CN121354801A

CN121354801A - Rehabilitation training optimization method, system, equipment and medium based on artificial intelligence

Info

Publication number: CN121354801A
Application number: CN202511485027.9A
Authority: CN
Inventors: 王芮; 彭世杰; 郝晓青; 张晓梅; 李博宇
Original assignee: Xiong'an Xuanwu Hospital
Current assignee: Xiong'an Xuanwu Hospital
Priority date: 2025-10-17
Filing date: 2025-10-17
Publication date: 2026-01-16

Abstract

The application relates to a rehabilitation training optimizing method, system, equipment and medium based on artificial intelligence. The method comprises the steps of obtaining basic physiological data of a user, wherein the basic physiological data at least comprise heart injury reasons, age parameters, heart rate parameters, heart pressure parameters and body quality indexes, generating an initialization motion plan of the user based on the basic physiological data, obtaining motion image data, real-time heart rate parameters, tired image data and tired voice data of the user when the initialization motion plan is executed, quantifying the actions of the initialization motion plan on the user based on the motion image data, the real-time heart rate parameters, the tired image data and the tired voice data to obtain plan feedback of the initialization motion plan, and updating the initialization motion plan by using the plan feedback and/or historical training data of the user to obtain the updated initialization motion plan. The heart rehabilitation training efficiency can be improved by adopting the method.

Description

Rehabilitation training optimization method, system, equipment and medium based on artificial intelligence

Technical Field

The invention belongs to the field of heart rehabilitation training, and particularly relates to an artificial intelligence-based rehabilitation training optimization method, system, equipment and medium.

Background

Along with the development of rehabilitation training technology, an intelligent rehabilitation monitoring system based on wearable equipment appears, and the system can continuously acquire basic physiological data such as heart rate and blood pressure of a patient and realize remote transmission, so that a currently mainstream 'periodic evaluation-static prescription-autonomous execution' heart rehabilitation mode is formed.

In the existing intelligent rehabilitation monitoring system based on the wearable equipment, a heart rehabilitation doctor firstly formulates an initialization rehabilitation scheme comprising exercise type, intensity, frequency and duration based on electronic health records and primary evaluation results of a patient, the patient performs training according to the established scheme in a home environment, heart rate changes are recorded through equipment such as an intelligent bracelet and the like, a training log is manually filled in, the doctor regularly acquires training data and subjective feeling of the patient through a video follow-up or outpatient review mode, the rehabilitation progress is judged by combining clinical experience, and a exercise prescription is adjusted in a stepwise mode in units of weeks or months, for example, the exercise intensity is properly increased or single training time is prolonged.

However, the existing rehabilitation mode has obvious limitations that firstly, a static exercise prescription is difficult to adapt to the physiological state and rehabilitation requirement of daily fluctuation of a patient, a targeted dynamic adjustment mechanism is lacked, secondly, manual evaluation depends on experience of doctors and has subjective deviation, accurate quantitative analysis cannot be realized, thirdly, the traditional method is most critical to lack of real-time monitoring and instant intervention capability in the training execution process, and when the posture of the patient is wrong or overtired, the patient is difficult to correct in time, so that the rehabilitation effect is influenced, potential safety hazards exist, and the heart rehabilitation training efficiency is low.

Disclosure of Invention

Based on the above, it is necessary to provide an artificial intelligence based rehabilitation training optimizing method, system, device and medium capable of improving the heart rehabilitation training efficiency.

In a first aspect, the present application provides an artificial intelligence based rehabilitation training optimization method, including:

Acquiring basic physiological data of a user, wherein the basic physiological data at least comprise heart injury reasons, age parameters, heart rate parameters, heart pressure parameters and body quality indexes;

Generating an initialized exercise plan of the user based on the basic physiological data, wherein the initialized exercise plan is used for indicating the exercise type, the exercise frequency, the exercise time and the exercise heart rate interval of the exercise of the user;

Acquiring moving image data, real-time heart rate parameters, tired image data and tired voice data of a user when performing an initialization movement plan;

Quantifying the effect of the initialized motion plan on the user based on the motion image data, the real-time heart rate parameter, the tired feeling image data and the tired feeling voice data, and obtaining plan feedback of the initialized motion plan;

and updating the initialized motion plan by using the planning feedback and/or the historical training data of the user to obtain the updated initialized motion plan.

Further, generating an initialized motion plan for the user based on the underlying physiological data, comprising:

Based on preset exercise intensity parameters, the following formulas are used for calculating exercise heart rate interval data by combining age parameters and heart rate parameters:

Wherein, the Is the exercise heart rate interval,Is a parameter of the heart rate,Is a first motion intensity parameter, which is a first motion intensity parameter,Is an age parameter, which is a function of the age,Is a second motion intensity parameter;

Inputting the cause of heart injury and the body quality index into a preset motion type classification model to obtain motion type data, wherein the motion type classification model is constructed based on a random forest algorithm;

inputting the heart pressure parameters and the body mass index into a preset multiple linear regression model to obtain movement frequency data and movement time data;

And obtaining an initialization exercise plan based on the exercise heart rate interval data, the exercise type data, the exercise frequency data and the exercise time data.

Further, quantifying the effect of the initialized motion plan on the user based on the motion image data, the real-time heart rate parameter, the tired image data and the tired voice data, and obtaining the plan feedback of the initialized motion plan, including:

based on a preset standard motion key point sequence and motion image data, quantifying the difference between the actual motion of a user and the correct motion of the user to obtain gesture difference data, wherein the gesture difference data comprises gesture angle change data and gesture key point change data;

based on the gesture key point change data, quantifying the difference between the coordinates of the key points when the user moves and the coordinates of the key points of the standard gesture to obtain a gesture accuracy score;

generating auditory feedback information based on the posture angle change data, the posture key point change data and the standard movement key point sequence, wherein the auditory feedback information is used for indicating a user to adjust the training posture through the auditory information;

predicting the fatigue degree of the user based on the tired image data and the tired voice data to obtain a comprehensive fatigue score;

Based on the real-time heart rate parameters and the initialized exercise plan, quantifying the heart rate normal degree of the user during exercise, and obtaining the heart rate coincidence degree;

according to preset weighting weights, carrying out weighted fusion on the gesture correctness scores and the heart rate coincidence degree to obtain the exercise success rate;

based on the auditory feedback information, the integrated fatigue score, and the motion success rate, planning feedback is generated.

Further, based on a preset standard motion key point sequence and motion image data, quantifying the difference between the actual motion of the user and the correct motion of the user, and obtaining gesture difference data, including:

extracting coordinates of key points from moving image data based on a preset key point positioning area to obtain a real-time moving key point coordinate sequence;

Calculating the angle change of the joints when the user moves based on the real-time movement key point coordinate sequence and a preset joint node set to obtain gesture angle change data;

Based on the gesture angle change data, extracting periodic characteristics of user movement to obtain actual movement periodic parameters;

determining a key point coordinate sequence of correct motion of a user based on the standard motion key point sequence and the actual motion period parameter to obtain a standard reference key point coordinate sequence;

Calculating the difference between the real-time motion key point coordinate sequence and the standard reference key point coordinate sequence to obtain gesture key point change data;

based on the posture key point change data and the posture angle change data, posture difference data is obtained.

