, the turntable 340 rotates to the tongue suction module. The robotic arm 300 drives the tongue suction module to slowly and smoothly move to the target position corresponding to this height. The user extends the tongue, and the tongue suction module generates a vacuum suction force to adsorb the user's tongue. Then, depending on the training program, the robotic arm 300 pulls the tongue muscle in different directions at a set frequency. After the training is completed, the suction force disappears, the robotic arm 300 returns to the starting position, and the interaction module prompts the user to perform a swallowing action. The sEMG sensor measures the myoelectric value of this swallowing action, and the training result includes the myoelectric value at this time.

[0046] The tongue pressure resistance module is used to provide detection points at multiple different positions for the tongue to contact. Further, at least one first sensor is provided at each detection point for detecting the pressure and duration applied by the tongue when the user's tongue contacts the detection point. The tongue pressure resistance module further includes a second sensor for detecting the position of the tongue pressure resistance module in the user's oral cavity. Exemplarily, both the above-mentioned first sensor and the second sensor adopt flexible pressure sensors.

[0047] For further explanation, during the execution of tongue pressure resistance feedback training, the user places the mandible on the support member 110. After the support member 110 rises to an appropriate height, the turntable 340 rotates to the tongue pressure resistance module, and the robotic arm 300 drives the tongue pressure resistance module to move slowly and smoothly to the target position corresponding to this height and extends into the user's oral cavity. After staying for a preset time, the robotic arm 300 drives the tongue pressure resistance module to move vertically upward so that the second sensor presses against the user's upper jaw. When the detection value of the second sensor reaches the preset threshold, it is considered to have reached the designated position. At this time, the interaction module guides the user to perform isometric exercise training of the tongue muscles. The user presses the tongue against different detection points in turn, and the first sensor measures the applied pressure and duration. After the training ends, the robotic arm 300 returns to the starting position, the interaction module prompts the user to perform a swallowing action, and the sEMG sensor measures the myoelectric value of this swallowing action. The training results include the myoelectric value at this time, as well as the above-mentioned applied pressure and duration.

[0048] An air pulse module, which is used to provide an air flow to blow air at a designated position in the user's oral cavity. In this embodiment, the designated position for blowing air refers to the user's tonsil position. Exemplarily, the user places the mandible on the support member 110. After the support member 110 rises to an appropriate height, the turntable 340 rotates to the air pulse module, and the robotic arm 300 drives the air pulse module to move slowly and smoothly to the target position corresponding to this height. The user opens the oral cavity, and the air pulse module continuously blows air into the oral cavity to stimulate the tonsil position. When the stimulation time is reached, the robotic arm 300 returns to the starting position, the interaction module prompts the user to perform a swallowing action, and the sEMG sensor measures the myoelectric value of this swallowing action. The training results include the myoelectric value at this time.

[0049] It should be noted that when the user uses the rehabilitation robot of this embodiment for the first time, the position information of the rehabilitation component 400 needs to be calibrated. This position information includes the position where the tongue suction module adsorbs the tongue, the position where the tongue pressure resistance module extends into the oral cavity, and the position where the air pulse module blows air on the tonsils. In this way, the robotic arm 300 can move to an appropriate position for the convenience of the user's training and use during the next use.

[0050] Optionally, the rehabilitation component 400 further includes: an electronic lollipop, an oral improvement massage rod, a respiratory muscle exerciser, etc., which can be installed according to different usage situations during actual use.

[0051] Optionally, the rehabilitation robot 400 further includes: a photobiomodulation module.

[0052] A photobiomodulation module is disposed on the support member 110 and is used to emit light to a specified position of the user to stimulate nerves. Specifically, a rotatable positioning member 140 is provided on the support member 110, and the irradiation module is disposed on the positioning member 140. After the user places the mandible on the support member 110, by rotating the positioning member 140, the light of the irradiation module can be emitted to the specified position for nerve stimulation. Exemplarily, the irradiation module is selected as an LED module, and red light and near-infrared light LEDs are used in combination. Further, the wavelength ranges of the red light and the near-infrared light are in the interval of 630 nm to 850 nm. Further, a wavelength of 810 nm is selected to perform biomodulation on nerve and muscle tissues. Further, the power intensity between 20 mw / cm 2 and 100 mw / cm 2 is used to irradiate the specified position. Further, the light irradiation is continuous or pulsed light irradiation and lasts for 20 - 40 minutes. Further, the irradiation position is the Jiache acupoint of the user.

[0053] Optionally, the photobiomodulation module can also be set as wearable and worn on the user's head during use, while performing rehabilitation training on the tongue muscles, auxiliary stimulation is performed on the acupoints and nerves of the user's head.

[0054] Optionally, the rehabilitation robot 400 further includes: a hyperbaric and hypobaric oxygen therapy module.

[0055] The hyperbaric and hypobaric oxygen therapy module can supply oxygen to the user while the user is performing rehabilitation training on the tongue muscles, increasing the plasticity of nerves. Exemplarily, the hyperbaric and hypobaric oxygen therapy module has a valve that can control the oxygen ratio according to a preset program (such as an intermittent hypoxic training protocol) and is connected to a breathing mask or nasal cannula at its own air outlet for the user to use.

[0056] It should be noted that the above-mentioned modules for rehabilitation training are only examples. In actual use, corresponding modules can also be set according to needs.

[0057] It should be noted that during actual training, the user can select a suitable training mode according to needs. Exemplarily, the user can select any one of the above-mentioned tongue suction device module, tongue pressure resistance module, air pulse module, or the mandible pressing the support member in different directions as the main training method, and then select any one or more from the auxiliary modules as the auxiliary training method. For example, the user can select the training method of the tongue suction device module + hyperbaric and hypobaric oxygen therapy module, or the tongue suction device module + hyperbaric and hypobaric oxygen therapy module + photobiomodulation module. A combined therapy with multiple trainings carried out simultaneously is realized.

[0058] The above-mentioned rehabilitation robot can perform targeted rehabilitation training on various symptoms of the user by setting rehabilitation modules with different uses and is applicable to a variety of application scenarios.

