Facial rehabilitation exercise training method and computer program product

Abstract

The present application relates to a kind of facial rehabilitation action training method and computer program product, belong to medical rehabilitation technical field.The method is by obtaining user video stream, extracts facial key point and generates facial expression parameter;From facial expression parameter, determine the target parameter associated with preset rehabilitation action, and the target parameter is compared with preset evaluation standard, obtain evaluation result;According to evaluation result, execute feedback operation and / or adjust training target.The method can be for closed-lip action, cheek-puffing action, tongue muscle stretch action, tongue muscle around action and neck massage action etc. many kinds of rehabilitation training are accurately evaluated, by monitoring the duration of facial expression parameter, repeat frequency, facial key point displacement etc. index, combined with training management model generates individualized training target and evaluation threshold.The present application provides a kind of intelligent, individualized facial rehabilitation training solution, can effectively improve the accuracy and effectiveness of facial muscle function rehabilitation training.

Description

Technical Field

[0001] This invention relates to the field of rehabilitation training technology, specifically to a facial rehabilitation exercise training method and computer program product. Background Technology

[0002] With the development of medical technology, facial rehabilitation training plays an increasingly important role in the rehabilitation treatment of diseases such as nasopharyngeal carcinoma and Parkinson's disease. Nasopharyngeal carcinoma is a common malignant tumor of the head and neck, and radiotherapy is one of its main treatment methods. However, patients often experience a series of complications after radiotherapy, including difficulty opening the mouth, swallowing disorders, speech disorders, and facial muscle fibrosis, which seriously affect their quality of life. Therefore, postoperative rehabilitation training is of great significance for improving patients' function and enhancing their quality of life. Facial rehabilitation training mainly helps patients restore facial muscle function and improve facial expressions and oral function through specific facial movement exercises. However, traditional facial rehabilitation training methods have many limitations and are difficult to meet the needs of modern rehabilitation medicine.

[0003] However, existing rehabilitation training models have the following technical problems: First, traditional rehabilitation assessments mainly rely on the subjective observation and experience of medical staff, lacking precise and quantitative methods for movement assessment. Different medical staff may have different understandings of "whether the movement is standard," leading to poor consistency and repeatability of assessments, and failing to provide objective data support for the rehabilitation process. Second, rehabilitation training for nasopharyngeal carcinoma is a long-term process, and patients need to complete the training independently at home most of the time. However, without the professional guidance and supervision of the hospital, patients find it difficult to ensure the standardization of their movements, which may lead to poor rehabilitation results or even secondary injuries due to incorrect movements. Third, existing computer vision and motion capture technologies are mainly designed for general scenarios such as fitness and sports. The number and accuracy of their key points are insufficient to capture the fine facial movements involved in nasopharyngeal carcinoma rehabilitation training, such as lip closure, cheek puffing, and tongue muscle extension, and cannot meet the needs of rehabilitation assessment for high-precision, fine-grained analysis. Summary of the Invention

[0004] To address the technical problems in existing technologies, such as the high subjectivity of nasopharyngeal carcinoma rehabilitation training assessment, the lack of dynamic adjustment of training programs, the lack of real-time feedback for patients' home training, and the limited accuracy of facial analysis, this invention provides a rehabilitation training method that achieves standardized and quantitative assessment of rehabilitation training movements, improves patient training compliance, enables in-depth personalization of training programs, enhances detection accuracy, and improves patients' quality of life.

[0005] The technical solution adopted by this invention to solve its technical problem is: to provide a facial rehabilitation exercise training method, including the following steps: acquiring a user's video stream; extracting facial key points from the video stream and generating facial expression parameters based on the facial key points; determining target parameters associated with preset rehabilitation exercises from the facial expression parameters, and comparing the target parameters with preset evaluation criteria to obtain the evaluation result of the preset rehabilitation exercise; and performing feedback operations and / or adjusting subsequent training objectives based on the evaluation result.

[0006] Preferably, the step of obtaining the evaluation result further includes: monitoring whether the duration and / or number of repetitions of facial expression parameters associated with the preset rehabilitation action meet preset standards.

[0007] Preferably, the step of obtaining the evaluation result further includes: monitoring whether the displacement of facial key points associated with the preset rehabilitation action meets the preset standard.

