Method for analyzing elderly body movements based on video AI and automatically generating detection reports

Abstract

The present invention discloses a method for analyzing the body movements of the elderly based on video AI and automatically generating a test report, and relates to the technical field of intelligent physical examination for home-based elderly care. The present invention includes: a pre-detection step; performing actions according to voice prompts and obtaining video data; a cloud server automatically calculates the height, body shape and shooting distance of the elderly according to the recorded video data, and automatically analyzes and calculates the maximum angle and maximum distance of each limb of the elderly according to the AI ​​algorithm, and feeds the above data back to the body movement recognition algorithm for identification and analysis, and compares it with the standard data or scoring model of the corresponding body parameters, and automatically generates a physical examination report for the elderly. The present invention can realize accurate detection of the bending of the body joints of the elderly, the speed of the movement of the limbs, whether the body has tremors, and the corresponding tremor amplitude, without the assistance of others, flexible use, and no time scene restrictions; it has a wide market application prospect.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent physical examination for elderly people at home, and in particular, relates to a method for analyzing the body movements of the elderly based on video AI and automatically generating a detection report. Background Art

[0002] With the improvement of people's living standards, the average life expectancy has increased, and the problems of population aging and elderly care have become increasingly serious. The corresponding home-based elderly care and home medical examination industries have also risen rapidly, especially the home medical examination service industry. The testing services for the physical health of elderly people at home are becoming more and more convenient. The physical examination service for the elderly at home is mainly carried out and implemented by medical staff visiting the homes of the elderly, and the main physical examination items include routine blood pressure, blood sugar, blood lipid tests, breathing tests and functional tests of various organs of the body, which are carried out by medical staff visiting the homes of the elderly through various corresponding physical examination equipment; and the detection of the movement and behavior ability of the elderly's hands, feet and limbs and the inconvenience of early limb behavior is usually predicted and judged through the actual intuitive judgment of medical staff, and there are no corresponding special tools and equipment to realize the intuitive judgment of hand and foot movements to obtain the corresponding limb movement effects; such as CN201810205754.9, named as an invention patent for a home-based elderly care service system, which only discloses the above-mentioned corresponding routine tests, and does not have targeted limb movement and behavior tests; therefore, in response to the above problems, it is of great significance to provide a method for analyzing the limb movements of the elderly based on video AI and automatically generating a test report. Summary of the invention

[0003] The present invention provides a method for analyzing the body movements of the elderly based on video AI and automatically generating a detection report, which solves the above problems.

[0004] To solve the above technical problems, the present invention is achieved through the following technical solutions:

[0005] The method of analyzing the body movements of the elderly based on video AI and automatically generating a detection report of the present invention comprises the following steps:

[0006] S01. Pre-test steps: open the terminal, select the elderly physical function test, fill in basic information including age, weight and height, and follow the voice prompts to perform the test steps;

[0007] S02, perform actions according to voice prompts and obtain video data: the side-viewed elderly person performs neck movements, left and right upper arm movements, left and right arm movements, and left and right wrist movements in sequence according to the voice prompts, and the camera on the terminal captures the corresponding actions and stores them as video data;

[0008] S03. Upload the recorded video data to the cloud server and perform AI algorithm analysis: The cloud server automatically calculates the elderly person’s height, body shape and shooting distance based on the recorded video data, and automatically analyzes and calculates the maximum angle and maximum distance of each limb of the elderly person based on the AI ​​algorithm. The above data is fed back to the limb movement recognition algorithm for identification and analysis, and compared with the standard data or scoring model of the corresponding body parameters, automatically generates a physical examination report for the elderly person, and recommends relevant self-care plans.

[0009] Furthermore, the AI ​​algorithm specifically captures the body movements of the subject and the corresponding body sampling points and forms coordinate data corresponding to each sampling point corresponding to the coordinate system:

[0010] According to the body shape of the elderly being tested, the OpenPose technology is used to capture the user's movements in the video data, and mark them into 25 sampling points according to the body joints, and assign the two-dimensional coordinate data of the corresponding sampling point to each sampling point. The two-dimensional coordinate data is specifically (X i , Y i ), is the coordinate of the i-th sampling point; the coordinate matrix corresponding to the overall body shape of the body is A 25×2 ;

[0011]

[0012] There are 25 coordinate points in total, each with an x-coordinate and a y-coordinate. (i) Represents the i-th sampling point of the A matrix;

[0013] P2. Calculate the angle:

[0014] Use the angle formula to calculate two key angle data:

[0015]

[0016] Among them, u and v represent the required angle between two vectors, |u| and |v| represent the modulus of u and v vectors respectively;

[0017] For human body structure, if we want to calculate A (i) ,A (j) ,A (m) ,A (n) Substituting the angle between the four points into equation (1), we can get the angle θ as:

[0018]

[0019] The θ obtained at this time is a radian value. For ease of understanding, it is converted into an angle value. The conversion formula is as follows:

[0020]

[0021] Substituting (2) into the above formula, we get the formula for the human body angle:

[0022]

[0023] Through the above formula, the bending formula of each part is calculated.