[0059] Optionally, the interaction module includes: a display screen 500 and a speaker.

[0060] The display screen 500 is disposed on the support body 100 and is used to display training data and training programs for the user. Further, the display screen 500 is disposed opposite to the support member 110, so that when the user places the lower jaw on the support member 110, the user can intuitively observe the display screen 500, which is convenient for observing data or assisting the user to relax. The speaker is used to cooperate with the display screen 500 to guide the user to complete the training program.

[0061] Optionally, the rehabilitation robot further includes: a storage module and a communication module.

[0062] The storage module is used to store training information, where the training information is all data involved in the training process, including training programs, training data, evaluation information, and training results. Further, the training information of the embodiments of the present application can also be stored in a cloud server.

[0063] Optionally, the user can communicate with the communication module through a terminal and view the training information in the storage module or the cloud server through the interaction module.

[0064] Wherein, the user can select any one of the above terminals to be connected to the communication module in a wired or wireless manner.

[0065] Optionally, the rehabilitation robot further includes: a switch assembly.

[0066] The switch assembly is used to turn on or off the rehabilitation robot. After the rehabilitation robot is turned on, a start instruction can be output to relevant modules. Further, the switch assembly can also be used to pause the rehabilitation robot. It should be understood that when the display screen 500 is a touch screen, the user can also perform the above operations through the touch screen.

[0067] Optionally, more visual sensors are added to the robotic arm 300 to monitor the position and dynamics of the robotic arm 300 in real time and provide feedback, further improving safety.

[0068] Optionally, the rehabilitation robot further includes: a main control module.

[0069] The main control module is used to control the actions of the support member 110, the robotic arm 300, and the turntable 340, and is also used to receive and process the data of the first sensor assembly, the first sensor, and the second sensor. At the same time, the main control module also displays training data and training programs through the above interaction module to guide the user to perform rehabilitation training.

[0070] The above-mentioned rehabilitation robot can automatically execute the set training program according to instructions. During the execution process, the robotic arm 300 can automatically switch to the corresponding rehabilitation component 400 according to the training program, drive the rehabilitation component 400 to move to the target position, and guide the user to perform rehabilitation training. The rehabilitation robot of the present application has a simple operation method, is applicable to users with various diseases, and can effectively improve the training efficiency and training effect.

[0071] Please refer to Figure 3 , based on the same inventive concept, the embodiment of the present application also provides a method for generating a personalized training program, which is applicable to the rehabilitation robot involved in the above-mentioned embodiment, and includes: Step S302, obtaining the evaluation information of the user; wherein, the evaluation information includes medical evaluation parameters and physiological ability parameters, the medical evaluation parameters include the dysphagia level and complication taboos, and the physiological ability parameters include the pressure reference value, tongue pressure ability value, and electromyogram reference value.

[0072] Specifically, the evaluation information is obtained by evaluating and measuring the user before training, and is used as a basis for the training data adopted in subsequent training.

[0073] Among them, the dysphagia level is rated by professionals according to the standard VFSS classification, including normal, mild, moderate, and severe. The complication taboos refer to other diseases related to swallowing of the user, such as limited neck movement. Generally speaking, dysphagia involves two aspects of muscle strength and swallowing skills. According to the above-mentioned dysphagia level and complication taboos, professionals can give the recommended weights of muscle strength and swallowing skills, and this weight can be used to generate a personalized training program subsequently.

[0074] The pressure reference value is obtained by a pressure sensor disposed on the surface of the supporting member. When measuring the pressure reference value, the user presses down the pressure sensor with the greatest strength.

[0075] The tongue pressure ability value is obtained by the first sensor disposed at each detection point of the tongue pressure resistance module. When measuring, the user presses the tongue against the above-mentioned detection point with the greatest strength. Exemplarily, when the number of detection points is four and each detection point is located in four quadrants respectively, the tongue pressure ability value is the average value of the data obtained at the four positions.

[0076] The electromyogram baseline value is obtained by the sEMG sensor set on the surface of the supporting member. During measurement, first let the user relax, and measure the electromyogram value of the user at this time as the initial value; then let the user perform n forceful swallows, each swallow taking t seconds, and relax and rest for t seconds after each swallow is completed. Record the maximum electromyogram value during each swallow, take the average value of n times, and subtract the initial value from this average value. The obtained value is the electromyogram baseline value. Specifically, in this embodiment, n is 5 and t is 30. This electromyogram baseline value can be used as the calibration value for training, and 80% of this value is taken as the target value for strength training.

[0077] Step S304, use the pre-constructed scheme generation model to process the evaluation information and generate a personalized training plan, so that the rehabilitation robot automatically performs rehabilitation training on the user's target object according to the training plan.

[0078] Specifically, the scheme generation model is obtained based on the user's historical training information. It should be understood that the composition of the historical training information is the same as that of the training information, and also includes the training plan, evaluation information, training data, and training results.

[0079] For the above method for generating a personalized training plan, a scheme generation model is pre-constructed, and the user's evaluation information is obtained. The constructed scheme generation model is used to generate a personalized training plan for the user, providing targeted training guidance to achieve a better rehabilitation effect; at the same time, the rehabilitation robot in this application automatically switches to the corresponding rehabilitation component according to this training plan during the training process, and drives the rehabilitation component to the target position through the robotic arm to guide the user to perform various rehabilitation trainings, which is convenient for the user to use.

[0080] Optionally, using the pre-constructed scheme generation model to process the evaluation information and generate a personalized training plan includes: Construct a scheme generation model based on fuzzy logic and decision trees; based on the constructed scheme generation model, perform fuzzy processing on the evaluation information to obtain multiple fuzzy variables; according to the fuzzy variables, use the decision tree algorithm to generate a personalized training plan.

[0081] Please refer to Figure 4 , optionally, the scheme generation model uses the decision tree algorithm to generate a personalized training plan according to the fuzzy variables, including: Step S402, construct a fuzzy membership function corresponding to the fuzzy variable.

[0082] Step S404, based on the preset fuzzy rules, comprehensively evaluate the fuzzy membership functions of each fuzzy variable to obtain a preliminary fuzzy output result.