[0008] Preferably, when the preset rehabilitation action is a lip-closing action, the step of obtaining the evaluation result includes: acquiring facial expression parameters related to lip muscle contraction, the parameters including lip closure parameters, upper lip curling parameters and lower lip curling parameters; determining whether the values ​​of the lip closure parameters, upper lip curling parameters and lower lip curling parameters meet a first preset threshold, and whether the duration of meeting the first preset threshold meets a first preset duration requirement.

[0009] Preferably, the step of obtaining the evaluation result further includes: obtaining the real-time distance between the key points of the upper lip and the key points of the lower lip; comparing the real-time distance with the reference distance, and using the comparison result to help determine whether there is an air leakage phenomenon.

[0010] Preferably, when the preset rehabilitation action is a cheek-puffing action, the step of obtaining the evaluation result includes: obtaining lip-puffing parameters and determining whether they meet a second preset threshold; in response to the lip-puffing parameters meeting the second preset threshold, monitoring the displacement of key points in the cheek area relative to the initial position; determining whether there is a first stage where the displacement exceeds a third preset threshold, and a second stage where the displacement returns to the initial position, and whether the total duration of the complete process consisting of the first stage and the second stage meets the second preset duration requirement.

[0011] Preferably, when the preset rehabilitation action is a tongue muscle extension action, the step of obtaining the evaluation result includes: acquiring mandibular movement parameters, the mandibular movement parameters including the degree of mandibular opening, the degree of mandibular leftward movement, and the degree of mandibular rightward movement; determining whether the parameters of the mandibular movement parameters corresponding to the current extension direction meet a fourth preset threshold, and whether the number of repetitions that meet the fourth preset threshold meets a first preset quantity requirement; and acquiring head posture parameters, and determining whether the angle changes of the head posture parameters in the pitch, yaw, and roll directions during the tongue muscle extension process meet a fifth preset threshold.

[0012] Preferably, when the preset rehabilitation action is a tongue muscle circling action, the step of obtaining the evaluation result includes: acquiring lip state parameters, the lip state parameters including lip pursing parameters and / or upper lip curling parameters; determining whether the lip state parameters meet a sixth preset threshold; in response to the lip state parameters meeting the sixth preset threshold, monitoring the periodic displacement changes of key points around the mouth caused by tongue tip movement; determining the number of circling cycles based on the number of periodic displacement changes, and determining whether the number of circling cycles meets a second preset quantity requirement.

[0013] Preferably, when the preset rehabilitation action is a neck massage action, the steps for obtaining the evaluation result include: acquiring key points of the hand and tracking their movement trajectory in real time; determining whether the difference between the endpoint and the starting point of the movement trajectory meets a seventh preset threshold, and whether the shape of the movement trajectory meets a preset closed loop requirement; constructing a three-dimensional model of the neck based on facial key points, and delineating a prohibited contact area containing a preset danger zone; and determining that the current neck massage action does not meet safety standards in response to any of the key points of the hand entering the prohibited contact area.

[0014] Preferably, the steps for adjusting subsequent training objectives include: inputting the user's training data into the training management model; and obtaining personalized training objectives and / or evaluation thresholds for the user output by the training management model.

[0015] Preferably, the training management model is trained through the following steps: obtaining a pre-trained base model; obtaining training data sequences and corresponding label data of sample users, wherein the label data is the optimal training target associated with the rehabilitation effect of the sample users; inputting the training data sequence into the base model to predict the training target; comparing the predicted training target with the label data, and fine-tuning the base model according to the comparison result until the training completion condition is met.

[0016] The beneficial effects of this invention are as follows: 1. Standardizing and quantifying the assessment of rehabilitation movements. By deeply integrating 3D facial key points and facial expression parameters, subjective experience is transformed into objective data, establishing a "movement profile" for each patient, greatly improving the accuracy and consistency of the assessment. 2. Providing real-time movement feedback to improve patient training compliance. The system can capture movement data in real time and compare it with preset standards. Once non-standard movements are detected, the system immediately prompts the patient through visual and auditory means, achieving a closed loop of "instant feedback - instant correction - instant encouragement." 3. Enhancing rehabilitation effects through personalized training. A large-scale training management model is introduced, dynamically adjusting the training goals for the next day based on the patient's historical training data, ensuring that the training difficulty always matches the patient's current ability, achieving a personalized rehabilitation plan tailored to each individual.

