An automatic evaluation method, device, medium and product for the exercise capacity of the elderly

Abstract

The application discloses an automatic evaluation method, device, medium and product of the exercise capacity of the elderly, and relates to the field of computer vision. The method comprises the following steps: acquiring a to-be-tested video of a subject; the to-be-tested video comprises the following videos: the subject puts both hands behind the head, squats down and sits down; extracting skeleton key point data from the to-be-tested video by using a key point detection algorithm; determining a skeleton feature by using a feature extraction network model trained based on a graph convolution network according to the skeleton key point data; and determining an exercise capacity evaluation result by using a multi-modal classification model trained based on a long short-term memory network and a full convolution network according to the skeleton feature. The application can realize simple implementation, high precision and high efficiency of the automatic evaluation of the exercise capacity of the elderly in a family environment.

Description

Technical Field

[0001] This application relates to the field of computer vision, and in particular to a method, device, medium, and product for automatically assessing the motor abilities of the elderly. Background Technology

[0002] With the increasing global trend of aging, the health of the elderly is receiving more and more attention. Exercise capacity is an important indicator of the health status of the elderly; appropriate exercise can effectively prevent many chronic diseases and improve their quality of life. However, due to the gradual decline in physical function among the elderly, scientific, objective, and continuous assessment of their exercise capacity is particularly important.

[0003] Currently, traditional motor ability assessments largely rely on manual measurement and observation by medical professionals. While these methods are accurate, they have limitations, such as complex assessment processes, dependence on specialized equipment, and high time and cost. Furthermore, older adults commonly experience declining motor skills and cognitive impairment, making it difficult for them to participate freely in complex assessment processes. Therefore, existing assessment methods struggle to achieve routine motor ability monitoring and management and are ill-suited to the individualized needs of the older population.

[0004] With the development of artificial intelligence and smart wearable devices, automated assessment methods based on sensor technology are gradually being applied in the healthcare field. By collecting motion data through sensors and using machine learning and deep learning algorithms for data analysis, it is possible to automatically assess the motor abilities of the elderly. However, most existing automated assessment algorithms based on sensor technology are complex and difficult to implement in ordinary households. Furthermore, these methods often fail to fully consider the movement characteristics and individualized needs of the elderly in real-life scenarios, thus limiting the accuracy and practicality of the assessment results.

[0005] Therefore, there is an urgent need for an automated assessment method that can be easily implemented in a home environment and is both highly accurate and efficient. Summary of the Invention

[0006] The purpose of this application is to provide an automated assessment method, device, medium, and product for the motor abilities of the elderly, which can be easily implemented in a home environment and is both highly accurate and efficient.

[0007] To achieve the above objectives, this application provides the following solution:

[0008] Firstly, this application provides an automatic assessment method for the motor function of elderly people, the automatic assessment method for the motor function of elderly people includes:

[0009] Acquire test videos of the subject; the test videos include: videos of the subject with their hands behind their head, squatting, and sitting;

[0010] Keypoint detection algorithms are used to extract skeletal keypoint data from the video under test.

[0011] Based on the skeletal keypoint data, a feature extraction network model trained on a graph convolutional network is used to determine the skeleton features;

[0012] Based on the skeleton features, a multimodal classification model trained on a long short-term memory network and a fully convolutional network is used to determine the motor ability assessment results.

[0013] Optionally, acquiring the test video of the subject specifically includes:

[0014] The subjects were filmed using video equipment;

[0015] Use OpenCV to acquire the video to be tested.

[0016] Optionally, the key point detection algorithm is the HrNet algorithm.

[0017] Optionally, the loss function of the multimodal classification model is the cross-entropy loss function.

[0018] Optionally, the test video of the subject is acquired, and the process further includes:

[0019] Perform data cleaning and data filling operations on the video to be tested.

[0020] Secondly, this application provides an automatic assessment device for the motor function of the elderly, the automatic assessment device for the motor function of the elderly comprising:

[0021] The test acquisition and input module is used to acquire the test videos of the subjects; the test videos include: videos of the subjects with their hands behind their heads, squatting, and sitting.