[0083] Step S406: Use the preliminary fuzzy result as the input of the decision tree, and determine the training data through branch judgments. The above training data constitutes a training plan. The training data includes the first type of data and the second type of data. The first type of data includes training modes and training items, and the second type of data includes training intensity, training duration, and progressive step length.

[0084] Specifically, the decision tree model is constructed based on the fuzzy logic processing result. Among them, the nodes of the decision tree represent different evaluation indicators, such as the level of swallowing disorder, the proportion of muscle strength, etc.; the branches represent different value ranges or conditions; the leaf nodes represent the final selection of training modes, such as muscle strength-dominated, swallowing skill-dominated. The generation process of the decision tree includes selecting an evaluation indicator, such as the level of swallowing disorder, the pressure reference value, etc., and recursively generating subtrees according to different values of the evaluation indicator until the stopping criterion is met. The stopping criterion includes achieving a preset training effect or reaching the maximum tree depth, etc. The decision tree algorithm is adopted because users have different physiological characteristics and rehabilitation needs. The decision tree algorithm can automatically select the optimal training path and intensity according to the specific situation of the patient, generate a personalized training plan, and provide targeted training guidance for users. It should be understood that before generating the training plan model, it is also necessary to pre-construct fuzzy rules according to the structure of the model. For example, set a rule "If the level of swallowing disorder is high and the proportion of muscle strength is low, then select the swallowing skill-dominated training mode", so that the model can generate a corresponding training plan according to the evaluation information based on the above fuzzy rules.

[0085] Exemplarily, for a user A, first, a professional conducts a comprehensive evaluation of the user A to obtain the evaluation information of the user A. The evaluation information includes the actual detection data of the user A and the suggestions of the professional.

[0086] Specifically in this embodiment, first, define fuzzy rules. The model associates the input fuzzy variables with the training mode and training intensity according to the fuzzy rules. For example, define Rule 1: If the degree of swallowing disorder is moderate and the muscle dominance is high, then set the training mode to muscle strength-dominated; for another example, define Rule 2: If the degree of swallowing disorder is moderate and the muscle dominance is low, then set the training mode to swallowing skill-dominated; for another example, define Rule 3: If the pressure reference value is medium and the tongue pressure ability value is medium, then set the disorder level coefficient to 0.5.

[0087] Furthermore, determine the evaluation information. The swallowing disorder level of User A is rated as moderate using the standard VFSS grading; in terms of weights, User A has relatively weak muscle strength, and it is recommended that the muscle strength accounts for 60% and the swallowing skill accounts for 40%; in addition, User A also has a complication taboo item of limited neck movement. Furthermore, by detecting the physiological ability of User A, the pressure baseline value of User A is obtained as 220N, the average value of the tongue pressure ability in the four quadrants is 180N, and the electromyogram baseline value is 40μV.

[0088] Furthermore, perform fuzzification processing on the above evaluation information to obtain multiple fuzzified variables. For example, for the information that the VFSS grading is moderate swallowing disorder, it is converted into a fuzzy variable "degree of swallowing disorder", and a fuzzy membership function is set for it. This function is used to describe the belonging of this variable to categories such as "mild", "moderate", or "severe". It should be understood that the above fuzzy membership function can be set according to the actual situation. For example, triangular, trapezoidal, or Gaussian functions can be used to describe the trend of membership degree changing with the degree of swallowing disorder.

[0089] Furthermore, according to the specific data of the evaluation information and the fuzzy membership function, determine the fuzzy membership degrees of each evaluation information; then comprehensively evaluate each fuzzy membership degree to obtain a preliminary fuzzy output result. For example: the current value of the degree of swallowing disorder is moderate, and its membership degree is set to 1; the muscle strength accounts for 60% and the swallowing skill accounts for 40%, and it is more muscle-dominated, so the membership degrees of the recommended weights are set to 0.8 and 0.2 respectively; the pressure baseline value is 220N, which is between medium (M) and high (H) and closer to medium (M), so its μ M is defined as 0.7, and μ H is defined as 0.3. According to the fuzzy rules and the membership degrees of the degree of swallowing disorder and the weights, determine that the training mode is mainly muscle strength and supplemented by swallowing skills. According to the membership degrees of the pressure baseline value and the tongue pressure ability value, determine that the intensity coefficient is 0.5; according to the membership degree of the electromyogram baseline value, determine that the training intensity should be reduced by 20%, so the finally determined intensity coefficient is 0.4. According to the membership degree of the complication taboo, determine that the relevant training for reducing the neck load should be reduced.

[0090] Further, input the above fuzzy inference results into a decision tree to further refine each piece of training data, and finally obtain a training plan. For example, in the obtained personalized training plan, the main training mode is set as a muscle training plan, and the swallowing skill is used as an auxiliary module. Among them, the main training module includes strength training for the chin and tongue, as well as coordination training for swallowing movements; the auxiliary module includes designing neck relaxation exercises and gentle head-turning movements for the taboos regarding neck movement restrictions to relieve neck pressure and avoid aggravating swallowing disorders at the same time. For chin pressure training, the initial intensity is 220×0.4 = 88N. For tongue pressure training, the initial intensity is 180N×0.4 = 72N. The gradual progress step = (100% - 40%) / preset cycle. If the preset cycle is 4 weeks and it gradually increases from the initial intensity to 100% of the benchmark value, then the chin pressure training increases by 33N per week (i.e., the difference between 88N and 220N divided by 4), and the tongue pressure training increases by 27N per week. In addition, according to the specific situation of the user, it is set that the chin pressure training is 10 times per group, each time lasting 5 seconds, with a 30-second break between groups; the tongue pressure training is 8 times per group, each time lasting 6 seconds, with a 40-second break between groups.

[0091] Optionally, during the training process, the training status of the user is recorded in real time. The recorded information includes the fatigue level of the patient, whether there is discomfort, whether the training items can be completed, etc.; in addition, the embodiments of the present application also monitor the training parameters of the user during or after the training. The training parameters include the pressure value of the lower jaw on the support, the pressure values of the tongue on each detection point of the tongue pressure resistance module and the duration, and the myoelectric value generated during swallowing.