[0024] Furthermore, the 25 sampling points include head marker points: 0, 15, 16, 17, 18; arm marker points: 5, 6, 7, 2, 3, 4; torso marker points: 1, 8; leg marker points: 9, 10, 11, 22, 23, 24, 12, 13, 14, 19, 20, 21.

[0025] Furthermore, for the neck bending of the tested elderly, the neck bending detection formula is obtained by the human body angle formula:

[0026]

[0027] The smaller the neck bending angle is, the greater the degree of forward bending of the neck of the elderly being tested is. This formula is used to detect the bending angle of the neck of the elderly being tested when moving forward, backward, leftward, and rightward.

[0028] Furthermore, the formula for the left upper arm raising angle of the tested elderly person is:

[0029]

[0030] The formula for the right upper arm raising angle of the tested elderly person is:

[0031]

[0032] Furthermore, the formula for the bending angle of the right arm of the tested elderly person is:

[0033]

[0034] The formula for the bending angle of the left arm of the tested elderly person is:

[0035]

[0036] Furthermore, according to the human body angle formula, the bending speed corresponding to the angle change of a certain part at two times can be calculated:

[0037]

[0038] Among them, deg1 and deg2 represent the bending angles of the same part during two motion captures, t represents the time difference between the motion captures on both sides, and F represents the bending rate. The larger the bending rate value is, the faster the bending frequency of the tested elderly person is.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The method provided by the present invention analyzes the limb movements of the elderly based on video AI and automatically generates a detection report. The terminal directly captures and identifies the limb movements of the elderly being tested and obtains the analysis results of the limb bending angles of the elderly being tested in combination with a limb bending angle algorithm. Accurate detection can be achieved for the bending of the elderly's body joints, the speed of limb movement, whether the body has tremors, and the corresponding tremor amplitude. No assistance from others is required, the method is flexible to use, and there is no restriction on time and scene. It has broad market application prospects.

[0041] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0043] Figure 1 This is a step diagram of the method of analyzing the body movements of the elderly based on video AI and automatically generating a detection report in the present invention;

[0044] Figure 2 It is a schematic diagram of marking body joints into 25 sampling points in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] See also Figure 1-2 As shown, the method of analyzing the body movements of the elderly based on video AI and automatically generating a detection report of the present invention comprises the following steps:

[0047] S01. Pre-test steps: open the terminal, select the elderly physical function test, fill in basic information including age, weight and height, and follow the voice prompts to perform the test steps;

[0048] S02, perform actions according to voice prompts and obtain video data: the side-viewed elderly person performs neck movements, left and right upper arm movements, left and right arm movements, and left and right wrist movements in sequence according to voice prompts, and the camera on the terminal captures the corresponding actions and stores them as video data; the video shooting in this process is uploaded by the user himself on the terminal, or the video is shot by a caregiver at home, and the terminal includes a mobile phone and a tablet computer;

[0049] S03. Upload the recorded video data to the cloud server and perform AI algorithm analysis: The cloud server automatically calculates the elderly person’s height, body shape and shooting distance based on the recorded video data, and automatically analyzes and calculates the maximum angle and maximum distance of each limb of the elderly person based on the AI ​​algorithm. The above data is fed back to the limb movement recognition algorithm for identification and analysis, and compared with the standard data or scoring model of the corresponding body parameters, automatically generates a physical examination report for the elderly person, and recommends relevant self-care plans.

[0050] The AI ​​algorithm is specifically: by capturing the subject's body movements and corresponding body sampling points and forming coordinate data corresponding to each sampling point corresponding to the coordinate system:

[0051] According to the body shape of the elderly being tested, OpenPose technology is used to capture the user's movements in the video data, using a method of reading 10 frames per second, and marking 25 sampling points according to body joints, and assigning each sampling point with the corresponding two-dimensional coordinate data of the sampling point. The two-dimensional coordinate data is specifically (X i ,y i ), is the coordinate of the i-th sampling point; the coordinate matrix corresponding to the overall body shape of the body is A 25×2 ;

[0052]

[0053] There are 25 coordinate points in total, each with an x-coordinate and a y-coordinate. (i) Represents the i-th sampling point of the A matrix;

[0054] P2. Calculate the angle:

[0055] Use the angle formula to calculate two key angle data:

[0056]

[0057] Among them, u and v represent the required angle between two vectors, |u| and |v| represent the modulus of u and v vectors respectively;

[0058] For human body structure, if we want to calculate A (i) ,A (j) ,A (m) ,A (n) Substituting the angle between the four points into equation (1), we can get the angle θ as:

[0059]

[0060] The θ obtained at this time is a radian value. For ease of understanding, it is converted into an angle value. The conversion formula is as follows:

[0061]

[0062] Substituting (2) into the above formula, we get the formula for the human body angle:

[0063]

[0064] Through the above formula, the bending formula of each part is calculated.