[0022] The motion capability feature extraction module is used to extract skeletal key point data from the video under test using a key point detection algorithm; based on the skeletal key point data, a feature extraction network model trained on a graph convolutional network is used to determine the skeleton features.

[0023] The physical motor ability assessment module is used to determine the motor ability assessment results based on skeletal features and a multimodal classification model trained on a long short-term memory network and a fully convolutional network.

[0024] Optionally, the automatic assessment device for the elderly's motor ability further includes: a physical motor ability assessment report module;

[0025] The physical activity assessment report module is used to generate an activity ability assessment report from the activity ability assessment results.

[0026] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned automatic assessment method for the motor abilities of the elderly.

[0027] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned automatic assessment method for the motor abilities of the elderly.

[0028] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned automatic assessment method for the motor abilities of the elderly.

[0029] According to the specific embodiments provided in this application, this application has the following technical effects:

[0030] This application provides an automated method, device, medium, and product for assessing the motor function of the elderly. It employs computer vision technology combined with a Graph Convolutional Network (GCN) and a multimodal network trained using Long Short-Term Memory (LSTM) and Fully Convolutional Networks for Semantic Segmentation (FCN), proposing a novel method for assessing motor function. By recording test videos, motor function assessment results for different movements can be obtained. The elderly can use this method independently at home, and the scoring is fully automated. The assessment is objective and quantifiable, aiding the elderly in home-based motor function training. Furthermore, the motor function assessment results obtained through this application have an encouraging effect on patients. Simultaneously, the scores recorded by date and the collected videos help doctors analyze the data and make corresponding training recommendations. The application of this application not only improves the scientific and personalized level of treatment but also effectively improves the quality of life for the elderly. Through continuous training and scientific assessment, the elderly can gradually improve their motor function and enhance their quality of life. At the same time, this application also provides medical personnel with a more convenient and efficient assessment tool, promoting the further development of motor function assessment for the elderly. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a diagram illustrating the application environment of an automatic assessment method for the motor abilities of the elderly, as described in one embodiment of this application.

[0033] Figure 2 This is a schematic flowchart of an automatic assessment method for the motor ability of the elderly provided in an embodiment of this application;

[0034] Figure 3 Provide a schematic diagram for the action;

[0035] Figure 4 A schematic diagram of key points in the skeleton;

[0036] Figure 5 This is a schematic diagram of the convolutional network structure of MALSTM-MFCN;

[0037] Figure 6 This is a schematic diagram of the MALSTM-MFCN model evaluation algorithm. Detailed Implementation

[0038] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0039] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] The automatic assessment method for the motor function of the elderly provided in this application can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on other servers. Terminal 102 can send the video to be tested to server 104. After receiving the video, server 104 extracts skeletal keypoint data from the video using a keypoint detection algorithm. Based on the skeletal keypoint data, it uses a feature extraction network model trained on a graph convolutional network to determine the skeleton features. Based on the skeleton features, it uses a multimodal classification model trained on a long short-term memory network and a fully convolutional network to determine the motor ability assessment result. Server 104 can feed back the obtained motor ability assessment result to terminal 102. Furthermore, in some embodiments, the automatic motor ability assessment method for the elderly can also be implemented independently by server 104 or terminal 102. For example, terminal 102 can directly process the video to be tested, or server 104 can obtain the video to be tested from the data storage system and process it.

[0041] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.

[0042] In one exemplary embodiment, such as Figure 2 As shown, an automatic assessment method for the motor ability of the elderly is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps S201 to S204. Wherein:

[0043] S201, Acquire the test video of the subject; the test video includes: videos of the subject with hands behind their head, squatting, and sitting; wherein, the three actions in the test video are determined according to the standards of the American College of Sports Medicine and are used to assess the motor function of the elderly; the action of placing hands behind the head can include elbow flexion and extension and shoulder extension; during the acquisition of the test video of the subject, subjects with decreased motor function may experience problems such as insufficient flexibility of the shoulder or elbow joints leading to inability to bend, inability to bend the knee joint normally, and dizziness when sitting and standing up. The action settings are as follows. Figure 3 As shown.

