Movement identification method for old people suffering from sarcopenia and movement APP for old people suffering from sarcopenia

Abstract

The invention provides a movement identification method for old people with sarcopenia and an old people sarcopenia movement APP, and relates to the technical field of movement identification. Action videos are matched from an action video library, communication with a positioning patch array is carried out, the action videos are processed through an action analysis network, and the movement identification result is obtained. Abnormal action identification is carried out on the positioning patch position time sequence information according to the positioning patch position target time sequence information, after deviation position identification time sequence information is obtained, an action video is segmented to obtain a guidance action video, and the problems that sarcopenia movement assistance lacks real-time and personalized movement guidance, feedback delay is caused, and the accuracy of movement guidance is poor are solved. The problems that the exercise effect and safety cannot be guaranteed and the injury risk is increased due to poor interactivity caused by the fact that most exercises need to be supervised by professionals are solved, the exercise accuracy and safety are improved, the exercise auxiliary response instantaneity, adaptability and exercise intervention effectiveness are improved, and the exercise effect and safety are improved. The exercise quality of the old with sarcopenia and the compliance of the exercise prescription are improved.

Description

Technical Field

[0001] The present invention relates to the field of motion identification technology, and specifically to a method for identifying motion for elderly people with sarcopenia and a sarcopenia motion APP for the elderly. Background Art

[0002] Sarcopenia is a common senile disease, which is mainly manifested by the gradual atrophy of skeletal muscles with age, leading to decreased physical strength, mobility problems and even loss of self-care ability in the elderly. At present, appropriate exercise intervention can effectively delay or even partially reverse the development of sarcopenia, but the elderly often face the problem of being unable to accurately perform exercise movements and difficulty in obtaining immediate feedback and guidance when exercising, which leads to poor compliance with medical advice. As for the issue of executing doctor-ordered movements, although there are a variety of sarcopenia exercise assistance products on the market, most products cannot provide real-time movement correction or feedback, and lack action plans customized for individual health conditions. In addition, existing solutions often need to be used under the supervision of a gym or medical institution, which is not a viable long-term solution for many elderly people with limited mobility.

[0003] In summary, existing sarcopenia exercise aids usually lack real-time and personalized exercise guidance, resulting in delayed feedback, and most of them need to be carried out under the supervision of professionals. The interactivity is poor, so the exercise effect and safety cannot be guaranteed, which increases the risk of injury. Summary of the invention

[0004] The present application provides a method for identifying exercise for elderly people with sarcopenia and an exercise APP for elderly people with sarcopenia, which are used to solve the technical problems that existing sarcopenia exercise assistance usually lacks real-time and personalized exercise guidance, resulting in delayed feedback, and most of them need to be conducted under the supervision of professionals, with poor interactivity, so that exercise effects and safety cannot be guaranteed, and the risk of injury is increased.

[0005] The present application provides a method for identifying movement of elderly people with sarcopenia, the method comprising:

[0006] In response to the motion scheme number selected by the client, a target motion video is matched from the motion video library; the positioning patch array is communicated with to receive a plurality of positioning patch position timing information, wherein the positioning patches are deployed at preset joints of the elderly; the target motion video is processed through a motion analysis network to obtain a plurality of positioning patch position target timing information; according to the plurality of positioning patch position target timing information, the plurality of positioning patch position timing information is marked with abnormal motion to obtain deviation position mark timing information; according to the deviation position mark timing information, the target motion video is segmented to obtain a guide motion video; and the guide motion video is sent to the client for display.

[0007] The present application provides a sarcopenia exercise APP for the elderly, which is used to implement a method for identifying exercise for sarcopenia elderly people.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] The present application provides a method for identifying the movement of elderly people with sarcopenia. In response to the movement plan number selected by the client, a target movement video is matched from the movement video library; the positioning patch array is communicated with to receive a plurality of positioning patch position timing information, wherein the positioning patches are deployed at preset joints of the elderly; the target movement video is processed through a movement analysis network to obtain a plurality of positioning patch position target timing information; based on the plurality of positioning patch position target timing information, the plurality of positioning patch position timing information is subjected to abnormal movement identification to obtain deviation position identification timing information; the target movement video is segmented based on the deviation position identification timing information to obtain a guidance movement video; the guidance movement video is transmitted to the target movement video. It is sent to the client for display, which solves the technical problem that existing sarcopenia exercise assistance usually lacks real-time and personalized exercise guidance, resulting in feedback delays, and most of the exercises need to be conducted under the supervision of professionals, with poor interactivity, which makes it impossible to guarantee the exercise effect and safety, and increases the risk of injury. Through real-time monitoring of positioning patches and motion analysis networks, it provides instant motion correction and personalized exercise plans, enhances the safety and effectiveness of exercise, and supports remote medical assistance, making exercise intervention more intelligent, personalized and convenient, achieving the technical effect of improving exercise accuracy and safety, increasing the immediacy, adaptability and effectiveness of exercise assistance response, and improving the exercise quality and compliance of exercise prescriptions for elderly people with sarcopenia. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A schematic flow chart of a method for identifying exercise for elderly people with sarcopenia is provided for this application.