Further, generating audible feedback information based on the posture angle change data, the posture key point change data, and the standard motion key point sequence, includes:

calculating the difference value between the angle change of the joint and the angle change of the standard posture when the user moves based on the standard reference key point coordinate sequence and the posture angle change data to obtain a joint angle deviation sequence;

Based on a preset deviation threshold, extracting joints and key points from the joint angle deviation sequence and gesture key point change data to obtain a deviation node sequence;

determining correction sequences of joints and key points in a deviation node sequence based on preset deviation weights to obtain a node feedback sequence;

and converting the node feedback sequence based on a preset language database and a sequencing rule to obtain a language feedback text, and generating auditory feedback information based on the language feedback text.

Further, predicting the fatigue degree of the user based on the tired image data and the tired voice data to obtain a comprehensive fatigue score, including:

According to tired feel image data, adopting a preset convolutional neural network model to analyze facial fatigue characteristics of a user to obtain facial fatigue scores;

extracting the mel frequency cepstrum coefficient characteristic of tired voice data to obtain a mel frequency cepstrum coefficient characteristic vector;

analyzing the time sequence characteristics of the mel frequency cepstrum coefficient feature vector based on a preset time sequence analysis model to obtain the voice fatigue score, wherein the time sequence analysis model is constructed based on a long-period and short-period memory network;

based on the attention mechanism, the facial fatigue score and the voice fatigue score are fused to obtain the comprehensive fatigue score.

Further, updating the initialized motion plan using the planning feedback and/or the historical training data of the user, resulting in an updated initialized motion plan, comprising:

Based on plan feedback and historical training data, quantifying the effect of initializing the movement plan to obtain a multi-target rewarding value, wherein the multi-target rewarding value comprises at least one of a movement effect rewarding value, a movement safety rewarding value and a movement comfort rewarding value;

Obtaining plan optimization parameters based on the multi-target reward values, the plan feedback, the historical training data and a preset strategy gradient updating model, wherein the plan optimization parameters are used for indicating and adjusting at least one aspect of motion types, motion frequencies, motion time and motion heart rate intervals of user motions;

and updating the initialization motion plan based on a preset safety constraint rule and plan optimization parameters to obtain an updated initialization motion plan.

In a second aspect, the present application also provides an artificial intelligence based rehabilitation training optimizing system, comprising:

The basic physiological data at least comprises a heart injury reason, an age parameter, a heart rate parameter, a heart pressure parameter and a body quality index;

The system comprises an initial plan making module, a motion model setting module and a motion model setting module, wherein the initial plan making module is used for generating an initial motion plan of a user based on basic physiological data, wherein the initial motion plan is used for indicating the motion type, the motion frequency, the motion time and the motion heart rate interval of the motion of the user;

The real-time data acquisition module is used for acquiring moving image data, real-time heart rate parameters, tired image data and tired voice data of a user when the initialization exercise plan is executed;

the plan feedback generation module is used for quantifying the effect of the initialized motion plan on the user based on the motion image data, the real-time heart rate parameters, the tired feeling image data and the tired feeling voice data, and obtaining the plan feedback of the initialized motion plan;

And the initial plan updating module is used for updating the initial motion plan by using the plan feedback and/or the historical training data of the user to obtain an updated initial motion plan.

In a third aspect, the present application further provides a computer device, including a memory and a processor, where the memory stores a computer program, and the processor implements any of the artificial intelligence based rehabilitation training optimization methods according to the first aspect of the present application when executing the computer program.

In a fourth aspect, the present application also provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor implements any of the artificial intelligence based rehabilitation training optimization methods according to the first aspect of the present application.

The artificial intelligence-based rehabilitation training optimizing method, system, equipment and medium are used for acquiring basic physiological data of a user, wherein the basic physiological data at least comprise heart injury reasons, age parameters, heart rate parameters, heart pressure parameters and body quality indexes, generating an initialization exercise plan of the user based on the basic physiological data, wherein the initialization exercise plan is used for indicating exercise types, exercise frequencies, exercise time and exercise heart rate intervals of the user, acquiring moving image data, real-time heart rate parameters, tired image data and tired voice data of the user when the initialization exercise plan is executed, quantifying the action of the initialization exercise plan on the user based on the moving image data, the real-time heart rate parameters, the tired image data and the tired voice data to obtain plan feedback of the initialization exercise plan, and updating the initialization exercise plan by using the plan feedback and/or the historical training data of the user to obtain an updated initialization exercise plan. The rehabilitation training is realized, the individual characteristics and the real-time state change of the user can be adapted, the safety and the controllability of the training process are ensured, the personal suitability of the rehabilitation effect is improved, meanwhile, the compliance and the comfort degree of the user participating in the training are effectively enhanced through the real-time feedback and dynamic adjustment mechanism, and the heart rehabilitation training efficiency is improved.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the related art, the drawings that are required to be used in the embodiments or the related technical descriptions will be briefly described, and it is apparent that the drawings in the following description are only some embodiments of the present application, and other drawings may be obtained according to the drawings without inventive effort for those skilled in the art.

FIG. 1 is a schematic diagram of an implementation environment of an artificial intelligence based rehabilitation training optimization method according to an embodiment of the present application;

FIG. 2 is a schematic flow chart of an artificial intelligence based rehabilitation training optimization method according to an embodiment of the present application;

FIG. 3 is a schematic diagram of a rehabilitation training optimizing system based on artificial intelligence according to an embodiment of the present application;

Fig. 4 is a schematic structural diagram of a computer device for an artificial intelligence based rehabilitation training optimizing method according to an embodiment of the present application.

Detailed Description

The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

The rehabilitation training optimizing method based on artificial intelligence provided by the embodiment of the application is described with reference to fig. 1, and the implementation environment of the embodiment of the application includes a planning monitoring terminal 101, a communication network 102 and a user training terminal 103. The program monitoring terminal 101 can be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, internet of things equipment and other electronic equipment, the user training terminal 103 can be, but not limited to, various wearable equipment such as augmented reality equipment, smart bracelets, smart watches and portable electrocardiographs, programs capable of receiving information and outputting information are installed on the program monitoring terminal 101 and the user training terminal 103, the program monitoring terminal 101 can acquire physiological information of a user receiving heart rehabilitation training and images of the user training in real time through the programs, and output rehabilitation training programs and action reminding information of training, and the user training terminal 103 can receive the rehabilitation training programs through the programs and output physiological information of the user and the images of the real-time training. The planned monitoring terminal 101 communicates with the user training terminal 103 via a communication network 102, and the communication network 102 may be a wireless network or a wired network.

Further, the program monitor terminal 101 also has a function of storing and processing information, and is used for making and adjusting a rehabilitation training program according to the physiological information of the user and the image of real-time training, and storing the past training data of the user.

In an exemplary embodiment, as shown in fig. 2, there is provided an artificial intelligence based rehabilitation training optimizing method, which is illustrated by taking an application of the method to the planned monitoring terminal in fig. 1 as an example, and includes the following steps S101 to S105, wherein:

s101, acquiring basic physiological data of a user, wherein the basic physiological data at least comprise heart injury reasons, age parameters, heart rate parameters, heart pressure parameters and body quality indexes.

Specifically, the planning monitoring terminal acquires basic physiological data of a user receiving heart rehabilitation training, wherein the basic physiological data at least comprises a heart injury reason, an age parameter, a heart rate parameter, a heart pressure parameter and a body quality index. The heart injury reasons are heart health conditions of a user and can be obtained according to doctor diagnosis results, medical records or operation reports, age parameters are actual age values of the user, heart rate parameters are resting heart rate of the user and are used for representing heart beat times per minute of the user in a resting state and can be obtained according to actual conditions of the user, heart pressure parameters comprise systolic pressure and diastolic pressure values and represent pressure load of a cardiovascular system, and body mass indexes are body mass indexes calculated according to height and weight.