[0044] S201 specifically includes:

[0045] S11, The subject is filmed using a camera; when the subject places his / her hands behind his / her head, he / she faces the camera; when the subject squats or sits, he / she faces the camera sideways.

[0046] S12 uses OpenCV to acquire the video to be tested. OpenCV performs frame-level processing and analysis on the video. OpenCV provides rich video processing functions, including video stream acquisition, saving, and playback, and has good cross-platform compatibility, allowing for convenient video manipulation in various environments.

[0047] To maximize data utilization and automatically clean up non-standard videos in batches, thereby reducing interference from non-standard data in the overall analysis and saving time and resources spent viewing non-standard videos, S201 and later also include:

[0048] Perform data cleaning and data filling operations on the video to be tested.

[0049] The main reasons for non-standard data are twofold: firstly, the patient's movements may not be fully detected, leading to errors in feature data extraction; secondly, incorrect shooting angles may cause problems. For example, for a specific action, the subject may be facing the camera directly, but if the subject is facing the camera from the side, the movement cannot be segmented according to the rules, resulting in inaccurate evaluation. To address these two main reasons, rules are used to limit the scope of the data. If the rules are not followed, the subsequent evaluation is not performed. For cases where the skeleton is not detected, data padding is used. Since the movement changes of elderly individuals between frames are not significant, data is added forward or backward to increase the data volume. The rules are set as shown in Table 1.

[0050] Table 1

[0051]

[0052] S202, using a key point detection algorithm to extract skeletal key point data from the video under test;

[0053] Existing keypoint detection algorithms can detect 17 keypoints of the human body in photos or videos. Specifically, for the upper limbs, 6 keypoints can be identified, including the shoulder, elbow, and wrist joints; for the lower limbs, 7 keypoints can be identified, including the hip, knee, and ankle joints. The information obtained for each keypoint includes coordinates and a confidence score. The x and y coordinates represent the horizontal and vertical coordinates, respectively. The x and y coordinates indicate the horizontal and vertical position of the skeletal keypoint in the image.

[0054] In one exemplary embodiment, this application utilizes the HrNet algorithm to extract skeletal keypoint data from the video under test. The HrNet algorithm can achieve real-time detection and tracking of skeletal keypoints in elderly individuals during movement. Through deep learning technology, the HrNet algorithm can identify skeletal keypoints of the human body from the video stream in real time; skeletal keypoints include those of the arms, hands, and other parts of the body, and track their movement trajectories.

[0055] Key skeletal points represent specific points in the human body. For example, key skeletal points for lower limb movement include predetermined points on the left hip joint, right hip joint, left knee joint, right knee joint, left ankle joint, and right ankle joint. Key skeletal points are as follows: Figure 4 As shown.

[0056] S203. Based on the skeletal key point data, a feature extraction network model trained on a graph convolutional network is used to determine the skeleton features.

[0057] S203 specifically includes:

[0058] The key point data of each skeleton in each frame of the video under test are represented as nodes in the graph, and the connections between nodes are represented as edges in the graph, thus forming a graph.

[0059] Due to the sparsity and connectivity of joints in motion capability assessment, using graph convolutional neural networks for skeleton feature extraction can effectively extract spatial features during the motion process.

[0060] Graph Convolutional Networks (GCNs) are models specifically designed for processing graph data, widely used to handle dependencies between nodes and edges in graph structures. GCNs learn local and global features of nodes in a graph through graph convolution operations, capturing the relationships between nodes and their neighbors. By stacking multiple graph convolutional layers, more complex graph features are extracted progressively.