[0065] Among them, the 25 sampling points include head marker points: 0, 15, 16, 17, 18; arm marker points: 5, 6, 7, 2, 3, 4; torso marker points: 1, 8; leg marker points: 9, 10, 11, 22, 23, 24, 12, 13, 14, 19, 20, 21.

[0066] Among them, for the neck bending of the tested elderly, the neck bending detection formula is obtained through the human body angle formula:

[0067]

[0068] The smaller the neck bending angle is, the greater the degree of forward bending of the neck of the elderly being tested is. This formula is used to detect the bending angle of the neck of the elderly being tested forward, backward, left, and right.

[0069] Specifically, it is done through the following steps:

[0070] (1) Voice prompts the elderly to put their heads into the appropriate detection position;

[0071] (2) Voice prompts the elderly to nod, raise their heads, and turn left and right, and the background records the maximum range of the elderly's movements;

[0072] (3) Compare and analyze the test results with the big data in the background to generate a neck function test report for the elderly.

[0073] Among them, the formula for the left upper arm raising angle of the tested elderly is:

[0074]

[0075] The formula for the right upper arm raising angle of the tested elderly is:

[0076]

[0077] Specifically, it is done through the following steps:

[0078] (1) Voice prompts the elderly to put their heads into the appropriate detection position;

[0079] (2) Voice prompts the elderly to nod, raise their heads, and turn left and right, and the background records the maximum range of the elderly's movements;

[0080] (3) Compare and analyze the test results with the big data in the background to generate a neck function test report for the elderly.

[0081] Among them, the formula for the bending angle of the right arm of the tested elderly is:

[0082]

[0083] The formula for the left arm bending angle of the tested elderly is:

[0084]

[0085] Specifically, it is done through the following steps:

[0086] (1) The sound prompts the elderly person’s arm to enter the appropriate testing position;

[0087] (2) Voice prompts the elderly to bend their upper arm and lower arm to the maximum extent they can bend, and the background records the maximum angle of the elderly's arm bending;

[0088] (3) Compare and analyze the test results with the big data in the background to generate an arm function test report for the elderly.

[0089] For left and right wrist function testing, the specific steps include:

[0090] (1) Voice prompts the elderly to put their wrist into the appropriate detection position;

[0091] (2) Voice prompts the elderly to bend their wrists forward and backward, and the background records the frequency and amplitude of the elderly's wrist bending;

[0092] (3) Compare and analyze the test results with the big data in the background to generate a wrist function test report for the elderly.

[0093] Among them, according to the human body angle formula, the bending speed corresponding to the angle change of a certain part at two times can be calculated:

[0094]

[0095] Among them, deg1 and deg2 represent the bending angles of the same part during two motion captures, t represents the time difference between the two-side motion captures, and F represents the bending rate. The larger the bending speed value, the faster the bending frequency of the tested elderly person.

[0096] Among them, this technical solution uses big data analysis to obtain relevant models and behavior standards, including whether certain actions are completed, body or joint bending angle standards, action limb movement speed standards, whether there is tremor in the body and the tremor amplitude, and compares the collected data with the scoring model.

[0097] In the initial stage of the project, the limb analysis results of a certain number of elderly people are collected. In this specific embodiment, 100 people are used. Based on big data analysis, the final amplitude values of the neck turning forward, backward, left, and right, the height value of the upper arm lifting, the angle value, the bending angle value of the arm, the bending frequency, and the amplitude value are obtained; only the initial data volume is small, and as the number of users increases, the data volume will continue to increase. The corresponding scoring model, that is, the corresponding reference standard range, will be continuously optimized as the data volume continues to increase.

[0098] In the actual scenario, when we extract the maximum amplitude of the left and right upper arms of the elderly based on the video, due to the decline of the elderly's physical functions, tremors inevitably occur during the lifting process, which will cause an error problem for accurately calculating the maximum amplitude of the lift. In order to reduce this error, we use the following improved algorithm to select key frames.