[0092] Optionally, after completing the above training, the user is further guided to conduct a swallowing skill training. The central myoelectric value of the target area for training is 50% of the calibration value, and the size of the area is 30% of the calibration value. The myoelectric value generated by the swallowing muscles and swallowing indicators are recorded during the training process. Among them, the swallowing indicators include: Acceleration, which is calculated by dividing the change in the myoelectric value generated by swallowing by the time taken to reach the highest point, and is used to reflect the strength of the tongue muscle.

[0093] Hit rate, which is calculated by dividing the number of accurate tasks successfully completed by the number of attempts. If the hit rate is greater than 80%, it indicates that this skill training has achieved the expectation.

[0094] Relative time error, which is the difference between the time point of the peak of the sEMG sensor and the set target area center time point divided by the duration of a single training session (usually set to 30 seconds).

[0095] Relative amplitude error, which is the difference between the voltage value of the peak of the sEMG sensor and the voltage value of the set target area center divided by the peak of the sEMG sensor.

[0096] The above training status, training parameters, and swallowing metrics constitute the training results.

[0097] Optionally, generating a training plan based on the user's evaluation information further includes: Using a large language model to process the evaluation information to obtain a personalized training plan; among them, the large language model includes a base model, a fine-tuning model, a vertical model, a distillation inference model, and an agent created by the above models.

[0098] Among them, the base model optimizes the current plan by retrieving a large amount of historical training data to enable it to generate more accurate answers or training plans by integrating external knowledge. Exemplarily, the base model can adopt GPT, DeepSeek, LLaMA, or PaLM-2.

[0099] The fine-tuning model is fine-tuned on the basis of the original model (the base model) through additional task data to make it more suitable for generating personalized training plans. Among them, the original model can be the above-mentioned pre-trained plan generation model, or any one of the base model, the vertical model, and the distillation inference model.

[0100] The vertical model is a model specifically trained directly on a large amount of data in a specific field for a certain field. Exemplarily, the vertical model can adopt BioGPT, FinGPT, or LegalGPT.

[0101] The distillation inference model compresses the capabilities of a large model into a small model through knowledge distillation technology. Exemplarily, the distillation inference model can adopt DistilBERT or Med-PaLM. It should be noted that in actual use, any one of the above large language models can be used to generate a personalized training plan, and an AI agent created based on these models can also be used to execute. It can also be combined with multiple models as needed, such as the base model + the fine-tuning model.

[0102] Optionally, the above method for generating a personalized training plan further includes: Obtain and analyze the training results, and adjust the training plan according to the analysis results.

[0103] Specifically, the purpose of the analysis is to determine whether the current training plan is suitable for the patient's physical condition and whether the expected training effect has been achieved. Through adjustment, the training plan can be made more suitable for the patient's current physiological data and achieve a better training effect.

[0104] Optionally, analyzing the training results and adjusting the training plan according to the analysis results includes: Compare the training parameters and swallowing indicators with the preset target values respectively to obtain difference information. If the difference information exceeds the preset range, adjust the training plan. Among them, adjusting the training plan includes adjusting the content and weight of the first type of data, and adjusting the threshold range of the second type of data.

[0105] The above preset target value is determined according to the physiological ability parameters and swallowing indicators, and can be an actual measured value or the value multiplied by a proportionality coefficient.

[0106] Further, if the difference information exceeds the preset range, it is considered that the training requirements are not met, and the training plan is automatically adjusted according to the preset logic. Exemplarily, during the training process, the patient feels pain or fatigue and cannot complete the training items as required. Then, according to the training parameters obtained in real time, update the membership degrees of each input variable in the fuzzy logic system, re-run the decision tree, obtain a new training mode and new training items, and adjust the weights of the new training mode and training items to form a new training plan. In addition, the threshold range of the second type of data can also be adjusted in real time. For example, if the muscle fatigue degree is high, sEMG fatigue index > 0.7, then reduce the current intensity parameter by 15%, shorten the training duration, extend the rest duration, and insert 5 minutes of photobiomodulation at the same time. If the completion rate reaches 90% for 3 consecutive days, appropriately increase the intensity and progressive step. Further, the adjustment methods also include: optimizing the node weights of the decision tree based on the preset rules of the decision tree; in addition, the rule base can also be iteratively updated, and when the current training plan is not suitable for this user, the rule base is automatically updated and replaced.

[0107] Optionally, analyzing the training results and adjusting the training plan according to the analysis results also includes: Periodically obtain the training results to obtain a training data set; generate a development curve or heat map of the swallowing ability according to the training data set, and then predict the training effect according to the development curve or heat map. If the predicted training effect is poor, the training plan can be adjusted based on the same steps as above.

[0108] Optionally, the embodiments of the present application also calculate the completion rate, fatigue degree, and error rate according to the above training results, and then according to the reward function: R = α×(completion rate) - β×(fatigue degree) - γ×(error rate), obtain a score, which is used to evaluate the effect of this training. Among them, α, β, and γ are all preset parameter values. If the score is lower than the preset score, the training plan can be adjusted based on the same steps as above.

[0109] The above method for generating a personalized training plan uses multiple algorithms to construct a plan generation model. The plan generation model generates a personalized training plan according to the evaluation information and adjusts the training plan according to the training results, providing targeted training guidance for users and effectively improving the rehabilitation effect.

[0110] Please refer to Figure 5 , based on the same inventive concept, the embodiment of the present application further provides a rehabilitation training method, which is applicable to the rehabilitation robot involved in the above embodiment, including: Step S502, according to the detection signal obtained by the first sensor component, confirm whether the supporting member reaches the target position.