And S102, generating an initialized exercise plan of the user based on the basic physiological data, wherein the initialized exercise plan is used for indicating the exercise type, the exercise frequency, the exercise time and the exercise heart rate interval of the exercise of the user.

Specifically, the planning monitoring terminal determines a safe and effective exercise heart rate interval according to the age parameter and the heart rate parameter through an exercise intensity calculation formula, adjusts a maximum heart rate estimated value through the age parameter, and defines a proper heart rate training interval by combining the heart rate parameter and a preset intensity coefficient. The planning monitoring terminal adopts a machine learning model to learn the relation between the heart injury reasons, the body mass indexes and the finally selected exercise types of the past cases in the historical data, builds an exercise type classification model, and comprehensively analyzes the heart injury reasons and the body mass indexes of the user through the exercise type classification model to determine the exercise type suitable for the user. The planning monitoring terminal also uses a regression analysis model to establish the corresponding relation between the heart pressure parameter and the body mass index and the movement frequency and movement time so as to calculate the proper movement frequency and movement time of the user according to the heart pressure parameter and the body mass index of the user, and integrates the calculated four parameters of the movement type, movement frequency, movement time and movement heart rate interval to generate an initialized movement plan of the user.

S103, when the initialization exercise program is executed, the motion image data, the real-time heart rate parameters, the tired image data and the tired voice data of the user are acquired.

Specifically, when the user trains the terminal to execute the initialized exercise program, the program monitor terminal acquires the user's motion image data, real-time heart rate parameters, tired image data, and tired voice data. Illustratively, the motion image data user performs successive frames of training actions for subsequent analysis of the correctness of the motion gesture, each frame of frames being time stamped. Optionally, the real-time heart rate parameter is a heart rate parameter continuously monitoring the movement process of the user, and the real-time heart rate parameter reflects the adaptation degree of the body of the user to the movement intensity. Illustratively, the tired image data is a facial expression close-up image of the user, the tired voice data is voice information of the user when moving, and the tired image data and the tired voice data are used for presuming subjective fatigue feeling of the user.

S104, quantifying the effect of the initialized motion plan on the user based on the motion image data, the real-time heart rate parameters, the tired image data and the tired voice data, and obtaining plan feedback of the initialized motion plan.

The method comprises the steps of determining a standard gesture of user movement according to an initialized movement plan by a planning monitoring terminal, obtaining gesture accuracy scores and auditory feedback information by comparing differences of movement image data and the standard gesture, calculating the condition that the real-time heart rate parameters fall into a movement heart rate interval of the initialized movement plan according to real-time heart rate parameters to obtain heart rate coincidence degree, estimating tired feeling of a user according to tired feeling image data and tired feeling voice data by the planning monitoring terminal to obtain comprehensive fatigue scores, comprehensively estimating accuracy of user execution actions according to the obtained gesture accuracy scores and the heart rate coincidence degree by the planning monitoring terminal to obtain movement success rate, and integrating the obtained movement success rate, auditory feedback information and the comprehensive fatigue scores by the planning monitoring terminal to obtain planning feedback of the initialized movement plan.

S105, updating the initialization motion plan by using the planning feedback and/or the historical training data of the user to obtain an updated initialization motion plan.

Specifically, the plan monitoring terminal analyzes the plan feedback and the historical training data of the current training according to the plan feedback and/or the historical training data of the user, establishes a mapping relation from training effect to plan parameter adjustment, identifies parameters needing to be optimized in four aspects of parameters including motion type, motion frequency, motion time and motion heart rate interval, and obtains the optimized parameters to form an updated initialization motion plan.

According to the rehabilitation training optimization method based on artificial intelligence, an initial exercise plan is generated by acquiring basic physiological data of a user and based on multi-parameter decisions, the user state is monitored in real time through multi-mode data in the plan execution process, the training effect of a comprehensive evaluation user is accurately evaluated through quantitative analysis of actual exercise and plan difference, plan feedback is obtained, and continuous optimization is achieved by dynamically adjusting an exercise plan based on the plan feedback and historical data. The rehabilitation training is realized, the individual characteristics and the real-time state change of the user can be adapted, the safety and the controllability of the training process are ensured, the personal suitability of the rehabilitation effect is improved, meanwhile, the compliance and the comfort degree of the user participating in the training are effectively enhanced through the real-time feedback and dynamic adjustment mechanism, and the heart rehabilitation training efficiency is improved.

In one embodiment, generating an initialized motion plan for a user based on underlying physiological data includes:

S201, based on preset exercise intensity parameters, combining the age parameters and the heart rate parameters, calculating exercise heart rate interval data by using the following formula:

Wherein, the Is the exercise heart rate interval,Is a parameter of the heart rate,Is a first motion intensity parameter, which is a first motion intensity parameter,Is an age parameter, which is a function of the age,Is a second motion intensity parameter.

Specifically, the planning monitoring terminal calculates exercise heart rate interval data of the user by using a formula according to preset exercise intensity parameters and combining age parameters and heart rate parameters. Illustratively, exercise heart rate intervalsIs a heart rate range value that includes upper and lower limits for indicating the range of motion of the heart rate of the user during exercise. Optionally, heart rate parametersIs the heart rate parameter and the age parameter in the acquired user basic physiological dataIs the age parameter in the previously acquired user base physiological data. Illustratively, a first exercise intensity parameterIs the recommended lower limit coefficient of intensity level in heart rehabilitation training, and the second exercise intensity parameterIs the recommended upper limit coefficient of intensity level in heart rehabilitation training, the first exercise intensity parameterAnd a second exercise intensity parameterCan be set according to heart rehabilitation guidelines and specific conditions of users, and default settingsThe total number of the components is 0.5,0.7.

S202, inputting the cause of heart injury and the body mass index into a preset motion type classification model to obtain motion type data, wherein the motion type classification model is constructed based on a random forest algorithm.

The method comprises the steps of inputting a heart injury reason and a body quality index into a preset motion type classification model by a planning monitoring terminal to obtain motion type data, wherein the motion type classification model is constructed based on a random forest algorithm, and constructing the preset motion type classification model by the planning monitoring terminal according to the heart injury reason, the body quality index and the selected motion type of different historical patients in preset historical rehabilitation data by adopting the random forest algorithm. The preset historical rehabilitation data can be recorded according to actual work, the heart damage cause is the heart health condition type of a historical patient, the heart damage cause can be obtained according to a doctor diagnosis result, a medical record or a surgery report, the body quality index is a body quality index calculated according to the height and weight of the historical patient, and the selected exercise type can be obtained according to exercise selected by a doctor or autonomously executed by the historical patient.

S203, inputting the heart pressure parameter and the body mass index into a preset multiple linear regression model to obtain movement frequency data and movement time data.