[0061] Skeletal keypoint data typically consists of multiple time frames, each containing several skeletal keypoints, each with corresponding spatial coordinates; that is, the skeletal keypoint data in each frame of the video under test. In this embodiment, human skeleton data is processed using a graph convolutional neural network to extract features. For each frame of image, each skeletal keypoint in the human skeleton data is represented as a node in a graph, and the connections between skeletal keypoints are represented as edges in the graph. After extracting the image data for each frame through the graph convolutional network, the graphs from different time frames are connected to form a new time series.

[0062] S204. Based on skeleton features, a multimodal classification model trained using a long short-term memory network and a fully convolutional network is employed to determine the motion capability assessment results. The loss function of the multimodal classification model is the cross-entropy loss function.

[0063] The evaluation algorithm for the multimodal classification model is either MLSTM-MFCN or MALSTM-MFCN. The multimodal classification model consists of a multi-length short-term memory network and a multi-scale fully convolutional network, which together form the motion capability evaluation network.

[0064] The input layer of the multimodal classification model receives skeleton features extracted by the GCN. The gate-based LSTM layer naturally extracts temporal dependencies between features, while the FCN with multi-scale filter banks can perceive spatial features of different ranges from the time-series curves. Adding an attention mechanism to the LSTM layer allows the model to better focus on temporally more important features. In the FCN, the dilated convolutional filter expands the receptive field without compressing information or increasing the number of parameters, which is beneficial for learning long-term features from high-dimensional and continuous time series. However, an excessively high dilation rate corresponds to extensive skipping of sequence nodes, which may lead to misperceptions of patterns. Therefore, the dilation rate of the multimodal classification model is less than or equal to 4. Finally, the outputs of the LSTM and FCN are concatenated to form a fully connected layer. Through linear transformation, the features are mapped to rating levels using the SoftMax activation function. The multimodal classification model is then trained, and the final output is the probability distribution of each sample belonging to each rating level, yielding the classification result, which is the motion ability assessment result.

[0065] The structure of the MALSTM-MFCN convolutional network in this application can be found in [reference needed]. Figure 5 As shown in the figure, the MLSTM-MFCN network is trained by extracting features from several sample videos using the GCN algorithm and training them with corresponding labels. The labels represent the elderly person's level of the current movement process.

[0066] Regarding the loss function used during training, since the dataset adapted to this application does not have the problem of data similarity between categories, and the number of categories at levels 0, 1, and 2 is relatively balanced, the cross-entropy loss function is used. If subsequent data imbalance occurs, it will be modified to weighted cross-entropy loss. Weighted cross-entropy loss can make the model pay more attention to these minority categories during training, thereby improving classification performance.

[0067] The MALSTM-MFCN model was trained using labeled motion data, with the motion results categorized into three levels: 0, 1, and 2. Through supervised learning, the MALSTM-MFCN model learned feature representations of normal and abnormal motion patterns. During the training process, appropriate learning rates and iteration counts were set based on the specific performance of the MALSTM-MFCN model and the distribution of the dataset, ultimately achieving good results in classification problems.

[0068] After training and obtaining the MALSTM-MFCN model (multimodal classification model) with the best performance metrics, this application extracts the coordinate information of key points of the human skeleton in the video stream and performs feature extraction on all frames using GCN to obtain new time-series data. This new data is then input into the trained MALSTM-MFCN model to obtain the motion capability evaluation result corresponding to the input video stream, achieving an automatic scoring effect. The evaluation algorithm flow of this application can be found in [link to relevant documentation]. Figure 6 .

[0069] This application provides an automated method for assessing the motor function of the elderly based on the GCN-MALSTM-MFCN model. Based on the assessment results, it can determine the motor function status of the subject's upper and lower limbs, including muscle strength, coordination, and flexibility, providing a reference and basis for medical personnel. This application utilizes GCN to extract skeletal features from existing samples and trains the MALSTM-MFCN model. The trained model can then be directly used to automatically assess new videos, making motor function assessment more efficient, and the resulting assessments can also assist medical personnel.