[0011] Figure 2 This application provides a schematic diagram of the process of constructing a motion analysis network in a method for identifying movements of elderly people with sarcopenia.

[0012] Figure 3 This is a schematic diagram of the structure of an exemplary electronic device of the present application.

[0013] Description of the reference numerals: processor 31 , memory 32 , input device 33 , output device 34 . DETAILED DESCRIPTION

[0014] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0015] Embodiment 1, as Figure 1 As shown, the present application provides a method for identifying movement of elderly people with sarcopenia, the method comprising:

[0016] In response to the motion scheme number selected by the client, a target motion video is matched from the motion video library.

[0017] Specifically, based on the exercise program number selected by the client, the corresponding target action video is first matched from the action video library. This process involves user interaction on the client interface, and the user selects a specific exercise program number through the client interface. The exercise program number is pre-set in the system database, and each number corresponds to a set of specific exercise videos and exercise instructions, which are designed to meet the treatment and prevention needs of different sarcopenia patients. After receiving the exercise program number, the action video library located on the server or cloud is immediately accessed. The action video library is a collection of various pre-recorded exercise action videos, each of which is equipped with detailed action execution instructions to ensure the correctness and effect of the exercise. The videos in the video library are associated with different exercise program numbers, so that the system can accurately retrieve the corresponding target action video after each user selection. During the retrieval process, the system uses the exercise program number as a key index to query the video file path and related attributes associated with the number in the database. Once a matching target action video is found, it is retrieved from the video library for further processing and display. This step is performed automatically, ensuring the speed and accuracy of the operation, allowing users to view the required exercise video without waiting. The retrieved target action video is then sent to the user's client for display. During this process, the video data may be transmitted over the network, involving data compression and buffering technology to optimize the transmission speed and viewing experience. Once the video is successfully loaded, the user can watch the video on the client and perform the corresponding exercises according to the instructions shown in the video, thereby achieving the purpose of treating or preventing sarcopenia.

[0018] Through the above steps, not only the operation process of elderly users is simplified, but also the selection of exercise plans and video playback are more intuitive and efficient, which greatly improves the user experience and convenience of treatment.

[0019] Communicate with the positioning patch array and receive the position timing information of a number of positioning patches, wherein the positioning patches are deployed at preset joints of the elderly.

[0020] Optionally, in order to achieve accurate monitoring of the elderly user's exercise movements, the communication mechanism between the positioning patch array and the client is used to collect and transmit the position timing information of the positioning patch. The positioning patch is a small device equipped with a sensor, which can be attached to the user's joint position such as the knee, shoulder or wrist. These patches can capture and record the specific position and movement trajectory of the joint during the movement based on the time sequence, and generate position timing information. When the user starts to exercise, each positioning patch monitors and records the spatial position changes of the joint in real time based on its built-in sensor. These change data are arranged in chronological order to form timing information, which can reflect the accuracy and continuity of the user's movement in detail. The positioning patch corresponding to each joint sends this data to the user's client device in real time through wireless means (such as Bluetooth, Wi-Fi, etc.), or directly transmits it to the cloud server. After receiving the data from each positioning patch, the system backend will summarize and analyze this information. In this way, real-time data of all action execution processes can be provided for further analysis of the user's sports performance, or detailed exercise execution reports can be provided to medical professionals. The application of this positioning patch array greatly enhances the accuracy and effectiveness of exercise intervention, providing a scientific and effective exercise treatment plan for elderly patients with sarcopenia.

[0021] The target action video is processed through an action analysis network to obtain a plurality of positioning patch position target time sequence information.

[0022] Further, in order to ensure that the actual action data monitored can be effectively compared with the ideal action model to judge the quality of the user's action execution, the target action video is processed by the action parsing network to obtain the position target timing information corresponding to the positioning patch. The action parsing network is a computational model built using deep learning technology specifically for analyzing and understanding complex action sequences. The network can identify key action features in the video and convert these features into target position information that the positioning patch should reach. In the present invention, the target action video is pre-recorded and shows the standard execution method of each action in a specific motion scheme. When the target action video is input into the action parsing network, the network first performs frame analysis on the video, that is, decomposing the video into a series of continuous image frames. Among them, each frame image contains the posture and joint position of the exercise performer at a specific moment. The action parsing network identifies the key joint positions of the characters in the video by analyzing these frame images, and these positions are the target positions that each joint should reach in an ideal state. Subsequently, the network uses the feature extraction and timing analysis techniques learned during the training process to convert these identified joint positions into timing information. This conversion process involves mapping the spatial information in the image (such as joint position coordinates) onto the time axis to form a dynamic data sequence that can be continuously updated over time, namely the position target timing information. This position target timing information reflects the specific position of each joint in the ideal motion state in the time sequence during the motion process. This information is then stored and used to compare with the real-time position timing information from the positioning patch on the user to evaluate the accuracy and fluency of the user's execution of the action.