Specifically, the planning monitoring terminal inputs the heart pressure parameter and the body quality index into a preset multiple linear regression model to obtain movement frequency data and movement time data. The preset multiple linear regression model is obtained by learning correlation modes among heart pressure parameters, body mass indexes, movement frequency and single movement time of different historical patients in preset historical movement data based on a multiple linear regression algorithm. The multiple linear regression model is a statistical analysis method, and predicts by establishing a linear relation between a plurality of independent variables and dependent variables. The preset historical exercise data can be obtained according to actual work recording, the heart pressure parameters are systolic pressure and diastolic pressure values of a historical patient and represent pressure loads of cardiovascular systems of the historical patient, the body mass index is a body mass index obtained by calculating according to height and weight of the historical patient, the exercise frequency defines the number of exercises per week, the exercise time sets the duration of single exercise, and the exercise frequency and the exercise time with the best exercise effect of the historical patient are selected to record.

S204, obtaining an initialized exercise plan based on the exercise heart rate interval data, the exercise type data, the exercise frequency data and the exercise time data.

Specifically, the plan monitoring terminal integrates exercise heart rate interval data, exercise type data, exercise frequency data and exercise time data to obtain an initialized exercise plan.

The rehabilitation training optimizing method based on artificial intelligence is characterized by determining a safe and effective exercise intensity range through formulation calculation of physiological parameters to obtain exercise heart rate interval data, selecting the most suitable exercise type according to heart conditions and body composition characteristics of a user by means of a machine learning model to obtain exercise type data, determining proper exercise frequency and time dosage according to cardiovascular states and body weight indexes by means of a regression analysis model to obtain exercise frequency data and exercise time data, and integrating the exercise heart rate interval data, the exercise type data, the exercise frequency data and the exercise time data to obtain an initialization exercise plan. The balance between the medical safety standard and the individuation requirement is realized, the subjectivity and the limitation of the traditional empirical scheme formulation are overcome, accurate, scientific, safe and reliable exercise guidance is provided for heart rehabilitation patients, a solid foundation is laid for subsequent training execution and dynamic optimization, and the heart rehabilitation training efficiency is improved.

In one embodiment, quantifying the effect of an initialized motion plan on a user based on motion image data, real-time heart rate parameters, tired image data, and tired voice data, resulting in plan feedback of the initialized motion plan, includes:

S301, based on a preset standard motion key point sequence and motion image data, quantifying differences between actual motion of a user and correct motion of the user to obtain gesture difference data, wherein the gesture difference data comprises gesture angle change data and gesture key point change data, and the standard motion key point sequence is used for representing the condition of a standard gesture when the user moves.

Specifically, the plan monitoring terminal extracts key point coordinates during user training from the moving image data, obtains key point coordinates of a standard gesture from a standard movement key point sequence according to the movement type of the initialized movement plan, and obtains gesture difference data comprising gesture angle change data and gesture key point change data by comparing the key point coordinates during user training and the key point coordinates of the standard gesture to quantify the difference between actual movement and correct movement of the user. The standard movement key point sequence is a group of predefined space coordinate data representing standard movement postures, and can be obtained according to a standard action template formulated by a professional rehabilitation doctor and used for providing a reference standard of movement postures, the posture angle change data are obtained by calculating the difference between joint angles and standard angles when a user actually moves and reflect the accuracy of a joint movement range in the movement process, and the posture key point change data are obtained by comparing the difference between body key point coordinates and standard coordinates when the user moves and represent the deviation degree of the space positions of all parts of the body. For example, the key points and joint nodes may be set according to actual work. Alternatively, the predetermined sequence of standard motion keypoints may be in the form ofWhereinIs the data of the type of movement,Is a period of a standard posture and,Is the key point coordinates at the time of the standard gesture. Illustratively, the gesture angle change data and the gesture key point change data are each provided with a time tag to characterize the user training when representing the time stamp corresponding to the time tag in the moving image data.

S302, based on gesture key point change data, difference between coordinates of key points when a user moves and coordinates of key points of a standard gesture is quantified, and a gesture accuracy score is obtained.

Specifically, the planning monitor terminal uses the following formula according to the posture key point change data: Calculating the difference between the coordinates of the key points in the movement of the user and the coordinates of the key points in the standard posture to obtain the difference degree of the training of the user, and using the following formula: and calculating to obtain a gesture accuracy score according to the difference degree trained by the user. In the foregoing equation of the present invention, Is the moment of timeIs used for the degree of difference of (2),Is the total number of key points,Is the moment of timeAny key point ofIs provided with the gesture key point change data,Is the moment of timeIs provided with a gesture accuracy score of (2),Is a scaling parameter for controlling the sensitivity of the score to the degree of variance, which can be set according to the actual work.

And S303, generating auditory feedback information based on the gesture angle change data, the gesture key point change data and the standard movement key point sequence, wherein the auditory feedback information is used for indicating a user to adjust the training gesture through the auditory information.

Specifically, the planning monitoring terminal determines deviation conditions of the gesture and the standard gesture during real-time training of the user according to the gesture angle change data, the gesture key point change data and the standard movement key point sequence, and generates auditory feedback information according to the deviation conditions so as to instruct the user to adjust the training gesture through the auditory information.

S304, predicting the fatigue degree of the user based on the tired image data and the tired voice data to obtain the comprehensive fatigue score.

The method comprises the steps that a planning monitoring terminal obtains a facial expression image sequence of a user through tired sense image data, visual features including but not limited to eye closure degree, facial muscle looseness and the like are extracted, the visual features are quantized to obtain facial fatigue scores, user voice signals collected through tired sense voice data comprise acoustic features such as speech speed, tone and pronunciation definition, the acoustic features are quantized to obtain the voice fatigue scores, and the planning monitoring terminal synthesizes the facial fatigue scores and the voice fatigue scores to obtain comprehensive fatigue scores.

S305, quantifying the heart rate normal degree of the user during exercise based on the real-time heart rate parameters and the initialized exercise plan, and obtaining the heart rate coincidence degree.

Specifically, the plan monitoring terminal counts the time proportion of the real-time heart rate parameter of the user in the exercise heart rate interval required by the exercise plan initialization during training, and obtains the heart rate coincidence degree.

S306, carrying out weighted fusion on the gesture correctness score and the heart rate coincidence degree according to preset weighted weights to obtain the motion success rate.

Specifically, the planning monitoring terminal uses the formula according to the preset weighting weight: and carrying out weighted fusion on the gesture correctness score and the heart rate coincidence degree to obtain the exercise success rate. In the foregoing equation of the present invention, Is the success rate of the movement, which is the success rate of the movement,Is the gesture correctness score weight,Is the gesture accuracy score of the gesture,Is the heart rate coincidence weight,Is heart rate compliance. The preset weighting weights comprise gesture correctness score weights and heart rate coincidence weights, and can be set according to actual work.

S307, generating planning feedback based on the auditory feedback information, the comprehensive fatigue score and the motion success rate.

Specifically, the planning monitoring terminal integrates the auditory feedback information, the comprehensive fatigue score and the exercise success rate to obtain planning feedback.

According to the rehabilitation training optimizing method based on artificial intelligence, the user action execution quality is quantified through accurate gesture difference analysis, specific hearing guidance is generated to help a user correct action deviation in real time, meanwhile, the fatigue state of the user is accurately estimated through multi-mode information fusion, the suitability of training intensity is evaluated through combination of heart rate monitoring data, and planning feedback comprehensively reflecting training effects is generated through comprehensive indexes. The comprehensive monitoring and evaluation of the training process are realized, timely correction guidance is provided, the safety, the effectiveness and the user experience of the rehabilitation training are remarkably improved through the organic combination of accurate quantitative evaluation and real-time feedback guidance, reliable technical support is provided for personalized heart rehabilitation, and the heart rehabilitation training efficiency is improved.