[0070] This application utilizes computer vision to detect the motor function of the upper and lower limbs of elderly individuals, enabling automatic assessment of their motor abilities, and aims to address the following issues:

[0071] 1. Enable automated detection and assessment of upper and lower limb movement in the elderly. Addressing the challenge of assessing motor function in the elderly, this system utilizes high-precision computer vision technology to capture the coordinates of key body points during movement, analyzes the patient's movement status, and achieves real-time monitoring and analysis of the elderly's motor abilities.

[0072] 2. Achieving objectivity and standardization in automated motor ability assessment. Traditional motor ability assessments typically rely on subjective observation and manual testing by medical personnel, thus exhibiting subjectivity and inaccuracy. This application, however, utilizes computer vision technology to achieve objective and automated detection of motor abilities in the elderly, thereby improving the objectivity and standardization of the assessment.

[0073] 3. Real-time automated assessment of motor function in the elderly. Traditional motor function assessments often require significant time and human resources, and the manual screening of large amounts of data and the instability of data collection can be a burden for physicians. This application enables real-time monitoring and analysis of motor function, greatly improving assessment efficiency and facilitating the timely detection of changes in the elderly's motor function and targeted rehabilitation interventions. By analyzing the motor function assessment results, appropriate rehabilitation training recommendations can be provided to the elderly.

[0074] 4. Achieving continuous and trackable assessment of motor function in the elderly. Decline in motor function is usually a continuous process requiring long-term tracking and assessment. This application enables continuous monitoring and long-term tracking of the motor function of the elderly, recording the subject's motor status and trends in real time. This provides data support for physicians, helping to develop personalized rehabilitation plans and treatment programs for elderly individuals with declining motor function. Real-time feedback can evaluate treatment effectiveness, allowing physicians to adjust rehabilitation plans promptly.

[0075] 5. Achieve data collection and organization in the field of motor ability assessment. Organizing a large amount of data on the elderly through databases will help further research on the algorithmic aspects of motor ability in the elderly; through the accumulation of data from a large number of subjects, combined with deep learning algorithms, we can continuously optimize multimodal classification models and improve the accuracy of diagnosis and treatment.

[0076] Based on the same inventive concept, this application also provides an automatic assessment device for elderly people's motor ability, which implements the above-mentioned automatic assessment method for elderly people's motor ability. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the automatic assessment device for elderly people's motor ability provided below can be found in the limitations of the automatic assessment method for elderly people's motor ability above, and will not be repeated here.

[0077] In one exemplary embodiment, an automatic assessment device for the motor function of the elderly is provided, comprising:

[0078] The test acquisition and input module is used to acquire the test videos of the subjects; the test videos include: videos of the subjects with their hands behind their heads, squatting, and sitting.

[0079] The motion capability feature extraction module is used to extract skeletal key point data from the video under test using a key point detection algorithm; based on the skeletal key point data, a feature extraction network model trained on a graph convolutional network is used to determine the skeleton features.

[0080] The physical motor ability assessment module is used to determine the motor ability assessment results based on skeletal features and a multimodal classification model trained on a long short-term memory network and a fully convolutional network.

[0081] In another exemplary embodiment, the automatic assessment device for the elderly's motor abilities further includes: a physical motor ability assessment report module;

[0082] The physical activity assessment report module is used to generate an activity ability assessment report from the activity ability assessment results.

[0083] In an exemplary embodiment, the test acquisition and input module further includes: a motion video input device and a motion data acquisition device; the information acquired by the motion video input device and the motion data acquisition device is input and used as the basic data and basic model of the body movement ability assessment module.

[0084] The automatic assessment device for the motor function of the elderly provided in this application has the following technical effects:

[0085] 1. The modules are simple and clear;

[0086] 2. The function of each module is clearly defined and the design is reasonable;

[0087] 3. Classifying the movements performed by the computer program avoids the subjectivity and error of human classification, making the test more accurate;

[0088] 4. Assessing the movement of the subject's upper and lower limbs separately can provide a comprehensive understanding of the subject's motor abilities and allow for targeted rehabilitation training suggestions for the upper or lower limbs.