[0111] Specifically, the first sensor component includes a pressure sensor and an sEMG sensor, and the detection signals obtained respectively include a pressure value and an electromyogram value. Among them, the pressure value represents the supporting force of the supporting member on the user's mandible. By analyzing this supporting force, it can be known whether the user's target posture is standard. In this embodiment, the judgment basis for whether the target posture is standard is to confirm whether the user's back is straight and the head is not too high or too low. Exemplarily, before training, the user is first guided to relax, and then the reference threshold value when the user is in the standard posture is detected. It should be understood that the reference threshold value includes a reference pressure value and a reference electromyogram value, and the reference threshold value has a certain range. During the rising process of the supporting member, the pressure value and the electromyogram value change in real time. When the pressure value and the electromyogram value reach the above reference threshold value, it is considered that the supporting member has risen to the target position.

[0112] Further, in the case where there are multiple pressure sensors, the pressure distribution of the mandible on the supporting member can be calculated according to each pressure value, and at this time, the reference threshold value includes a reference pressure distribution value and a reference electromyogram value.

[0113] It should be noted that when the user's swallowing muscles are in a relaxed state, the electromyogram value is a stable straight line close to zero. Therefore, the reference electromyogram value is a value close to zero. When tense or exerting force, the electromyogram value will be unstable and the fluctuation amplitude will increase.

[0114] Step S504, when the supporting member reaches the target position, execute the training plan so that the robotic arm drives the rehabilitation component to move to the target position along a preset trajectory, guiding the user to perform rehabilitation training on the tongue muscles; wherein, the training plan is pre-generated by using the generation method of the personalized training plan disclosed in the above embodiment.

[0115] For the above rehabilitation training method, the rehabilitation robot can automatically execute the training plan according to the instruction, drive different rehabilitation components to the target position by the robotic arm, guide the user to perform rehabilitation training, with simple operation and convenient for the user to use.

[0116] Optionally, the above rehabilitation training method further includes: Calculate the pressure distribution of the user's mandible on the supporting member according to the pressure value; confirm whether the placement position of the user's mandible is correct according to the pressure distribution.

[0117] Specifically, before and during training, the user's lower jaw moves left and right, resulting in the rehabilitation component not fully matching the user's target object, and the training effect is not good. Exemplarily, the reference pressure value of the user at the correct position is detected in advance. If the pressure value detected in real time exceeds the allowable range of the reference pressure value, it is considered that the user's lower jaw is offset, and the user is reminded to move to the correct position.

[0118] Please refer to Figure 6 , optionally, the preset trajectory is calculated based on the starting position and the target position of the robotic arm. The steps of calculating the preset trajectory include: Step S602, configure a coordinate system for each joint of the robotic arm and determine the motion parameters of each joint. The motion parameters include joint length, joint twist angle, joint offset, and joint angle.

[0119] Step S604, construct the homogeneous transformation matrix of each joint according to the motion parameters, multiply the homogeneous transformation matrices in sequence to obtain the total transformation matrix, and the total transformation matrix is used to represent the pose of the target coordinate system relative to the starting coordinate system.

[0120] Step S606, according to the total transformation matrix, determine the mapping relationship between the motion parameters and the pose through forward kinematics and inverse kinematics.

[0121] Step S608, based on the mapping relationship, use the interpolation algorithm to generate a series of intermediate poses between the starting pose and the target pose. The starting pose, the target pose, and each intermediate pose constitute the preset trajectory.

[0122] Exemplarily, the robotic arm of the embodiment of the present application is driven by 6 independent stepping motors. To achieve a given pose of the end effector of the robotic arm in six-dimensional space, first establish a kinematic model of the robotic arm by the Denavit-Hartenberg (DH) parameter method, which is used to calculate the forward and inverse kinematic relationships between the position and attitude of the end effector and the motion angles of each axis. Specifically, allocate a three-dimensional Cartesian coordinate system for each joint of the robotic arm according to the DH convention, then the homogeneous transformation matrix of a single joint can be constructed according to the above motion parameters, and the coordinates of two adjacent joints are transformed from the coordinate system of the previous joint to the coordinate system of the next joint, so as to obtain the coordinate transformation relationship between each coordinate system. Furthermore, according to the total transformation matrix and forward kinematics, given the joint angles of each joint, the exact position and attitude of the end effector are determined. Then, according to the total transformation matrix and inverse kinematics, given the desired pose of the end, the motion angles of each joint are obtained by solving the non-linear equations.

[0123] Further, determine the starting pose and the target pose according to the training requirements, and use linear interpolation or circular interpolation to generate a series of intermediate poses, constituting a smooth motion trajectory. For each interpolation point, solve the corresponding joint angle vector according to the inverse kinematics of the robotic arm to ensure that the trajectory is smooth and continuous in both the joint space and the working space.

[0124] Optionally, the steps of calculating the preset trajectory further include: Use quintic spline interpolation to calculate the change curve of the joint angle of each joint; the change curve is used to achieve smooth acceleration or deceleration of the robotic arm.

[0125] Specifically, for each joint, set the starting angle, the ending angle, and the intermediate key frames, and use a quintic polynomial to plan the change curve of the joint angle over time. In this way, the quintic polynomial can ensure that the displacement, velocity, and acceleration are all continuously zero at the starting and ending points, thereby achieving smooth acceleration and deceleration and avoiding shocks and jitters during the movement process.

[0126] Further, convert the continuous joint angle trajectory obtained by quintic spline interpolation into drive instructions for the corresponding stepper motors, and achieve the given pose of the end effector of the robotic arm in the six-dimensional space by precisely controlling the movement of each axis, completing the coordinated movement of multiple axes of the robotic arm.

[0127] The above rehabilitation training method establishes a kinematic model of the robotic arm based on the DH parameter method, realizes the mapping between the pose and the joint angle through forward and inverse kinematics; then uses an interpolation algorithm to generate a smooth motion path, and then uses quintic spline interpolation to design smooth acceleration and deceleration for the trajectory; finally, generates specific control instructions to drive the stepper motors to complete the realization of the target pose. This application sets the starting position and the target position according to the training needs, and then plans a smooth motion path according to the above positions, which can make the movement of the robotic arm smoother.

[0128] Optionally, when a second sensor component is provided on the robotic arm, the rehabilitation training method of this application further includes: Execute a protection strategy according to the protection data detected by the second sensor component.