In one embodiment, based on a preset standard motion key point sequence and motion image data, quantifying the difference between the actual motion of the user and the correct motion of the user to obtain gesture difference data, including:

S401, extracting coordinates of key points from the moving image data based on a preset key point positioning area to obtain a real-time moving key point coordinate sequence.

Specifically, the planning monitoring terminal extracts coordinates of key points from the moving image data according to a preset key point positioning area to obtain a real-time moving key point coordinate sequence. The preset key point positioning area is a predefined human body joint point detection area in the image, and includes main motion joint positions of shoulder joint, elbow joint, hip joint, knee joint and the like, and can be set according to actual work, and the key point information is the same as the key point information of a preset standard motion key point sequence. Optionally, the planning monitoring terminal adopts a key point detection algorithm based on deep learning, performs feature extraction and joint point positioning on an input image through a convolutional neural network, and outputs continuous key point coordinate data to obtain a real-time motion key point coordinate sequence. Optionally, the real-time motion key point coordinate sequence is provided with a time tag to represent training conditions of the user when corresponding time tags in the motion image data are represented.

S402, calculating the angle change of joints when a user moves based on a real-time movement key point coordinate sequence and a preset joint node set to obtain gesture angle change data, wherein the joint node set is used for representing the coordinates of joint nodes participating in actions when the user moves.

Specifically, the planning monitoring terminal determines coordinate data of key points belonging to preset joint nodes in the real-time motion key point coordinate sequence according to the real-time motion key point coordinate sequence and a preset joint node set, calculates an angle value obtained by vector included angles formed by adjacent key points according to the coordinate data of the preset joints, combines the angle values according to a time sequence of the real-time motion key point coordinate sequence, and constructs an angle change curve of each joint in the motion process to obtain gesture angle change data. Wherein the joint node set is used for representing the coordinates of joint nodes involved in actions when a user moves, and can be set according to actual work, such as hip, knee and ankle. The planning monitoring terminal determines any timestamp in the real-time motion key point coordinate sequence according to the real-time motion key point coordinate sequence and a preset joint node setCoordinate data of key points belonging to hip in preset joint nodesCoordinate data of key points of kneeAnd coordinate data of key points of ankleUsing the formulaObtaining a first knee joint vectorUsing the formulaObtaining a second knee joint vectorReuse formulaObtaining a time stampKnee joint angle of (2)Knee joint angle within time stamp for moving image dataThe knee joint angle change data is obtained by deriving time, and is a component part of the posture angle change data, and the joint angle change data included in the posture angle change data is determined according to the number of joints included in a preset joint node set.

S403, based on the gesture angle change data, extracting the periodic characteristics of the user movement to obtain the actual movement periodic parameters.

Specifically, the planning monitoring terminal analyzes the periodic change of the joint angle through a signal processing technology according to the posture angle change data, and identifies the repeated pattern of the movement. The gait cycle is determined, for example, by periodic fluctuations in the knee angle during a walking movement. Alternatively, the dominant frequency and period length may be detected from the sequence of angular changes by means of autocorrelation analysis or fourier transform.

S404, determining a key point coordinate sequence of correct motion of a user based on the standard motion key point sequence and the actual motion period parameter, and obtaining a standard reference key point coordinate sequence.

Specifically, the planning monitoring terminal determines a key point coordinate sequence of correct motion of a user from a standard motion key point sequence according to actual motion period parameters, and obtains a standard reference key point coordinate sequence. The predetermined sequence of standard motion keypoints may be, for example, in the form ofWhereinIs the data of the type of movement,Is a period of a standard posture and,Is the key point coordinates at the time of the standard gesture. The planned monitoring terminal determines a key point coordinate sequence of correct motion of the user by matching the actual motion period parameter with the standard gesture period.

S405, calculating the difference between the real-time motion key point coordinate sequence and the standard reference key point coordinate sequence to obtain gesture key point change data.

Specifically, the planning monitoring terminal calculates Euclidean distance between each key point in the real-time motion key point coordinate sequence and the corresponding key point in the standard reference key point coordinate sequence to obtain gesture key point change data.

S406, based on the gesture key point change data and the gesture angle change data, gesture difference data is obtained.

Specifically, the planning monitor terminal combines the posture key point change data and the posture angle change data to obtain posture difference data.

The rehabilitation training optimization method based on artificial intelligence provided by the embodiment extracts space information of human body movement from visual data through a key point detection technology to obtain a real-time movement key point coordinate sequence, calculates and quantifies the movement states of all joints in the movement process through joint angles to obtain gesture angle change data, periodically analyzes and identifies rhythm characteristics of individual movement and realizes accurate alignment with a standard template, quantifies deviation degree of actual movement and standard movement through difference calculation to obtain gesture key point change data, and generates gesture difference data by integrating the gesture key point change data and the gesture angle change data. The multi-dimensional accurate quantitative analysis of the movement posture of the user is realized, the space accuracy of the movement and the matching degree of the time rhythm are considered, a reliable data base is provided for real-time movement guidance and quality evaluation, and the pertinence and the effectiveness of the action correction in the rehabilitation training are remarkably improved through accurate posture difference detection and detailed deviation analysis.

In one embodiment, generating audible feedback information based on the pose angular change data, the pose keypoint change data, and the standard motion keypoint sequence includes:

S501, calculating the difference value between the angle change of the joint and the angle change of the standard posture during the movement of the user based on the standard reference key point coordinate sequence and the posture angle change data, and obtaining a joint angle deviation sequence.

Specifically, the planning monitoring terminal calculates a standard reference key point coordinate sequence of a correct gesture of a user action according to the standard reference key point coordinate sequence, combines a standard gesture period in the standard reference key point coordinate sequence and a preset joint node set, solves to obtain standard joint angle change data in a standard gesture, and obtains a joint angle deviation sequence by calculating the difference value of the standard joint angle change data and the gesture angle change data of each joint. Illustratively, as knee angle deviation = actual knee angle in pose angle change data-standard knee angle in standard joint angle change data. Optionally, the data in the sequence of joint angular deviations is provided with a joint tag and a time tag.

S502, extracting joints and key points from the joint angle deviation sequence and posture key point change data based on a preset deviation threshold value to obtain a deviation node sequence.

The method comprises the steps of selecting a preset deviation threshold value, namely a joint angle deviation threshold value and a key point deviation threshold value, selecting key points of joints with angle change larger than the joint angle deviation threshold value from a joint angle deviation sequence according to the joint angle deviation threshold value by a planning monitoring terminal, and finding out key points with the key point deviation larger than the key point deviation threshold value from gesture key point change data according to the key point deviation threshold value to form a deviation node sequence. For example, the preset deviation threshold may be set according to the angle deviation requirement of each joint and the deviation requirement of the key point in actual operation, for example, the angle deviation threshold of the knee joint is ±5 degrees, and the angle deviation threshold of the hip joint is ±3 degrees.

S503, determining correction sequences of joints and key points in the deviation node sequence based on preset deviation weights, and obtaining a node feedback sequence.

Specifically, the planning monitoring terminal determines the correction sequence of joints and key points in the deviation node sequence according to preset deviation weights, and obtains a node feedback sequence. For example, the pre-set bias weights are used to characterize the importance of the keypoint adjustment in motion and the priority of safety, which may be set based on medical knowledge and the principles of motion biomechanics, such as the knee joint being more critical in supporting body weight than the elbow joint, and therefore being weighted higher. The planning monitoring terminal arranges the nodes according to the correction sequence from emergency to normal, and forms a node feedback sequence.