[0017] By utilizing the facial 3D key point detection module and facial expression parameters of the MediaPipe library, standardized and quantitative assessment of nasopharyngeal carcinoma rehabilitation training movements has been achieved. Complex facial movements are transformed into quantifiable and comparable data indicators, completely eliminating the subjectivity and ambiguity of traditional assessments. A real-time movement feedback mechanism allows patients to correct errors promptly, ensuring training quality and significantly improving patient compliance. Deep learning analysis of patients' historical training data using a large-scale training management model dynamically optimizes training goals and assessment standards, achieving deep personalization of rehabilitation training programs and significantly improving rehabilitation outcomes. The integration of a dual verification mechanism of 3D spatial positioning and muscle activity quantification improves the robustness and accuracy of detection, adapting to individual differences among patients and ensuring accuracy in various complex environments. Through precise, dynamic, and personalized training guidance, patients effectively improve sequelae such as difficulty opening their mouths and swallowing disorders, enhance eating, speech, and social abilities, and improve their quality of life. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of this specification and form part of this specification, illustrate exemplary embodiments and are used to explain this specification, but do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating the facial rehabilitation exercise training method in an embodiment of the present invention. Detailed Implementation

[0019] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings and embodiments. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] like Figure 1 As shown, this embodiment provides a facial rehabilitation exercise training method 100, which includes the following steps: 110 Acquiring the user's video stream.

[0021] By capturing real-time video data streams of the user's face using a camera, the front-facing camera of the patient's terminal device (such as a smartphone, tablet, or computer) can be utilized to ensure that the user's face is clearly visible in the video frame for subsequent processing and analysis. The video stream is typically captured at a frame rate of 30 frames per second with a resolution of at least 720p to ensure that subtle facial expressions are accurately captured.

[0022] Step 120 extracts facial key points from the video stream and generates facial expression parameters based on the facial key points.

[0023] This embodiment uses the MediaPipe framework to process the video stream. The MediaPipe Face Mesh module can detect and track 468 3D key points on the user's face from the real-time video stream of a monocular camera at high frame rate and high precision, forming a detailed 3D mesh model that changes in real time with facial expressions. These key points include the iconic locations of facial areas such as eyes, eyebrows, nose, lips, and chin. Based on this, the system further extracts the MediaPipeBlendshapes parameters, which are the intensity values ​​of 52 standard facial expression units, such as lip closure parameters, upper lip curling parameters, lower lip curling parameters, and lip pouting parameters. Each of these parameters represents intensity with a value between 0 and 1. These expression units accurately correspond to the main muscle activities of the human face; for example, MOUTH_CLOSED corresponds to closed mouth, MOUTH_PUCKER corresponds to pouting, and JAW_OPEN corresponds to open mouth.

[0024] Step 130 determines the target parameters associated with the preset rehabilitation movements from the facial expression parameters, and compares the target parameters with the preset assessment criteria to obtain the assessment results of the preset rehabilitation movements.

[0025] Based on the type of rehabilitation movement being trained, the system filters out target parameters related to that movement from all facial expression parameters. For example, for lip-closing movements, the system focuses on lip closure parameters, upper lip curling parameters, and lower lip curling parameters; for cheek-puffing movements, the system focuses on lip pursing parameters and the displacement of key points in the cheek area. The system compares these target parameters with pre-set evaluation criteria, including parameter thresholds, movement duration, and repetition count requirements. Through comparison, the system obtains an evaluation result for the current rehabilitation movement, such as "movement completion rate 80%" or "movement duration insufficient."

[0026] During the assessment, the system monitors whether the duration and repetition count of facial expression parameters associated with preset rehabilitation movements meet preset standards. For example, for lip-closing movements, the system calculates whether the duration for which the user keeps their lips closed meets the preset requirement of 2 seconds; for tongue muscle circling movements, the system calculates whether the number of circling cycles completed by the user meets the preset requirement of 5 repetitions.

[0027] Simultaneously, the system also monitors whether the displacement of key facial points associated with preset rehabilitation movements meets preset standards. For example, for cheek puffing movements, the system monitors whether the displacement of key points in the cheek area relative to their initial positions reaches a preset threshold; for tongue muscle extension movements, the system monitors the displacement of key points in the mandible.