[0023] Through the above steps, the motion analysis network provides an accurate benchmark for monitoring and guiding the user's exercise training, ensuring that each movement is performed in accordance with established health standards, thereby improving the effectiveness and safety of treatment, and significantly improving the scientificity and practicality of exercise intervention for sarcopenia.

[0024] According to the target timing information of the positions of the plurality of positioning patches, abnormal action marking is performed on the timing information of the positions of the plurality of positioning patches to obtain deviation position marking timing information.

[0025] Exemplarily, by comparing the position target timing information of the plurality of positioning patches with the actual positioning patch position timing information, the system can identify abnormal movements and identify deviation positions. This step is a key link to ensure the accuracy and safety of exercise execution, thereby accurately adjusting and optimizing the exercise therapy for elderly patients with sarcopenia. First, the position target timing information obtained by the system from the motion analysis network represents the change of the specific position of each joint patch over time under ideal conditions. This information forms a standard template for evaluating the accuracy of the user's actual movement. The positioning patch collects the actual position of each joint of the user during the movement in real time and records it in the form of timing data, that is, the actual position timing information. Subsequently, the standard position target timing information is compared with the actual position timing information. This comparison process is performed through a series of algorithms, such as calculating the Euclidean distance between the two sets of data or other suitable distance measurement methods. The actual joint position at each time point will be compared with its corresponding target position, and then the preset threshold is used to determine whether there is a significant difference between the two. When the distance between the actual position and the target position exceeds the set threshold, the system identifies the deviation as an abnormal action. The specific moment and joint position of the abnormal action will be marked and recorded as the deviation position identification timing information. This information not only indicates the specific error in the execution of the action, but also records the specific time when the error occurred, providing an accurate error correction reference for subsequent training. For example, when performing a knee bending exercise, if the user's knee positioning patch exceeds the threshold when it reaches the lowest point and the target position, the system will mark a deviation point at that moment. After this deviation point, specific correction suggestions or animations can be displayed on the client to help users understand and correct their exercise posture. Through this detailed action monitoring and deviation identification mechanism, not only the accuracy of exercise therapy is improved, but also the personalization and effectiveness of treatment are greatly increased, ensuring that every elderly patient with sarcopenia can benefit from exercise to the maximum extent under the premise of safety.

[0026] The target action video is segmented according to the deviation position identification timing information to obtain a guidance action video.

[0027] The guiding action video is sent to the client for display.

[0028] Specifically, after the system identifies and records the timing information of the deviation position identification, it then uses this information to accurately segment the target action video and generate the guidance action video, which is then sent to the client for display to help the user correct his or her movement. First, the specific segments of the target action video that need to be paid attention to are determined based on the recorded timing information of the deviation position identification. These timing information indicate which specific time points and action postures of the user have deviations during the movement. The corresponding execution software on the system uses this information as input for video editing and selects the video segments corresponding to these deviations. Subsequently, these selected segments are carefully edited and segmented. During the editing process, additional visual effects such as arrows, highlights, or slow-motion playback can be added to highlight the correct movement methods and postures, ensuring that users can clearly understand their movement deviations and how to correct them. In addition, the playback speed of the guidance action video will be adjusted to a slower speed to adapt to the viewing and learning speed of elderly users, ensuring that they have enough time to understand and imitate the correct movements. Once the guidance action videos are prepared, the system sends these videos to the user's client device via the network. On the client side, the video will display relevant text and guidance information in a larger font, and will be presented in a slow-motion playback mode that is easy for users to understand, helping users to clearly see the details of each movement. At the same time, the system can also be configured to automatically repeat specific action clips until the user feels that he or she can master and imitate the correct actions shown in the video. In addition, in order to enhance the learning effect, the client application will also be equipped with reminder functions, such as visual or sound reminders, to encourage users to review these videos regularly in daily practice and continuously improve the accuracy of their exercise execution. In this way, not only is precise guidance on exercise therapy for patients with sarcopenia technically achieved, but the intelligent functions of the client also greatly improve the user's participation and the effect of movement correction, thereby effectively assisting patients with sarcopenia to improve their athletic ability and quality of life.