S504, converting the node feedback sequence based on a preset language database and a sequencing rule to obtain a language feedback text, and generating auditory feedback information based on the language feedback text.

Specifically, the planning monitoring terminal forms a language feedback text from the node feedback sequence according to a preset language database and a sequencing rule, and generates auditory feedback information according to the language feedback text. The language database is a pre-stored voice command database and comprises standard correction statement templates aiming at various deviation types, the ordering rule prescribes the organization mode of a plurality of pieces of feedback information, such as descending order of priority, merging similar items and the like, and the language database can be set according to actual work.

The rehabilitation training optimization method based on the artificial intelligence provided by the embodiment comprises the steps of calculating and quantifying specific differences between user actions and standard postures through joint angle deviation to obtain a joint angle deviation sequence, identifying key problem positions needing to be corrected through a threshold screening mechanism to obtain a deviation node sequence, determining an optimal feedback sequence by combining preset deviation weights, and finally generating auditory feedback information through a language conversion technology. The method has the advantages that full-automatic processing from motion data detection to voice guidance generation is realized, a user is helped to correct action deviation in time through real-time voice guidance, quality and efficiency of rehabilitation training are effectively improved, cognitive burden of the user is lightened, and pertinence and effectiveness of action correction in rehabilitation training are remarkably improved.

In one embodiment, predicting the fatigue of the user based on the tired image data and the tired voice data, resulting in a composite fatigue score, comprises:

S601, analyzing facial fatigue characteristics of a user by adopting a preset convolutional neural network model according to tired feel image data to obtain facial fatigue scores.

Specifically, the planning monitoring terminal analyzes facial fatigue characteristics of a user by adopting a preset convolutional neural network model to obtain facial fatigue scores. The preset convolutional neural network model can be constructed according to the requirement of facial feature extraction in actual work, so that multi-layer feature extraction is performed on an input facial image, visual features related to fatigue, such as eyelid closing frequency, pupil size change, mouth angle sagging degree and the like, are identified, a facial fatigue score between 0 and 1 is output, and the higher the numerical value is, the more serious the fatigue degree is.

S602, extracting the characteristic of the mel frequency cepstrum coefficient of the tired voice data to obtain a characteristic vector of the mel frequency cepstrum coefficient.

The method comprises the steps of carrying out pre-emphasis, framing and windowing on tired feeling voice data by a planning monitoring terminal to obtain preprocessed voice data, converting the preprocessed voice data in a time domain into frequency domain representation through fast Fourier transform to obtain frequency domain voice data, simulating the frequency perception characteristic of human ears by a Mel scale filter bank, and obtaining Mel frequency cepstrum coefficient feature vectors through logarithmic operation and discrete cosine transform. The mel frequency cepstrum coefficient is a voice characteristic representation method based on the auditory characteristics of human ears, and the characteristics of voice signals are extracted by simulating the perception sensitivity of human ears to sounds with different frequencies. Illustratively, the mel-frequency cepstral coefficient feature vector can reflect fatigue-induced changes in speech characteristics, such as slow speech, pitch changes, blurry pronunciation, etc.

S603, analyzing time sequence characteristics of the mel frequency cepstrum coefficient feature vector based on a preset time sequence analysis model to obtain the voice fatigue score, wherein the time sequence analysis model is constructed based on a long-term and short-term memory network.

Specifically, the planning monitoring terminal inputs the mel frequency cepstrum coefficient feature vector into a preset time sequence analysis model to obtain the voice fatigue score. The long-term memory network is a special cyclic neural network structure, and long-term dependency relationship in sequence data can be effectively learned and memorized by introducing an input gate, a forgetting gate and an output gate mechanism. The preset time sequence analysis model can be constructed by adopting a long-short-period memory network according to the time length and other characteristics of tired feeling voice data in actual work, so as to analyze the change mode of voice characteristics in the time dimension through a gating mechanism in the long-short-period memory network, capture the voice dynamic characteristic change caused by fatigue, such as the smoothness of intonation change, the stability of speech speed and the like, and output a voice fatigue score between 0 and 1 to reflect the fatigue degree detected from a voice signal.

S604, based on the attention mechanism, fusing the facial fatigue score and the voice fatigue score to obtain the comprehensive fatigue score.

Specifically, the planning monitoring terminal performs weighted fusion on the facial fatigue score and the voice fatigue score according to the attention mechanism to obtain the comprehensive fatigue score. The attention mechanism dynamically calculates the fusion weight of the two scores from the signal-to-noise ratio, the preset credibility and the default value of the tired image data and the tired voice data, and fuses the facial fatigue score and the voice fatigue score according to the obtained fusion weight to obtain the comprehensive fatigue score.

The rehabilitation training optimization method based on artificial intelligence is used for accurately extracting visual fatigue characteristics from face images through a convolutional neural network to obtain face fatigue scores, capturing acoustic fatigue characteristics from voice signals through mel frequency cepstrum coefficient analysis to obtain mel frequency cepstrum coefficient characteristic vectors, modeling time sequence change modes of the voice characteristics through a long-term and short-term memory network to obtain the voice fatigue scores, and adaptively fusing visual and auditory mode assessment results through an attention mechanism to obtain comprehensive fatigue scores. The limitation of single-mode evaluation is overcome, the real fatigue state of the user is comprehensively and accurately reflected, and the intelligent level and the safety guarantee capability of the rehabilitation training system are remarkably improved.

In one embodiment, updating the initialized motion plan using the planning feedback and/or historical training data of the user, resulting in an updated initialized motion plan, comprising:

s701, based on plan feedback and historical training data, initializing the effect of the exercise plan in a quantized mode to obtain a multi-target rewarding value, wherein the multi-target rewarding value comprises at least one of an exercise effect rewarding value, an exercise safety rewarding value and an exercise comfort rewarding value.

Specifically, the plan monitoring terminal calculates a sports effect rewarding value, a sports safety rewarding value and a sports comfort rewarding value respectively according to hearing feedback information, comprehensive fatigue scores and a sports success rate in plan feedback and in combination with the plan feedback of historical training included in the historical training data, so as to form a multi-target rewarding value. Illustratively, the athletic performance benefit reward value is calculated using the following equation: Wherein Is a sports effect bonus value; Is a first weight for calculating the influence of the difference between the actual motion success rate and the target motion success rate on the motion effect rewarding value, Is the second weight for calculating the influence of the difference value of the current exercise success rate and the last exercise success rate on the exercise effect rewarding value, the first weight and the second weight can be set according to actual work, and the sum is 1; Is the success rate of the exercise; is the target movement success rate, can set the training requirement according to the actual work, The difference between the current training and the last training can be obtained according to the historical training data. Optionally, the sports safety prize value is calculated using the following formula:, is the value of the exercise safety reward, Is the coincidence of the heart rate and the degree of compliance,Is the total fatigue score of the product,Is the threshold value for fatigue and,Is a third weight that is a third weight,Is the fourth weight. Illustratively, the sports comfort prize value is calculated using the following formula:, is a value of a sports comfort prize, Is a fifth weight that is a third weight,Is the total fatigue score of the product,Is a sixth weight that is to be applied,The training time length can be obtained according to the initialization exercise program,Is the historical average training time length, and can be obtained according to the historical training data. Alternatively, the historical training data may be derived from a user's historical training situation record.