[0129] Specifically, the protection data obtained by the second sensor component includes the force parameters of the robotic arm, the rotation angle parameters of each joint on the robotic arm, and the current parameters of the motors driving the robotic arm.

[0130] Optionally, executing the protection strategy according to the protection data includes: When the force parameter is greater than the preset force threshold, the robotic arm returns to the starting position.

[0131] When the rotation angle parameter is greater than the preset rotation threshold, the joint returns to the initial angle.

[0132] When the current parameter is greater than the preset current threshold, enter the stall protection mode to reduce or stop the current supply.

[0133] For the above rehabilitation training method, the rehabilitation robot can automatically switch to the corresponding rehabilitation component according to the training plan, and drive the rehabilitation component to the target position through the robotic arm to guide the user to perform rehabilitation training, which is convenient for the user to use.

[0134] Furthermore, the present application also monitors the user's training results in real time during and after the training, and adjusts the training plan in real time according to the training results, providing targeted training guidance for the user and improving the rehabilitation effect.

[0135] Furthermore, the present application establishes the kinematic model of the robotic arm based on the DH parameter method, uses the interpolation algorithm to generate a smooth motion path, and then uses the quintic spline interpolation to design the smooth acceleration and deceleration of the trajectory, planning a smooth motion path for the robotic arm, which can make the robotic arm move more smoothly.

[0136] Furthermore, the present application also monitors the protection data of the robotic arm in real time and executes the protection strategy in a timely manner according to the protection data, ensuring the safety of the rehabilitation robot.

[0137] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are sequentially shown according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or steps in other steps.

[0138] Based on the same inventive concept, the embodiment of the present application also provides a system for generating a personalized training plan. This system is applicable to the above method for generating a personalized training plan. The implementation solution provided by this system to solve the problem is similar to the implementation solution described in the above method. Therefore, the specific limitations in one or more device embodiments provided below can refer to the limitations on the method in the above text, and will not be repeated here.

[0139] Please refer to Figure 7 , in one embodiment, the system for generating a personalized training plan includes: an acquisition module and a plan generation module.

[0140] An acquisition module for acquiring the user's evaluation information; the evaluation information includes medical evaluation parameters and physiological ability parameters, the medical evaluation parameters include the dysphagia level and complication taboos, and the physiological ability parameters include the stress reference value, tongue pressure ability value, and electromyogram reference value.

[0141] A scheme generation module for processing the evaluation information using a pre-constructed scheme generation model to generate a personalized training scheme, so that the rehabilitation robot automatically performs rehabilitation training on the user's target object according to the training scheme.

[0142] Optionally, the scheme generation module processes the evaluation information to generate a personalized training scheme, including: constructing a scheme generation model based on fuzzy logic and decision trees; performing fuzzy processing on the evaluation information based on the constructed scheme generation model to obtain multiple fuzzy variables; generating a personalized training scheme using a decision tree algorithm according to the fuzzy variables.

[0143] Optionally, the scheme generation model generates a personalized training scheme using a decision tree algorithm according to the fuzzy variables, including: constructing a fuzzy membership function corresponding to the fuzzy variables; comprehensively evaluating the fuzzy membership functions of the fuzzy variables based on preset fuzzy rules to obtain a preliminary fuzzy output result; using the preliminary fuzzy result as the input of the decision tree, and determining the training data through branch judgment, and the above training data constitutes the training scheme. Among them, the training data includes training mode, training item, training intensity, training duration, and progressive step length.

[0144] Optionally, the scheme generation module is also used to process the evaluation information using a large language model to obtain a personalized training scheme; among them, the large language model includes a base model, a tuning model, a vertical model, a distilled inference model, and an agent created by the above models.

[0145] Optionally, the scheme generation module is also used to obtain and analyze the training results, and adjust the training scheme according to the analysis results.

[0146] Optionally, the scheme generation module analyzes the training results and adjusts the training scheme according to the analysis results, including: comparing the training parameters and swallowing indicators with preset target values respectively to obtain difference information, and if the difference information exceeds the preset range, adjusting the training scheme; among them, adjusting the training scheme includes adjusting the content and weight of the first type of data, and adjusting the threshold range of the second type of data.

[0147] Optionally, the scheme generation module analyzes the training results and adjusts the training scheme, and also includes: periodically obtaining the training results to obtain a training data set; generating a development curve or heat map of swallowing ability according to the training data set, and then predicting the training effect according to the development curve or heat map.

[0148] The above-mentioned personalized training plan generation system constructs a plan generation model using multiple algorithms. This plan generation model generates a personalized training plan based on the evaluation information and adjusts the training plan according to the training results, providing targeted training guidance for users and effectively improving the rehabilitation effect.

[0149] Based on the same inventive concept, the embodiment of the present application also provides a rehabilitation training system. This rehabilitation training system is applicable to the above-mentioned rehabilitation training method. The implementation solution provided by this system for solving problems is similar to the implementation solution described in the above method. Therefore, the specific limitations in one or more of the following device embodiments can refer to the limitations on the method in the above text and will not be repeated here.

[0150] Please refer to Figure 8 , in one embodiment, the rehabilitation training system includes: a position confirmation module and an execution module.

[0151] The position confirmation module is used to obtain the detection signal of the first sensor component and confirm whether the supporting member reaches the target position according to the detection signal.

[0152] The execution module is used to execute the training plan when the supporting member reaches the target position, so that the robotic arm moves along a preset trajectory and drives the rehabilitation component to move to the target position to guide the user to perform rehabilitation training on the tongue muscles. Among them, the training plan is pre-generated by using the personalized training plan generation method disclosed in the above embodiment.

[0153] Optionally, the position confirmation module confirms whether the supporting member reaches the target position according to the detection signal, including: calculating the pressure distribution of the user's lower jaw on the supporting member according to the pressure value; and considering that the supporting member reaches the target position when the pressure distribution and the myoelectric value reach the preset reference threshold.

[0154] Optionally, the position confirmation module is also used to confirm whether the placement position of the user's lower jaw is correct according to the pressure distribution.