[0029] Furthermore, the target action video is processed through the action parsing network to obtain several positioning patch position target time sequence information, including:

[0030] The motion analysis network includes a preset joint positioning channel and a preset joint motion extraction channel, wherein the preset joint is a joint where the positioning patch is deployed.

[0031] The target action video is segmented into frames according to the first step length to obtain a first target action video frame sequence.

[0032] The target action video is segmented into frames according to a second step length to obtain a second target action video frame sequence, wherein the second step length is 1 / 8 of the first step length.

[0033] The first target action video frame sequence is input into the preset joint positioning channel, the second target action video frame sequence is input into the preset joint action extraction channel, and the plurality of positioning patch position target timing information is output.

[0034] In a specific embodiment, the motion parsing network is designed to process the target motion video and extract the key positioning patch position target timing information from it. The network includes a preset joint positioning channel and a preset joint motion extraction channel to enhance the analysis accuracy and efficiency of specific joint motions. First, the target motion video, that is, a video containing a complete motion sequence of demonstration, is input into the motion parsing network. This video is first frame segmented according to the first step length to decompose the video content into a series of continuous frames, which record every action detail in the motion process in detail. The first step length is pre-set according to the complexity of the video content and the action to ensure that enough frames are used to capture key action changes, thereby forming a first target action video frame sequence. Subsequently, the same video is frame segmented according to a finer second step length, where the second step length is 1 / 8 of the first step length, which allows higher frequency frame capture to capture more subtle actions and joint changes, and generate a second target action video frame sequence. This denser frame sequence helps capture fast or less obvious action details and provides more data to support action extraction analysis. Next, the first target action video frame sequence is input into the preset joint positioning channel, which is specifically used to determine the joint position of the character in the video, and locates the main position changes of the joints during the movement by analyzing the lower frequency frame sequence. The processing result of this stage is the precise position of the main stop points and turning points of the joints during the movement. At the same time, the second target action video frame sequence is input into the preset joint action extraction channel, which processes higher frequency frame sequences and focuses on analyzing the continuous movement of the joints during the movement, such as the smoothness of the movement and the slight adjustment of the joints. This meticulous analysis helps to reveal details that may be overlooked in the execution of the movement, such as slight muscle tremors or atypical movement trajectories of the joints. Finally, through the collaborative work of these two channels, the system can output accurate positioning patch position target timing information, which exhaustively reflects every position point and its time sequence that the preset joint should reach in the ideal movement execution. These data provide a scientific basis for subsequent movement quality evaluation and user movement guidance, ensuring the accuracy and effectiveness of exercise training, which is particularly critical for the rehabilitation treatment of patients with sarcopenia.

[0035] Furthermore, if Figure 2 As shown, the action parsing network construction steps include:

[0036] A slowfast model topology is obtained, wherein the slowfast model topology includes a slow processing channel and a fast processing channel.

[0037] Collect character action videos and preset joint position timing information.

[0038] The character action video is segmented into frames according to the first step length to obtain a first training action video frame sequence.

[0039] The character action video is segmented into frames according to the second step length to obtain a second training action video frame sequence.

[0040] The first training action video frame sequence is input into the slow processing channel, and the second training action video frame sequence is input into the fast processing channel. The slowfast model topology is trained using the preset joint position timing information as supervision data.

[0041] When the output loss value of the slowfast model topology is less than or equal to the output loss threshold for a preset number of consecutive times, the slow processing channel is set to the preset joint positioning channel, the fast processing channel is set to the preset joint motion extraction channel, and the motion parsing network is generated.

[0042] Furthermore, the slowfast model is a dual-path network, which consists of two processing channels, namely the slow processing channel and the fast processing channel. One of the channels focuses on capturing fast actions in the video (fast path), and the other channel captures relatively slow actions (slow path). This dual-channel design can simultaneously capture slow and fast dynamic changes in action videos, improving the accuracy and efficiency of action recognition. First, the construction of the action parsing network begins with obtaining the slowfast model topology. This model topology is designed based on the spatiotemporal characteristics of the action. The slow channel aims to capture the slow, basic action changes in the video, while the fast channel focuses on fast, instantaneous action details. This design allows the network to more comprehensively understand and analyze complex human actions. Subsequently, in order to train this model, it is necessary to collect human action videos and the corresponding preset joint position timing information. These videos should record different types of human actions in detail, and the preset joint position timing information provides the ideal movement path of each joint in these actions as the supervision data for subsequent training. After collecting enough training materials, the character action video is then segmented according to the predetermined first step length to obtain the first training action video frame sequence. This step length usually selects a frame rate that can capture the key action nodes to ensure that the slow channel can effectively learn the basic process of the action and the main motion trajectory of the joints. At the same time, the video needs to be segmented according to the second step length to generate the second training action video frame sequence. The second step length is shorter than the first step length, usually 1 / 8 of it. This setting enables the fast channel to capture more subtle action changes, such as fast gestures or short balance adjustments. These details are crucial to understanding the integrity of the action. After that, the first training action video frame sequence is input into the slow processing channel, and the second training action video frame sequence is input into the fast processing channel. During the training process, the two channels learn and adjust their own parameters respectively to best simulate and predict the movement of the joints. The training uses the preset joint position timing information as the supervision data to ensure that the motion trajectory output by the model is as consistent as possible with the standard trajectory of the actual action. The performance of the model during the training process is evaluated by calculating the output loss value, which represents the error between the model's predicted output and the actual data. Once the output loss value is less than or equal to the preset output loss threshold for several consecutive times, it indicates that the model is stable and accurate enough. At this time, the slow channel can be determined as the preset joint positioning channel, and the fast channel can be determined as the preset joint motion extraction channel, completing the final construction of the motion parsing network.