And S702, updating a model based on the multi-target rewarding value, the planning feedback, the historical training data and the preset strategy gradient to obtain planning optimization parameters, wherein the planning optimization parameters are used for indicating and adjusting at least one aspect of the movement type, the movement frequency, the movement time and the movement heart rate interval of the movement of the user.

Specifically, the plan monitoring terminal inputs plan feedback and historical training data into a preset strategy gradient update model to obtain plan optimization parameters. The method comprises the steps of obtaining a strategy gradient updating model, wherein the strategy gradient updating model is preset, a multi-target rewarding value is used as an optimization target, planning feedback and historical training data are used as state input, the gradient direction of a strategy function is calculated, model parameters are updated, and according to the learned optimal mapping relation from a user state to a planning adjustment action, the planning optimization parameters are specific adjustment suggestions output by the model, and comprise change indication of a movement type, adjustment amplitude of movement frequency, increment and decrement of movement time and upper limit and lower limit change values of a movement heart rate interval, and the parameters provide specific operation guidance for subsequent planning updating. The preset strategy gradient updating model can be constructed by adopting a strategy gradient updating algorithm according to historical training data.

S703, updating the initialization motion plan based on the preset safety constraint rules and the plan optimization parameters to obtain an updated initialization motion plan.

Specifically, the plan monitoring terminal screens the plan optimization parameters according to preset safety constraint rules, adjusts parameters which do not accord with the preset safety constraint rules, and updates the initialization motion plan by using the adjusted plan optimization parameters to obtain an updated initialization motion plan. The preset safety constraint rules are used for limiting the safety range of the adjustment of the planning optimization parameters, and can be set according to the heart rehabilitation medical standard, including but not limited to the safety range of exercise heart rate, the longest time limit of single training and the exercise types corresponding to different heart injury types, the planning monitoring terminal maps the planning optimization parameters to specific planning parameters conforming to the safety rules, if the heart rate interval in the planning optimization parameters exceeds the safety range, the planning optimization parameters are truncated to the allowable maximum or minimum value, and when the suggested exercise types are not matched with the heart condition of a user, the planning optimization parameters are replaced by safe alternative exercise types.

The rehabilitation training optimization method based on the artificial intelligence provided by the embodiment comprehensively measures the comprehensive performance of a training program in the aspects of effect, safety and comfort through multi-dimensional quantitative evaluation to obtain a multi-target rewarding value, intelligently generates a program optimization direction based on historical data and real-time feedback through a reinforcement learning algorithm to obtain a program optimization parameter, and finally ensures that all optimization adjustment is within a medical allowable range through a preset safety constraint rule to update an initialization motion program to obtain the updated initialization motion program. Continuous self-optimization and personalized adaptation of a training plan are realized, the risk controllability of the training process is ensured through safety constraint, so that the rehabilitation training can be dynamically adjusted along with the change of the state of a user, the limitation that the traditional static plan cannot adapt to individual progress is avoided, potential safety hazards possibly caused by excessive optimization are prevented, and the pertinence and the effectiveness of action correction in the rehabilitation training are remarkably improved.

The artificial intelligence-based rehabilitation training optimizing method, system, equipment and medium are used for acquiring basic physiological data of a user, wherein the basic physiological data at least comprise heart injury reasons, age parameters, heart rate parameters, heart pressure parameters and body quality indexes, generating an initialization motion plan of the user based on the basic physiological data, wherein the initialization motion plan is used for indicating motion types, motion frequencies, motion time and motion heart rate intervals of the user, acquiring motion image data, real-time heart rate parameters, tired image data and tired voice data of the user when the initialization motion plan is executed, quantifying the actions of the initialization motion plan on the user based on the motion image data, the real-time heart rate parameters, the tired image data and the tired voice data to obtain plan feedback of the initialization motion plan, and updating the initialization motion plan by using the plan feedback and/or the historical training data of the user to obtain the updated initialization motion plan. The rehabilitation training is realized, the individual characteristics and the real-time state change of the user can be adapted, the safety and the controllability of the training process are ensured, the personal suitability of the rehabilitation effect is improved, meanwhile, the compliance and the comfort degree of the user participating in the training are effectively enhanced through the real-time feedback and dynamic adjustment mechanism, and the heart rehabilitation training efficiency is improved.

It should be understood that, although the steps in the flowcharts related to the embodiments described above are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.

Based on the same inventive concept, the embodiment of the application also provides an artificial intelligence-based rehabilitation training optimizing system for realizing the artificial intelligence-based rehabilitation training optimizing method. The implementation of the solution provided by the system is similar to the implementation described in the above method, so the specific limitations in one or more embodiments of the artificial intelligence based rehabilitation training optimization system provided below may be referred to above for limitations of the artificial intelligence based rehabilitation training optimization method, and are not repeated here.

In one exemplary embodiment, as shown in FIG. 3, an artificial intelligence based rehabilitation training optimization system 200 is provided, comprising:

the basic data acquisition module 201 is configured to acquire basic physiological data of a user, where the basic physiological data at least includes a cause of heart injury, an age parameter, a heart rate parameter, a heart pressure parameter, and a body mass index;

An initial plan making module 202 for generating an initial exercise plan for the user based on the basic physiological data, the initial exercise plan being used for indicating an exercise type, an exercise frequency, an exercise time and an exercise heart rate interval of the exercise of the user;

A real-time data acquisition module 203 for acquiring moving image data, real-time heart rate parameters, tired image data, and tired voice data of a user when performing an initialization exercise program;

The plan feedback generation module 204 is configured to quantify an effect of the initialized motion plan on the user based on the motion image data, the real-time heart rate parameter, the tired feeling image data and the tired feeling voice data, and obtain plan feedback of the initialized motion plan;

The initial plan updating module 205 is configured to update the initial motion plan using the plan feedback and/or the historical training data of the user, and obtain an updated initial motion plan.

Further, the initial planning module is further configured to:

Further, the plan feedback generation module includes:

the gesture difference calculation unit is used for quantifying the difference between the actual motion of the user and the correct motion of the user based on a preset standard motion key point sequence and motion image data to obtain gesture difference data, wherein the gesture difference data comprises gesture angle change data and gesture key point change data;

A gesture difference quantification unit for quantifying the difference between the coordinates of the key points when the user moves and the coordinates of the key points of the standard gesture based on the gesture key point change data, so as to obtain a gesture accuracy score;

the system comprises an auditory information generating unit, an auditory feedback information generating unit and a training unit, wherein the auditory information generating unit is used for generating auditory feedback information based on gesture angle change data, gesture key point change data and standard motion key point sequences;

the comprehensive fatigue calculation unit is used for predicting the fatigue degree of the user based on the tired image data and the tired voice data to obtain comprehensive fatigue scores;

the heart rate normal quantifying unit is used for quantifying the heart rate normal degree of a user during exercise based on the real-time heart rate parameters and the initialized exercise plan, and obtaining the heart rate coincidence degree;

The exercise comprehensive calculation unit is used for carrying out weighted fusion on the gesture correctness score and the heart rate coincidence degree according to preset weighted weights to obtain an exercise success rate;

and the plan feedback generation unit is used for generating plan feedback based on the auditory feedback information, the comprehensive fatigue score and the exercise success rate.