[0028] When the preset rehabilitation movement is a lip-closing movement, the assessment results are obtained as follows: First, facial expression parameters related to lip muscle contraction are acquired, including lip closure parameters (MOUTH_CLOSED), upper lip curling parameters (MOUTH_ROLL_UPPER), and lower lip curling parameters (MOUTH_ROLL_LOWER). Then, it is determined whether the values ​​of these parameters meet the first preset threshold (e.g., lip closure parameter ≥ 0.7, upper lip curling parameter ≥ 0.3, lower lip curling parameter ≥ 0.3), and whether the duration of meeting the first preset threshold meets the first preset duration requirement (e.g., ≥ 2 seconds). In addition, the real-time distance between the key points of the upper lip and the key points of the lower lip is acquired and compared with the reference distance (usually the lip distance when the user is relaxed). The comparison results are used to help determine whether there is air leakage. If the ratio of the real-time distance to the reference distance is less than 0.2, it is judged as good closure; if the ratio is between 0.2 and 0.4, it is judged as mild air leakage; if the ratio is greater than 0.4, it is judged as severe air leakage.

[0029] When the preset rehabilitation movement is cheek puffing, the assessment results are obtained as follows: First, the lip pucker parameter (MOUTH_PUCKER) is acquired, and it is determined whether it meets the second preset threshold (e.g., ≥0.5). If the lip pucker parameter meets the second preset threshold, the system monitors the displacement of key points in the cheek area relative to the initial position. Then, it determines whether there is a first stage (puffing stage) where the displacement exceeds the third preset threshold (e.g., ≥5mm), and a second stage where the displacement returns to the initial position (release stage), and whether the total duration of the complete process consisting of the first and second stages meets the second preset duration requirement (e.g., ≥3 seconds). Only when all three conditions are met simultaneously—"lip closure preparation state meets the standard," "movement duration is sufficient," and "includes a complete puffing-release cycle"—is the system considered a valid cheek puffing movement.

[0030] When the preset rehabilitation movement is tongue muscle extension, the assessment results are obtained as follows: First, the mandibular movement parameters are acquired, including the degree of mandibular opening (JAW_OPEN), the degree of mandibular leftward movement (JAW_LEFT), and the degree of mandibular rightward movement (JAW_RIGHT). Then, it is determined whether the parameters of the mandibular movement parameters corresponding to the current extension direction meet the fourth preset threshold (e.g., when extending to the left, the degree of mandibular leftward movement is ≥0.4), and whether the number of repetitions that meet the fourth preset threshold meets the first preset quantity requirement (e.g., ≥5 times). At the same time, the system will also acquire head posture parameters and determine whether the changes in the angles of the head posture parameters in the pitch, yaw, and roll directions during the tongue muscle extension meet the fifth preset threshold (e.g., the angle changes in each direction are ≤10 degrees) to ensure that the user maintains head stability during training.

[0031] When the preset rehabilitation movement is a tongue muscle circling movement, the assessment results are obtained as follows: First, lip state parameters are acquired, including lip pucker parameters (MOUTH_PUCKER) and upper lip curling parameters (MOUTH_ROLL_UPPER); it is determined whether the lip state parameters meet the sixth preset threshold (e.g., lip pucker parameter ≥ 0.4, upper lip curling parameter ≤ 0.2); if the lip state parameters meet the sixth preset threshold, the system will monitor the periodic displacement changes of key points around the mouth caused by tongue tip movement; finally, the number of circling cycles is determined based on the number of periodic displacement changes, and it is determined whether the number of circling cycles meets the second preset quantity requirement (e.g., ≥ 3 times).

[0032] When the preset rehabilitation movement is a neck massage movement, the evaluation results are obtained as follows: First, the key points of the hands are acquired and their movement trajectory is tracked in real time. In this embodiment, the MediaPipe Hand Landmarks module is used to track 21 key points of both hands. It is determined whether the difference between the end position and the starting position of the movement trajectory meets the seventh preset threshold (e.g., ≤20mm), and whether the shape of the movement trajectory meets the preset closed loop requirement (e.g., circle or ellipse). At the same time, the system constructs a three-dimensional model of the neck based on the facial key points and delineates a prohibited contact area containing preset danger areas (e.g., sensitive areas such as the carotid artery and thyroid gland). If any key point of the hand enters the prohibited contact area, the system will determine that the current neck massage movement does not meet the safety standards and immediately issue a warning.

[0033] Step 140: Based on the evaluation results, perform feedback operations and adjust subsequent training goals. The system provides real-time feedback to the user based on the evaluation results, including visual feedback (such as on-screen prompts indicating whether the action is correct), auditory feedback (such as a voice prompt "Please strengthen lip closure"), or tactile feedback (such as device vibration). Simultaneously, the system dynamically adjusts the difficulty and goals of subsequent training based on the user's performance. For example, the system inputs the patient's training data (including daily scores, durations, repetitions, etc.) into the training management model, which then outputs personalized training goals and / or evaluation thresholds for the patient, achieving dynamic optimization of the training plan.