[0043] Through the above steps, an efficient and accurate motion analysis network was realized, which can provide detailed analysis for various movements, support subsequent movement quality assessment and guidance, and provide scientific and effective data support for the rehabilitation treatment of sarcopenia patients.

[0044] Further, according to the plurality of positioning patch position target timing information, abnormal action identification is performed on the plurality of positioning patch position timing information to obtain deviation position identification timing information, including:

[0045] According to the target timing information of the positions of the plurality of positioning patches, the target position of the first positioning patch at the first moment is extracted.

[0046] The first moment position of the first positioning patch is extracted according to the position timing information of the plurality of positioning patches.

[0047] When the distance between the first moment position of the first positioning patch and the first moment target position of the first positioning patch is greater than or equal to a distance threshold, the first moment position of the first positioning patch is added to the deviation position identification timing information.

[0048] Specifically, the position target timing information of each joint is extracted from the target action video through the action analysis network or other appropriate data processing methods, that is, the first moment target position of the first positioning patch. This position information represents the precise position that the joint or positioning patch should reach at a specific moment in the ideal action execution. Then, the system synchronously monitors and records the positioning patch position at the same moment in the user's actual movement, that is, the first moment position of the first positioning patch. This step is completed by real-time capture and analysis of the signal sent back by the positioning patch worn by the user, ensuring the real-time and accuracy of the data. After that, the two position information are compared. During this comparison process, the distance between the two positions is calculated and compared with the preset distance threshold. The distance threshold is pre-set based on kinematics and medical standards in order to determine what degree of deviation will have a negative impact on the exercise effect or may cause sports injuries. If the calculated distance is greater than or equal to this threshold, the action execution at this moment is deemed to be an abnormal action. At this time, the actual position at this specific moment is marked as a deviation position and recorded in the deviation position identification timing information. This deviation information will be used for subsequent user action guidance and training optimization, providing users with specific correction suggestions or automatically adjusting subsequent training plans. For example, if the actual position of the user's knee positioning patch at the lowest point exceeds the set threshold from the target position during a squat, the system will record this deviation and may send visual or audio prompts to the user through the client to guide the user to adjust the angle or position of the knee to achieve a safer and more effective exercise effect.

[0049] Through the above steps, key deviations during exercise can be accurately identified and recorded, and timely feedback and corrective measures can be provided, which greatly improves the safety and effectiveness of exercise training. This technology provides important support, especially for elderly patients with sarcopenia who have weak motor control ability.

[0050] Furthermore, the first training action video frame sequence is input into the slow processing channel, the second training action video frame sequence is input into the fast processing channel, and the preset joint position timing information is used as supervision data to train the slowfast model topology, including:

[0051] Construct the training loss function:

[0052]

[0053] Among them, Loss represents the loss value of any training, x ij0 Represents the predicted position of the i-th preset joint at the j-th moment, x ij1 represents the supervised position of the i-th preset joint at the j-th moment, Q represents the total number of preset joints, and M represents the total number of moments.

[0054] According to the training loss function, the first training action video frame sequence is input into the slow processing channel, the second training action video frame sequence is input into the fast processing channel, and the slowfast model topology is trained using the preset joint position timing information as supervision data.