Further, the posture difference calculating unit is further configured to:

further, the auditory information generating unit is further configured to:

Further, the comprehensive fatigue calculation unit is further configured to:

Further, the initial plan update module is further configured to:

In one embodiment, as provided in FIG. 4, a computer device includes:

At least one processor 301, and a memory 302 communicatively coupled to the at least one processor 301, the memory storing application code executable by the at least one processor, the application code being executable by the at least one processor to enable the at least one processor to perform the artificial intelligence based rehabilitation training optimization method as described above.

The computer device may also include a sensor 303.

The processor 301, memory 302, and sensor 303 may be connected by a bus or other means, for example.

In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when executed by a processor, implements the steps of the method embodiments described above.

For the device embodiments, reference is made to the description of the method embodiments for the relevant points, since they essentially correspond to the method embodiments. The above-described apparatus embodiments are merely illustrative, wherein the components illustrated as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the objectives of the disclosed solution. Those of ordinary skill in the art will understand and implement the present invention without undue burden.

The above examples merely represent a few implementations of the examples of the present application, which are described in more detail and are not to be construed as limiting the scope of the claims of the examples. It should be noted that it will be apparent to those skilled in the art that various modifications and improvements can be made to the present application without departing from the spirit of the embodiments of the application.

Claims

1. An artificial intelligence-based rehabilitation training optimization method, which is characterized by comprising the following steps:

Generating an initialization exercise plan of the user based on the basic physiological data, wherein the initialization exercise plan is used for indicating the exercise type, the exercise frequency, the exercise time and the exercise heart rate interval of the exercise of the user;

Acquiring moving image data, real-time heart rate parameters, tired image data and tired voice data of the user while executing the initialization exercise program;

quantifying the effect of the initialization exercise plan on the user based on the motion image data, the real-time heart rate parameter, the tired image data and the tired voice data, and obtaining plan feedback of the initialization exercise plan;

And updating the initialization motion plan by using the plan feedback and/or the historical training data of the user to obtain the updated initialization motion plan.

2. The method of claim 1, wherein the generating the user's initialized motion plan based on the base physiological data comprises:

Based on preset exercise intensity parameters, the following formulas are used for calculating exercise heart rate interval data by combining the age parameters and the heart rate parameters:

Inputting the cause of the heart injury and the body mass index into a preset motion type classification model to obtain motion type data, wherein the motion type classification model is constructed based on a random forest algorithm;

Inputting the heart pressure parameter and the body mass index into a preset multiple linear regression model to obtain movement frequency data and movement time data;

and obtaining the initialization exercise plan based on the exercise heart rate interval data, the exercise type data, the exercise frequency data and the exercise time data.

3. The method of claim 1, wherein quantifying the effect of the initialized motion plan on the user based on the motion image data, real-time heart rate parameters, tired image data, and tired voice data, resulting in a plan feedback for the initialized motion plan, comprises:

Based on a preset standard motion key point sequence and the motion image data, quantifying the difference between the actual motion of the user and the correct motion of the user to obtain gesture difference data, wherein the gesture difference data comprises gesture angle change data and gesture key point change data;

Generating auditory feedback information based on the gesture angle change data, the gesture key point change data and the standard motion key point sequence, wherein the auditory feedback information is used for indicating the user to adjust training gesture through the auditory information;

based on the real-time heart rate parameters and the initialized exercise plan, quantifying the heart rate normal degree of the user during exercise, and obtaining heart rate coincidence degree;

According to preset weighting weights, carrying out weighting fusion on the gesture correctness scores and the heart rate coincidence degree to obtain a motion success rate;

The planned feedback is generated based on the auditory feedback information, the integrated fatigue score, and the motion success rate.

4. A method according to claim 3, wherein said quantifying the difference between the actual motion of the user and the correct motion of the user based on a preset sequence of standard motion keypoints and the motion image data, to obtain gesture difference data, comprises:

Extracting coordinates of key points from the moving image data based on a preset key point positioning area to obtain a real-time moving key point coordinate sequence;

based on the gesture angle change data, extracting periodic characteristics of the user movement to obtain actual movement periodic parameters;

Determining a key point coordinate sequence of correct motion of the user based on the standard motion key point sequence and the actual motion period parameter to obtain a standard reference key point coordinate sequence;

calculating the difference between the real-time motion key point coordinate sequence and the standard reference key point coordinate sequence to obtain the gesture key point change data;

and obtaining the gesture difference data based on the gesture key point change data and the gesture angle change data.

5. The method of claim 4, wherein the generating audible feedback information based on the posture angle change data, the posture key point change data, and the standard motion key point sequence comprises:

extracting the joints and the key points from the joint angle deviation sequence and the posture key point change data based on a preset deviation threshold value to obtain a deviation node sequence;

Determining correction sequences of the joints and the key points in the deviation node sequence based on preset deviation weights to obtain a node feedback sequence;

And converting the node feedback sequence based on a preset language database and a sequencing rule to obtain a language feedback text, and generating the auditory feedback information based on the language feedback text.

6. A method according to claim 3, wherein predicting the degree of fatigue of the user based on the tired image data and tired voice data, resulting in a composite fatigue score, comprises:

analyzing facial fatigue characteristics of the user by adopting a preset convolutional neural network model according to the tired feel image data to obtain facial fatigue scores;

Extracting the mel frequency cepstrum coefficient characteristic of the tired feeling voice data to obtain a mel frequency cepstrum coefficient characteristic vector;

analyzing the time sequence characteristics of the mel frequency cepstrum coefficient feature vector based on a preset time sequence analysis model to obtain a voice fatigue score, wherein the time sequence analysis model is constructed based on a long-term and short-term memory network;

and based on an attention mechanism, fusing the facial fatigue score and the voice fatigue score to obtain the comprehensive fatigue score.

7. A method according to claim 3, wherein updating the initialized motion plan using the plan feedback and/or the user's historical training data, resulting in an updated initialized motion plan, comprises:

based on the plan feedback and the historical training data, quantifying the effect of initializing the movement plan to obtain a multi-target rewarding value, wherein the multi-target rewarding value comprises at least one of a movement effect rewarding value, a movement safety rewarding value and a movement comfort rewarding value;

Obtaining a plan optimization parameter based on the multi-objective reward value, the plan feedback, the historical training data and a preset strategy gradient update model, wherein the plan optimization parameter is used for indicating and adjusting at least one aspect of the movement type, the movement frequency, the movement time and the movement heart rate interval of the movement of the user;

And updating the initialization motion plan based on a preset safety constraint rule and the plan optimization parameter to obtain the updated initialization motion plan.

8. An artificial intelligence based rehabilitation training optimization system, the system comprising:

the initial plan making module is used for generating an initial exercise plan of the user based on the basic physiological data, wherein the initial exercise plan is used for indicating the exercise type, the exercise frequency, the exercise time and the exercise heart rate interval of the exercise of the user;

The real-time data acquisition module is used for acquiring moving image data, real-time heart rate parameters, tired image data and tired voice data of the user when the initialization exercise plan is executed;

The plan feedback generation module is used for quantifying the effect of the initialization exercise plan on the user based on the motion image data, the real-time heart rate parameters, the tired image data and the tired voice data, and obtaining plan feedback of the initialization exercise plan;

And the initial plan updating module is used for updating the initial motion plan by using the plan feedback and/or the historical training data of the user to obtain the updated initial motion plan.

9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the computer program is executed.

10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.