[0089] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The I / O interfaces of the computer device are used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements an automatic assessment method for the motor abilities of the elderly.

[0090] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0091] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0092] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0093] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0094] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0095] In this application, all actions to acquire signals, information, or data are carried out in compliance with the relevant data protection laws and policies of the country where the location is situated, and with the authorization granted by the owner of the relevant device.

[0096] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0097] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. An automatic assessment method for the motor function of the elderly, characterized in that, The automatic assessment method for the motor function of the elderly includes: Acquire test videos of the subject; the test videos include: videos of the subject with their hands behind their head, squatting, and sitting; Keypoint detection algorithms are used to extract skeletal keypoint data from the video under test. Based on the skeletal keypoint data, a feature extraction network model trained on a graph convolutional network is used to determine the skeleton features; Based on skeleton features, a multimodal classification model trained using a Long Short-Term Memory (LSTM) network and a fully convolutional network is used to determine the motor ability assessment results. The multimodal classification model is divided into multiple LSTM networks and a multi-scale fully convolutional network. The skeleton features are extracted by the input layer receptive map convolutional network of the multimodal classification model. The time dependencies between features are extracted by the gate-based LSTM network, and the fully convolutional network with a multi-scale filter bank perceives spatial features of different ranges from the time series curves. An attention mechanism is added to the LSTM network. In the fully convolutional network, the dilated convolutional filter expands the receptive field without compressing information or increasing the number of parameters, which is beneficial for learning long-term features from high-dimensional and continuous time series. The dilation rate of the multimodal classification model is less than or equal to 4.

2. The automatic assessment method for the motor ability of the elderly according to claim 1, characterized in that, The acquisition of the test video of the subject specifically includes: The subjects were filmed using video equipment; Use OpenCV to acquire the video to be tested.

3. The automatic assessment method for the motor ability of the elderly according to claim 1, characterized in that, The keypoint detection algorithm is the HrNet algorithm.

4. The automatic assessment method for the motor ability of the elderly according to claim 1, characterized in that, The loss function used in the multimodal classification model is the Cross-Entropy Loss.

5. The automatic assessment method for the motor ability of the elderly according to claim 1, characterized in that, Acquire the test video of the subject, and then include: Perform data cleaning and data filling operations on the video to be tested.

6. An automatic assessment device for the motor function of the elderly, characterized in that, The automatic assessment device for the elderly's motor ability includes: The test acquisition and input module is used to acquire the test videos of the subjects; the test videos include: videos of the subjects with their hands behind their heads, squatting, and sitting. The motion capability feature extraction module is used to extract skeletal key point data from the video under test using a key point detection algorithm; based on the skeletal key point data, a feature extraction network model trained on a graph convolutional network is used to determine the skeleton features. The physical motor ability assessment module is used to determine the motor ability assessment results based on skeletal features using a multimodal classification model trained on a Long Short-Term Memory (LSTM) network and a fully convolutional network. The multimodal classification model consists of multiple LSTM networks and a multi-scale fully convolutional network. The input layer of the multimodal classification model uses a receptive graph convolutional network to extract skeletal features. A gate-based LSTM network extracts the temporal dependencies between features, and a fully convolutional network with a multi-scale filter bank perceives spatial features of different ranges from time-series curves. An attention mechanism is added to the LSTM network. In the fully convolutional network, dilated convolutional filters expand the receptive field without compressing information or increasing the number of parameters, which is beneficial for learning long-term features from high-dimensional and continuous time series. The dilation rate of the multimodal classification model is less than or equal to 4.

7. The automatic assessment device for the motor ability of the elderly according to claim 6, characterized in that, The automatic assessment device for the elderly’s motor ability also includes: a physical motor ability assessment report module; The physical activity assessment report module is used to generate an activity ability assessment report from the activity ability assessment results.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the automatic assessment method for the motor function of the elderly as described in any one of claims 1-5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the automatic assessment method for the motor function of the elderly as described in any one of claims 1-5.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the automatic assessment method for the motor function of the elderly as described in any one of claims 1-5.

Images