[0055] Optionally, during the training process, the system uses the preset joint position timing information as supervision data to train the two channels in the slowfast model. The specific training method uses the loss function calculation, which is defined as Among them, Loss represents the loss value of any training, M represents the total number of moments in the training sample, Q represents the total number of joints, and x ij0 Represents the predicted position of the i-th preset joint at the j-th moment, x ij1 Characterizes the i-th preset joint supervision position at the j-th moment, and d is the distance function used to calculate the error between the predicted position and the actual position. This loss function helps optimize the model output to make it as close to the actual action execution as possible and reduce the prediction error. When the output loss value of the slowfast model is less than or equal to the preset loss threshold after multiple consecutive training iterations, it indicates that the model is stable and accurate enough, and the slow processing channel can be determined as the preset joint positioning channel, and the fast processing channel can be determined as the preset joint action extraction channel. At this time, the construction of the action parsing network is completed. This action parsing network based on comprehensive spatiotemporal features provides an efficient and accurate technical method for action recognition and evaluation, and is particularly suitable for application scenarios that require high action accuracy, such as sports training, physical therapy, and other health-related monitoring and analysis tasks.

[0056] Furthermore, in response to the motion scheme number selected by the client, the method further includes:

[0057] The basic information of the elderly with sarcopenia is obtained, wherein the basic information of the elderly with sarcopenia includes an illness duration identifier, a patient gender identifier, a patient age identifier, and a patient weight identifier.

[0058] Based on the exercise program number library, the exercise program historical configuration data is extracted, wherein the exercise program historical configuration data includes the historical configuration illness duration, the historical configuration patient gender, the historical configuration patient age, and the historical configuration patient weight.

[0059] Based on the historically configured illness duration, the historically configured patient gender, the historically configured patient age, and the historically configured patient weight, a list of exercise program numbers and a list of exercise program configuration frequencies configured with the illness duration identifier, the patient gender identifier, the patient age identifier, and the patient weight identifier are counted.

[0060] The motion scheme number list is sorted from large to small according to the motion scheme configuration frequency list, and a motion scheme number sequence is obtained and sent to the client for selection.

[0061] Exemplarily, the system responds to the exercise program number selected by the client. In this process, the system first needs to obtain the patient's basic information, which includes the duration of illness, the patient's gender, the patient's age and the patient's weight. These basic information provide important reference value for the subsequent personalized exercise program, ensuring that the recommended exercise program can adapt to the patient's specific health needs. Subsequently, the exercise program number library is accessed to extract historical configuration data similar to the current patient. These historical configuration data include the duration of illness, gender, age and weight of the historical configuration. This step aims to predict and optimize the exercise program that may be suitable for the current patient through historical cases. Based on the collected basic information of the patient and the historical configuration data, statistical analysis is further performed to compare the historical configuration of the duration of illness, gender, age, weight and the corresponding information of the current patient, and the exercise program number that matches the current patient's condition is counted. In addition, a list of exercise program configuration frequencies is generated, showing the frequency with which different exercise programs are configured, which helps to identify the most commonly used and effective exercise programs. After the statistical analysis is completed, the list of exercise program numbers is sorted from high to low according to the list of exercise program configuration frequencies. The purpose of this step is to recommend the most suitable and popular exercise programs to patients first, ensuring that patients can obtain the best treatment effect. Finally, the sorted sequence of exercise program numbers will be sent to the client for patients or medical service providers to choose. When displaying this information, the client will use an easy-to-understand and easy-to-operate interface design, so that patients can simply and intuitively choose the exercise program that best suits them, so that they can continue to carry out appropriate physical activities and rehabilitation training in a home environment. Through this coherent and systematic step, not only the accuracy of exercise program recommendations is improved, but also the patient's treatment compliance and overall treatment effect are enhanced. Especially for the special group of elderly people with sarcopenia, this personalized exercise program provision method is particularly important.

[0062] Furthermore, the method further comprises:

[0063] The timing information of the positions of the plurality of positioning patches is sent to the medical management end to obtain the doctor's feedback information.

[0064] When the doctor feedback information has an updated exercise program number, the updated exercise program number is sent to the client for recommendation.