[0034] During the adjustment of subsequent training goals, the system inputs the user's training data into the training management model. This data includes the user's performance on movements, parameter trends, training frequency, etc. The training management model analyzes this data and outputs personalized training goals and evaluation thresholds for the user. For example, if the system finds that the user's lip-closing ability has significantly improved, it may increase the required duration of lip closure from 2 seconds to 3 seconds, or increase the threshold for lip closure parameters from 0.7 to 0.8, to continuously challenge the user's ability limits and promote rehabilitation progress.

[0035] The training management model is obtained through the following steps: First, obtain a pre-trained basic model, such as an existing open-source large model, which already has basic data analysis and prediction capabilities. Then, the training data sequence and corresponding label data of the sample users are obtained. The label data is the optimal training target associated with the rehabilitation effect of the sample users. For example, the training data sequence includes rehabilitation training data of the sample patients for several consecutive days, such as daily lip closure score, tongue muscle extension score, training duration, completion rate, etc. The label data is the optimal training target associated with the rehabilitation effect of the sample patients. That is, for patients with good final rehabilitation effect, the training target on the N+1th day is used as the label. The training data sequence is then input into the base model to predict the training target; Finally, the predicted training objectives are compared with the labeled data, and the base model is fine-tuned based on the comparison results until the training completion conditions are met (e.g., prediction accuracy reaches over 95%). During the fine-tuning process, the underlying parameters of the base model remain unchanged, and only the top-level parameters are adjusted to adapt to the specific task of rehabilitation training. In this way, a high-performance training management model can be obtained with relatively little sample data.

[0036] In this way, facial rehabilitation exercise training can provide scientific, precise, and personalized training guidance for patients recovering from diseases such as nasopharyngeal carcinoma. It effectively improves sequelae caused by radiotherapy, such as difficulty opening the mouth, swallowing disorders, and muscle fibrosis, thereby improving patients' quality of life. The systematic quantitative assessment capability makes the rehabilitation process visible and traceable, the real-time feedback mechanism improves patient training compliance, and the personalized training plan ensures maximum rehabilitation effectiveness.

[0037] The present invention provides a computer program product comprising instructions that, when executed on a computer, cause the computer to perform the facial rehabilitation exercise training method described in Embodiment 1.

[0038] The computer program product can be stored in a computer-readable storage medium, such as ROM, RAM, disk, or optical disk. When the instructions in the computer program product are executed by a computer, they will implement all the functions and steps of the facial rehabilitation movement training method described in Embodiment 1, including acquiring the user's video stream; extracting facial key points from the video stream and generating facial expression parameters; determining and evaluating target parameters associated with preset rehabilitation movements from the facial expression parameters; and performing feedback operations and adjusting training objectives based on the evaluation results.

[0039] In a preferred embodiment, the computer program product is implemented using a hybrid C++ and Python programming approach, wherein the video processing and facial key point extraction modules are written in C++ to improve operational efficiency, while the training management model and user interface are implemented in Python to accelerate development.

[0040] In another preferred embodiment, the computer program product can be deployed as a mobile application on a smartphone or tablet, using the device's built-in front-facing camera to capture user video streams, enabling rehabilitation training anytime, anywhere. Simultaneously, the program can also upload training data to a cloud server, facilitating remote monitoring of the patient's rehabilitation progress by doctors.

[0041] This computer program product digitizes and automates facial rehabilitation exercise training methods, achieving standardized and quantitative assessment of nasopharyngeal carcinoma rehabilitation exercises. It provides real-time feedback to improve patient adherence and enhances rehabilitation outcomes through personalized training, thereby improving patients' quality of life. The program's automated assessment and feedback functions allow patients to receive professional rehabilitation guidance at home, effectively addressing the problems of insufficient professional guidance and difficulty in assessing training effects in traditional rehabilitation training.

[0042] The above description is merely a preferred embodiment of the present invention, and the present invention is not limited to the above embodiments. It is understood that other improvements and variations that are directly derived or conceived by those skilled in the art without departing from the spirit and concept of the present invention should be considered to be included within the protection scope of the present invention.