[0155] Optionally, the above-mentioned rehabilitation training system further includes: a trajectory planning module.

[0156] A trajectory planning module is used to plan a preset trajectory for the robotic arm. The preset trajectory is calculated based on the starting position and the target position of the robotic arm. The steps of calculating the preset trajectory include: configuring a coordinate system for each joint of the robotic arm and determining the motion parameters of each joint. The motion parameters include joint length, joint twist angle, joint offset, and joint angle; based on the motion parameters, constructing the homogeneous transformation matrix of each joint, multiplying the homogeneous transformation matrices in sequence to obtain the total transformation matrix, and the total transformation matrix is used to represent the pose of the target coordinate system relative to the starting coordinate system; according to the total transformation matrix, through forward kinematics and inverse kinematics, determining the mapping relationship between the motion parameters and the pose; based on the mapping relationship, using an interpolation algorithm to generate a series of intermediate poses between the starting pose and the target pose. The starting pose, the target pose, and each intermediate pose constitute the preset trajectory.

[0157] Optionally, the trajectory planning module is further used to determine the starting pose and the target pose according to the training requirements, and use linear interpolation or circular arc interpolation to generate a series of intermediate poses to form a smooth motion trajectory. For each interpolation point, the corresponding joint angle vector is solved according to the inverse kinematics of the robotic arm to ensure that the trajectory is smooth and continuous in both the joint space and the working space.

[0158] Optionally, the trajectory planning module is further used to use quintic spline interpolation to calculate the change curve of the joint angle of each joint; the change curve is used to achieve smooth acceleration or deceleration of the robotic arm.

[0159] Optionally, the above rehabilitation training system further includes: a protection module.

[0160] The protection module is used to execute a protection strategy according to the protection data detected by the second sensor component, and the protection strategy includes: when the force parameter is greater than the preset force threshold, the robotic arm returns to the starting position; when the rotation angle parameter is greater than the preset rotation threshold, the joint returns to the initial angle; when the current parameter is greater than the preset current threshold, enter the stall protection mode to reduce or stop the current supply.

[0161] In the above rehabilitation training system, the rehabilitation robot can automatically switch to the corresponding rehabilitation component according to the training plan, and drive the rehabilitation component to the target position through the robotic arm to guide the user to perform rehabilitation training, which is convenient for the user to use.

[0162] Furthermore, the present application also monitors the training results of the user in real time during and after the training, and adjusts the training plan in real time according to the training results to provide targeted training guidance for the user and improve the rehabilitation effect.

[0163] Furthermore, this application establishes a kinematic model of the robotic arm based on the DH parameter method, generates a smooth motion path using an interpolation algorithm, and then uses quintic spline interpolation to design smooth acceleration and deceleration for the trajectory, planning a smooth motion path for the robotic arm, which can make the robotic arm move more smoothly.

[0164] Furthermore, this application also monitors the force condition of the robotic arm, the rotation condition of the joints, and the current condition of the motors in real time, and executes protection strategies in a timely manner according to the monitored conditions to ensure the safety of the rehabilitation robot.

[0165] Each module in the above-mentioned personalized training plan generation system and rehabilitation training system can be implemented in whole or in part by software, hardware, and their combinations. Each of the above modules can be embedded in or independent of the processor in a computer device in the form of hardware, or stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0166] In a feasible embodiment, a computer device is provided. This computer device can be a terminal, and its internal structure diagram can be as Figure 9 shown. This computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of this computer device is used for the processor to exchange information with external devices. The communication interface of this computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be achieved through WIFI, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it realizes the above-mentioned rehabilitation training method. The display unit of this computer device is used to form a visually visible picture, which can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of this computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad set on the shell of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0167] Those skilled in the art can understand, Figure 9The structure shown is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0168] In a feasible embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the method steps in the above-mentioned generation method of the personalized training plan and the rehabilitation training method are implemented.

[0169] In a feasible embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method steps in the above-mentioned generation method of the personalized training plan and the rehabilitation training method are implemented.

[0170] In a feasible embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the method steps in the above-mentioned generation method of the personalized training plan and the rehabilitation training method are implemented.

[0171] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions.

[0172] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0173] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0174] The above-described embodiments merely represent several implementation manners of the present application. Their descriptions are relatively specific and detailed, but they should not be construed as limitations on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method for generating a personalized training program, applicable to a rehabilitation robot, characterized in that: include: Acquire the user's evaluation information; the evaluation information includes medical evaluation parameters and physiological ability parameters, the medical evaluation parameters include swallowing disorder level and complication contraindications, and the physiological ability parameters include pressure reference value, tongue pressure ability value and myoelectric reference value; The evaluation information is processed using a pre-built solution generation model to generate a personalized training solution, so that the rehabilitation robot automatically performs rehabilitation training on the user's target object based on the training solution.

2. The method according to claim 1, characterized in that The method of using a pre-built solution generation model to process the evaluation information and generate a personalized training solution includes: Constructing the solution generation model based on fuzzy logic and decision tree; Based on the constructed solution generation model, the evaluation information is fuzzy processed to obtain a plurality of fuzzy variables; and according to the fuzzy variables, a decision tree algorithm is used to generate a personalized training solution.

3. The method according to claim 2, characterized in that The solution generation model generates a personalized training solution based on the fuzzy variables using a decision tree algorithm, including: Constructing a fuzzy membership function corresponding to the fuzzy variable; Based on the preset fuzzy rules, the fuzzy membership function of each fuzzy variable is comprehensively evaluated to obtain a preliminary fuzzy output result; The preliminary fuzzy result is used as the input of the decision tree, and training data is determined through branch judgment; wherein the training data includes a first type of data and a second type of data, the first type of data includes a training mode and a training item, and the second type of data includes a training intensity, a training duration, and a progressive length; The training data mentioned above constitute the training plan.

4. The method according to claim 1, characterized in that The method of using a pre-built solution generation model to process the evaluation information and generate a personalized training solution includes: Constructing the solution generation model based on the large language model; Based on the constructed solution generation model, the evaluation information is processed to obtain a personalized training solution; wherein the large language model includes a base model, a tuning model, a vertical model, a distillation reasoning model, and an intelligent agent created by the above models.