[0065] Specifically, in order to ensure that the exercise plan of sarcopenia patients can be monitored and fed back by medical professionals in real time, the position timing information collected by the positioning patch can be sent to the medical management end, and the exercise plan can be adjusted and optimized according to the doctor's feedback. First, the positioning patch device in the system monitors the joint position of the elderly with sarcopenia in real time when they perform the predetermined exercise plan and records the generated position timing information, which includes but is not limited to the specific position coordinates and corresponding timestamps of each joint during the exercise process to ensure the detailed and accurate data. Then, the collected positioning patch position timing information is sent to the medical management end through a secure data transmission protocol. The medical management end is operated by a professional medical team, whose members include physical therapists, sports rehabilitation experts, etc. They are responsible for reviewing the patient's exercise data, evaluating the accuracy of exercise execution and its potential impact on the patient's health. After the medical management end receives and analyzes the position timing information of the positioning patch, the doctor will provide specific feedback information based on the data analysis results. These feedbacks may include suggestions for adjusting the current exercise plan, such as modifying the intensity of exercise, changing the type of exercise, or adjusting the frequency of exercise, etc., to better adapt to the patient's physical condition and rehabilitation needs. When the doctor's feedback includes suggestions for updating the exercise plan number, the system records these updated exercise plan numbers and processes them through the background management system. During the processing, the feasibility and safety of the new plan will be checked, and after confirmation, the updated exercise plan number will be sent to the client. Finally, after receiving the new exercise plan number, the client will show it to the patient and recommend that they update the exercise plan according to the doctor's professional advice. The client interface will friendly guide patients on how to switch to the new exercise plan, and provide necessary operating instructions and visual assistance to ensure that patients can switch and start a new exercise plan without obstacles. Through this coherent information flow and feedback mechanism, it is ensured that the exercise plan of patients with sarcopenia can be adjusted and optimized in a timely manner under the supervision of medical professionals, which greatly improves the adaptability and effectiveness of the treatment plan, while also ensuring the safety of patients' exercise.

[0066] Through the technical solutions of the above embodiments, the method for identifying movement of elderly people with sarcopenia provided by this application has the following technical effects:

[0067] 1. By combining the motion parsing network and the slowfast model to process the elderly's exercise videos, the fast and slow movement changes that may occur in the elderly when performing sarcopenia exercises can be captured. This can more accurately analyze and understand the elderly's movement characteristics, and provide correct feedback and guidance for movements of different speeds, greatly improving the safety and effectiveness of exercise and avoiding the risk of injury caused by improper execution of movements.

[0068] 2. Use the positioning patch array deployed on the joints of the elderly to monitor the execution of movements in real time, and work in conjunction with the motion analysis network to not only capture the execution of movements, but also compare the target movements in real time, identify and mark the deviations in movement execution, and then provide immediate movement correction suggestions to help the elderly correctly perform various sarcopenia prevention and treatment exercises, thereby improving the efficiency and effectiveness of exercise implementation, and reducing possible errors and injuries during exercise.

[0069] 3. Based on basic information such as the duration of illness, gender, age and weight of the elderly, the most suitable exercise plan is matched and recommended through intelligent algorithms. It can analyze historical data and user feedback, dynamically adjust and optimize the recommendation algorithm, and personalized exercise plans can more accurately meet the physical conditions and health needs of different elderly people, thereby improving the applicability of exercise plans and user participation, thereby more effectively delaying the development of sarcopenia.

[0070] 4. The plan not only includes the remote transmission of motion data to the medical management end, but also includes a mechanism for obtaining feedback from doctors and updating exercise plans. This enables doctors to provide professional medical advice and adjust exercise plans based on the remotely monitored exercise execution. Through the remote participation of doctors, the professionalism and timeliness of the exercise plan are enhanced, so that every elderly person can perform safe and effective exercise under the guidance of medical professionals, and it also facilitates doctors to conduct continuous health management of patients.

[0071] Embodiment 2, a sarcopenia exercise APP for the elderly, is used to implement a method for identifying exercise for sarcopenia elderly people provided in embodiment 1.

[0072] Embodiment 3, Figure 3 The schematic structural diagram of the electronic device provided in the second embodiment of the present invention shows a block diagram of an exemplary electronic device suitable for implementing the implementation mode of the present invention. Figure 3 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention. Figure 3 As shown, the electronic device includes a processor 31, a memory 32, an input device 33 and an output device 34; the number of processors 31 in the electronic device can be one or more. Figure 3 Taking a processor 31 as an example, the processor 31, memory 32, input device 33 and output device 34 in the electronic device can be connected through a bus or other means. Figure 3 The example of connecting through bus is taken in the following.

[0073] The memory 32, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to a method for identifying movement of elderly people with sarcopenia in an embodiment of the present invention. The processor 31 executes various functional applications and data processing of the electronic device by running the software programs, instructions and modules stored in the memory 32, that is, realizing the above-mentioned method for identifying movement of elderly people with sarcopenia.

[0074] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0075] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0076] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for identifying movement of elderly people with sarcopenia, characterized in that: include: In response to the motion scheme number selected by the client, matching the target motion video from the motion video library; Communicate with the positioning patch array and receive the position timing information of a number of positioning patches, wherein the positioning patches are deployed at preset joints of the elderly; The target action video is processed through an action analysis network to obtain a plurality of positioning patch position target time sequence information; According to the target timing information of the plurality of positioning patches, abnormal action marking is performed on the timing information of the plurality of positioning patches to obtain deviation position marking timing information; Segmenting the target action video according to the deviation position identification timing information to obtain a guiding action video; The guiding action video is sent to the client for display.