[0099] The data set A is ordered, A = {F i |i = 1,…,N}, Sequentially connect the data points in A to form a curve in an a-dimensional space; calculate a true subset K of A that satisfies a certain error and approximates A, and at the same time, as many elements in K as possible belong to the set C; And F i ,F n ∈C.

[0100] A is represented by a two-layer data structure. The lower layer is A, and the upper layer is the true subset C of A. First, a simple single-layer curve algorithm is used in the upper layer. If the error of the result is too large, a single-layer curve algorithm is used in the lower layer, and the obtained result is added to the upper layer. Then the upper layer continues to calculate, and it recurs until a set K that meets the conditions is obtained.

[0101] Calculate the shortest distance from the i-th element in the set A to the straight line connected by the n1-th element and the n2-th element (n1 < i < n2), which can be obtained by the following formula.

[0102]

[0103] Beneficial effects:

[0104] The method provided by the present invention analyzes the limb movements of the elderly based on video AI and automatically generates a detection report. The terminal directly captures and identifies the limb movements of the elderly being tested and obtains the analysis results of the limb bending angles of the elderly being tested in combination with a limb bending angle algorithm. Accurate detection can be achieved for the bending of the elderly's body joints, the speed of limb movement, whether the body has tremors, and the corresponding tremor amplitude. No assistance from others is required, the method is flexible to use, and there is no restriction on time and scene. It has broad market application prospects.

[0105] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for analyzing the body movements of the elderly based on video AI and automatically generating a detection report, characterized in that: The steps include: S01. Pre-test steps: open the terminal, select the elderly physical function test, fill in basic information including age, weight and height, and follow the voice prompts to perform the test steps; S02, perform actions according to voice prompts and obtain video data: the side-viewed elderly person performs neck movements, left and right upper arm movements, left and right arm movements, and left and right wrist movements in sequence according to the voice prompts, and the camera on the terminal captures the corresponding actions and stores them as video data; S03. Upload the recorded video data to the cloud server and perform AI algorithm analysis: The cloud server automatically calculates the height, body shape and shooting distance of the elderly based on the recorded video data, and automatically analyzes and calculates the maximum angle and maximum distance of each limb of the elderly based on the AI ​​algorithm, and feeds the above data to the limb movement recognition algorithm for recognition and analysis, and compares it with the standard data or scoring model of the corresponding body parameters, automatically generates a physical examination report for the elderly, and recommends relevant self-care plans; The AI ​​algorithm is specifically: by capturing the body movements of the subject and the corresponding body sampling points and forming coordinate data corresponding to each sampling point corresponding to the coordinate system: According to the body shape of the elderly being tested, the OpenPose technology is used to capture the user's movements in the video data, and mark them into 25 sampling points according to the body joints, and assign the two-dimensional coordinate data of the corresponding sampling point to each sampling point. The two-dimensional coordinate data is specifically (X i , Y i ), is the coordinate of the i-th sampling point; the coordinate matrix corresponding to the overall body shape of the body is A 25×2 ; There are 25 coordinate points in total, each with an x-coordinate and a y-coordinate; P2. Calculate the angle: Use the angle formula to calculate two key angle data: Among them, u and v represent the required angle between two vectors, |u| and |v| represent the modulus of u and v vectors respectively; For human body structure, if we want to calculate A( i ),A (j) ,A (m) ,A (n) , the angle between the four points, substituting into formula (1), we can get the angle θ as: The θ obtained at this time is a radian value. For ease of understanding, it is converted into an angle value. The conversion formula is as follows: Substituting (2) into the above formula, we get the formula for the human body angle: Through the above formula, the bending formula of each part is calculated; The 25 sampling points include head marking points: 0, 15, 16, 17, 18; arm marking points: 5, 6, 7, 2, 3, 4; trunk marking points: 1, 8; leg marking points: 9, 10, 11, 22, 23, 24, 12, 13, 14, 19, 20, 21; the schematic diagram of the 25 sampling points is shown in Figure 2; For the neck bending of the tested elderly, the neck bending detection formula is obtained by the human body angle formula: The smaller the neck bending angle is, the greater the degree of forward bending of the neck of the elderly being tested is. This formula is used to detect the bending angle of the neck of the elderly being tested forward, backward, left, and right. The formula for the left upper arm raising angle of the tested elderly person is: The formula for the right upper arm raising angle of the tested elderly person is: The formula for the bending angle of the right arm of the tested elderly person is: The formula for the bending angle of the left arm of the tested elderly person is: According to the human body angle formula, the bending speed corresponding to the angle change of a certain part at two times can be calculated: Among them, deg1 and deg2 represent the bending angles of the same part during two motion captures, t represents the time difference between the motion captures on both sides, and F represents the bending rate. The larger the bending rate value is, the faster the bending frequency of the tested elderly person is.

Images