5. The method according to claim 1 or 4, characterized in that: The method further comprises: Get training results; Analyze the training results, and adjust the training program according to the analysis results; wherein the training results include training parameters and swallowing indicators; the training results are obtained by the rehabilitation robot guiding the user to perform rehabilitation training based on the training program.

6. The method according to claim 5, characterized in that The analyzing the training results and adjusting the training scheme according to the analyzing results include: Comparing the training parameters and the swallowing index with the preset target values ​​respectively to obtain difference information, and if the difference information exceeds the preset range, adjusting the training plan; Wherein, adjusting the training scheme includes adjusting the content and weight of the first type of data, and adjusting the threshold range of the second type of data.

7. The method according to claim 5, characterized in that The step of analyzing the training results and adjusting the training scheme according to the analysis results further includes: Periodically acquiring the training results to obtain a training data set; generating a development curve or a heat map of swallowing ability according to the training data set; The training effect is predicted according to the development curve or the heat map, and the training program is adjusted according to the training effect.

8. A system for generating a personalized training program, characterized in that: The system comprises: An acquisition module, used to acquire the user's evaluation information; the evaluation information includes medical evaluation parameters and physiological ability parameters, the medical evaluation parameters include swallowing disorder level and complication contraindications, and the physiological ability parameters include pressure reference value, tongue pressure ability value and myoelectric reference value; The program generation module is used to process the evaluation information using a pre-built program generation model to generate a personalized training program so that the rehabilitation robot can automatically perform rehabilitation training on the user's target object based on the training program.

9. A rehabilitation training method, characterized in that: Applicable to a rehabilitation robot, the rehabilitation robot comprises a supporting member, a mechanical arm, a rehabilitation component and a first sensor component arranged on the supporting member; the method comprises: confirming whether the supporting member has reached the target position according to the detection signal obtained by the first sensor assembly; the detection signal includes an electromyographic value; When the supporting member reaches the target position, a training plan is executed so that the robotic arm drives the rehabilitation component to move to the target position according to a preset trajectory, guiding the user to perform rehabilitation training on the target object; wherein the training plan is pre-generated according to the method for generating a personalized training plan according to any one of claims 1-7.

10. The method according to claim 9, characterized in that The detection signal also includes a pressure value; The step of confirming whether the supporting member has reached the target position according to the detection signal obtained by the first sensor assembly includes: Calculating the pressure distribution of the user's lower jaw on the supporting member according to the pressure value; When the pressure distribution and the myoelectric value reach a preset reference threshold, it is considered that the supporting member has reached the target position.

11. The method according to claim 9, characterized in that The rehabilitation robot includes an interaction module, a storage module and a communication module; the training information is stored in the storage module or a cloud server; the training information includes a training plan, evaluation information, training data and training results; The user is connected to the communication module through a terminal, and views the training information in the storage module or the cloud server through the interaction module.

12. The method according to claim 9, characterized in that The rehabilitation component includes a tongue suction module, a tongue pressure resistance module and an air pulse module; The step of guiding the user to perform rehabilitation training on the target object includes: The tongue suction module fixes the user's tongue and pulls the tongue to perform rehabilitation exercises through the movement of the mechanical arm; Or, the tongue pressure resistance module provides a plurality of detection points at different positions for the tongue to contact; Or, the air pulse module blows air to a designated position in the user's mouth; Alternatively, the user is guided to place the lower jaw on the supporting member and press down in different directions.

13. The method according to claim 12, characterized in that The detection signal also includes a pressure value; the method also includes: Calculating the pressure distribution of the user's lower jaw on the supporting member according to the pressure value; Whether the user's jaw is positioned correctly is determined based on the pressure distribution.

14. The method according to claim 12, characterized in that The rehabilitation robot also includes a photobiomodulation module and a high and low oxygen treatment module; The guiding the user to perform rehabilitation training on the target object also includes: The training is performed using any module and / or auxiliary module of the rehabilitation component; wherein the auxiliary module is any one or more of the photobiomodulation module and the hyperoxia and hypoxia therapy module.

15. The method according to claim 9, characterized in that The preset trajectory is calculated based on the starting position and the target position of the robot arm, and the step of calculating the preset trajectory includes: Configuring a coordinate system for each joint of the robotic arm and determining motion parameters of each joint, wherein the motion parameters include joint length, joint torsion angle, joint offset and joint angle; According to the motion parameters, construct a homogeneous transformation matrix of each joint, and multiply the homogeneous transformation matrices in sequence to obtain a total transformation matrix, wherein the total transformation matrix is ​​used to represent the position and posture of the target coordinate system relative to the starting coordinate system; According to the total transformation matrix, the mapping relationship between the motion parameters and the posture is determined through forward kinematics and inverse kinematics; Based on the mapping relationship, an interpolation algorithm is used to generate a series of intermediate postures between the starting posture and the target posture; The starting posture, the target posture and each intermediate posture constitute the preset trajectory.

16. The method according to claim 15, characterized in that The step of calculating the preset trajectory also includes: Quintic spline interpolation is used to calculate a variation curve of the joint angle of each joint; the variation curve is used to achieve smooth acceleration or deceleration of the robotic arm.

17. The method according to claim 9, characterized in that A second sensor assembly is provided on the mechanical arm; The second sensor component is used to obtain protection data of the mechanical arm, wherein the protection data includes force parameters, rotation angle parameters of each joint on the mechanical arm, and current parameters of the motor driving the mechanical arm; A protection strategy is executed according to the protection data.

18. The method according to claim 17, characterized in that The executing the protection strategy according to the protection data includes: When the force parameter is greater than a preset force threshold, the robotic arm returns to the starting position; When the rotation angle parameter is greater than a preset rotation threshold, the joint returns to an initial angle; When the current parameter is greater than a preset current threshold, the device enters a stall protection mode to reduce or stop current supply.

19. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 or 9 to 18 are implemented.

20. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 or 9 to 18 are implemented.

Images