Claims

1. A facial rehabilitation exercise training method, characterized in that, Includes the following steps: Obtain the user's video stream; Facial key points are extracted from the video stream, and facial expression parameters are generated based on the facial key points; Target parameters associated with preset rehabilitation actions are determined from the facial expression parameters, and the target parameters are compared with preset evaluation criteria to obtain the evaluation results of the preset rehabilitation actions; Based on the evaluation results, perform feedback operations and / or adjust subsequent training objectives.

2. The method according to claim 1, characterized in that, The steps to obtain the evaluation results also include: Monitor whether the duration and / or number of repetitions of facial expression parameters associated with the preset rehabilitation movements meet preset standards.

3. The method according to claim 1, characterized in that, The steps to obtain the evaluation results also include: Monitor whether the displacement of key facial points associated with the preset rehabilitation movements meets the preset standards.

4. The method according to claim 1, characterized in that, When the preset rehabilitation movement is a lip-closing movement, the steps to obtain the assessment result include: Obtain facial expression parameters related to lip muscle contraction, including lip closure parameters, upper lip curling parameters, and lower lip curling parameters; Determine whether the values ​​of the lip closure parameter, upper lip curling parameter, and lower lip curling parameter meet a first preset threshold, and whether the duration of meeting the first preset threshold meets a first preset duration requirement.

5. The method according to claim 4, characterized in that, The steps to obtain the evaluation results also include: Get the real-time distance between the key points of the upper lip and the key points of the lower lip; The real-time distance is compared with the reference distance, and the comparison result helps to determine whether there is an air leak.

6. The method according to claim 1, characterized in that, When the preset rehabilitation exercise is a cheek-puffing exercise, the steps to obtain the assessment result include: Obtain the lip pout parameters and determine whether they meet the second preset threshold. In response to the lip pursing parameter meeting a second preset threshold, the displacement of key points in the cheek area relative to the initial position is monitored; Determine whether there is a first stage where the displacement exceeds a third preset threshold, and a second stage where the displacement returns to the initial position, and whether the total duration of the complete process consisting of the first and second stages meets a second preset duration requirement; or When the preset rehabilitation exercise is a tongue muscle extension exercise, the steps to obtain the assessment result include: Obtain mandibular movement parameters, including mandibular opening degree, mandibular leftward displacement degree, and mandibular rightward displacement degree; Determine whether the parameters corresponding to the current extension direction in the mandibular movement parameters meet a fourth preset threshold, and whether the number of repetitions meeting the fourth preset threshold meets a first preset quantity requirement; and, Acquire head posture parameters and determine whether the angular changes of the head posture parameters in the pitch, yaw, and roll directions during tongue muscle extension meet a fifth preset threshold; or When the preset rehabilitation movement is a tongue muscle circling movement, the steps to obtain the assessment result include: Obtain lip state parameters, including lip pout parameters and / or upper lip curling parameters; Determine whether the lip state parameters meet the sixth preset threshold; In response to the lip state parameters meeting the sixth preset threshold, periodic displacement changes of key points around the mouth caused by tongue tip movement are monitored; The number of orbiting cycles is determined based on the number of periodic displacement changes, and it is determined whether the number of orbiting cycles meets the second preset quantity requirement.

7. The method according to claim 1, characterized in that, When the preset rehabilitation movement is a neck massage movement, the steps to obtain the evaluation result include: Acquire key points of the hand and track its movement trajectory in real time; Determine whether the difference between the endpoint and starting position of the motion trajectory meets a seventh preset threshold, and whether the shape of the motion trajectory meets a preset closed loop requirement; and, A three-dimensional model of the neck is constructed based on facial key points, and a no-touch zone containing preset danger areas is delineated; If any of the aforementioned key hand points enter the prohibited contact area, the current neck massage action is determined to be non-compliant with safety standards.

8. The method according to claim 1, characterized in that, The steps for adjusting subsequent training objectives include: Input the user's training data into the training management model; Obtain the personalized training objectives and / or evaluation thresholds for the user, output by the training management model.

9. The method according to claim 8, characterized in that, The training management model is obtained through the following steps: Obtain the pre-trained base model; Obtain the training data sequence and corresponding label data of the sample users, wherein the label data is the optimal training target associated with the rehabilitation effect of the sample users; The training data sequence is input into the base model to predict the training target; The predicted training objective is compared with the labeled data, and the base model is fine-tuned based on the comparison results until the training completion conditions are met.

10. A computer program product, characterized in that, It includes instructions that, when executed on a computer, cause the computer to perform the method as described in any one of claims 1-9.

Images