2. A method for identifying movement of elderly people with sarcopenia as claimed in claim 1, characterized in that: The target action video is processed through the action parsing network to obtain several positioning patch position target time sequence information, including: The motion analysis network includes a preset joint positioning channel and a preset joint motion extraction channel, wherein the preset joint is a joint where the positioning patch is deployed; Segment the target action video by the number of frames according to the first step length to obtain a first target action video frame sequence; Performing frame segmentation on the target action video according to a second step length to obtain a second target action video frame sequence, wherein the second step length is 1 / 8 of the first step length; The first target action video frame sequence is input into the preset joint positioning channel, the second target action video frame sequence is input into the preset joint action extraction channel, and the plurality of positioning patch position target timing information is output.

3. A method for identifying movement of elderly people with sarcopenia as claimed in claim 2, characterized in that: The action parsing network construction step includes: Obtaining a slowfast model topology, wherein the slowfast model topology includes a slow processing channel and a fast processing channel; Collect character action videos and preset joint position timing information; Segment the character action video by the number of frames according to the first step length to obtain a first training action video frame sequence; Performing frame segmentation on the character action video according to the second step length to obtain a second training action video frame sequence; Inputting the first training action video frame sequence into the slow processing channel, inputting the second training action video frame sequence into the fast processing channel, and using the preset joint position timing information as supervision data to train the slowfast model topology; When the output loss value of the slowfast model topology is less than or equal to the output loss threshold for a preset number of consecutive times, the slow processing channel is set to the preset joint positioning channel, the fast processing channel is set to the preset joint motion extraction channel, and the motion parsing network is generated.

4. A method for identifying movement of elderly people with sarcopenia as claimed in claim 1, characterized in that: According to the plurality of positioning patch position target time sequence information, abnormal action identification is performed on the plurality of positioning patch position time sequence information to obtain deviation position identification time sequence information, including: Extracting the first moment target position of the first positioning patch according to the target timing information of the plurality of positioning patches; Extracting the first moment position of the first positioning patch according to the plurality of positioning patch position timing information; When the distance between the first moment position of the first positioning patch and the first moment target position of the first positioning patch is greater than or equal to a distance threshold, the first moment position of the first positioning patch is added to the deviation position identification timing information.

5. A method for identifying movement of elderly people with sarcopenia as claimed in claim 3, characterized in that: Inputting the first training action video frame sequence into the slow processing channel, inputting the second training action video frame sequence into the fast processing channel, and using the preset joint position timing information as supervision data to train the slowfast model topology, including: Construct the training loss function: Among them, Loss represents the loss value of any training, x ij0 Represents the predicted position of the i-th preset joint at the j-th moment, x ij1 represents the i-th preset joint supervision position at the j-th moment, Q represents the total number of preset joints, and M represents the total number of moments; According to the training loss function, the first training action video frame sequence is input into the slow processing channel, the second training action video frame sequence is input into the fast processing channel, and the slowfast model topology is trained using the preset joint position timing information as supervision data.

6. A method for identifying movement of elderly people with sarcopenia as claimed in claim 1, characterized in that: In response to the motion plan number selected by the client, the method further includes: Obtaining basic information of the elderly with sarcopenia, wherein the basic information of the elderly with sarcopenia includes an identification of the duration of illness, a patient's gender, a patient's age, and a patient's weight; Extracting historical configuration data of the exercise program based on the exercise program number library, wherein the historical configuration data of the exercise program includes the historical configuration illness duration, the historical configuration patient gender, the historical configuration patient age, and the historical configuration patient weight; Based on the historically configured illness duration, the historically configured patient gender, the historically configured patient age, and the historically configured patient weight, a list of exercise program numbers and a list of exercise program configuration frequencies configured with the illness duration identifier, the patient gender identifier, the patient age identifier, and the patient weight identifier are counted; The motion scheme number list is sorted from large to small according to the motion scheme configuration frequency list, and a motion scheme number sequence is obtained and sent to the client for selection.

7. A method for identifying movement of elderly people with sarcopenia as claimed in claim 1, characterized in that: Also includes: Sending the position sequence information of the plurality of positioning patches to the medical management terminal to obtain doctor feedback information; When the doctor feedback information has an updated exercise program number, the updated exercise program number is sent to the client for recommendation.

8. A sarcopenia exercise APP for the elderly, characterized in that: Used to implement a method for identifying movement in elderly people with sarcopenia as described in any one of claims 1-7.

9. An electronic device, characterized in that: The electronic device comprises: A memory for storing executable instructions; A processor is used to implement a method for identifying movement of elderly people with sarcopenia as described in any one of claims 1 to 7 when executing executable instructions stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements a method for identifying movement in elderly people with sarcopenia as described in any one of claims 1-7.

Images