Ataxia judgment method and system based on image information

Abstract

The invention provides a method and a system for judging ataxia based on image information. The method comprises the following steps: displaying a guide video and a guide action at the front end of an applet; uploading a finger and nose action video recorded according to the guidance video; performing framing processing on the finger and nose action video, and extracting each frame of picture; key point information of each frame of picture is positioned and analyzed, and a position relation is obtained; according to the position relationship, judging the accuracy and continuity of the finger and nose actions to obtain a judgment conclusion; and returning the judgment conclusion to the applet page and explaining in detail. According to the method, an ataxia diagnosis algorithm is adopted, position information of key points in an image is extracted, the situation that the position relation changes along with time is analyzed, an ataxia judgment conclusion is obtained, a cloud platform architecture is adopted, the accuracy and smoothness of finger and nose actions of a patient are analyzed through a deep learning algorithm deployed on a cloud platform, ataxia judgment is assisted, and the accuracy and smoothness of the finger and nose actions are improved. Influences of subjective factors are eliminated, hardware requirements are simplified, and therefore real-time diagnosis of ataxia is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of smart medical care, and specifically, relates to an ataxia judgment method and system based on image information. Background Art

[0002] Currently, the diagnosis of ataxia after a stroke mainly relies on the experience of clinicians and the patient's physical signs, such as the patient's walking style, finger-nose test, or ability to complete specific tasks. Manual clinical diagnosis methods are subjective and difficult to diagnose accurately in the early stages. In addition, although conventional monitoring systems can record patients' behavioral patterns, there are delays in data processing and real-time feedback, making it difficult to achieve real-time monitoring and identification of ataxia symptoms.

[0003] In the prior art, the high-precision movement disorder monitoring device disclosed in the patent document "A movement disorder monitoring device based on laser radar" (CN118078267A) can locate the position of movement disorder patients from the scene and monitor their movement trajectory, but it has extremely high hardware requirements, is difficult to popularize, and lacks interactivity and real-time performance.

[0004] The recognition method disclosed in the patent document "Unnoticeable Hand Movement Recognition Method for Stroke Patients Based on Deep Learning" (CN115311737A) aims to evaluate the recovery of upper limbs of stroke patients by recognizing hand movements, using non-contact devices such as Kinect sensors to collect hand movement data of patients and perform visual-based movement recognition. The present invention collects corresponding data for the diagnosis of ataxia, and analyzes the accuracy and coherence of movements based on image-based deep learning algorithms. Summary of the invention

[0005] In view of the defects in the prior art, an object of the present invention is to provide a method and system for determining ataxia based on image information.

[0006] According to the present invention, a method for determining ataxia based on image information includes:

[0007] Step S1: Displaying a guidance video on the front end of the mini program to guide the action;

[0008] Step S2: Upload the finger-to-nose action video recorded according to the instruction video;

[0009] Step S3: Process the finger-to-nose action video frame by frame and extract each frame of the image;

[0010] Step S4: locate and analyze key point information for each frame of the image to obtain the position relationship;

[0011] Step S5: judging the accuracy and continuity of the finger-to-nose action according to the position relationship, and obtaining a judgment conclusion;

[0012] Step S6: Return the judgment conclusion to the mini program page and explain it in detail.

[0013] Preferably, the instruction video is an instruction video explaining the finger-to-nose test action requirements and precautions, guiding the user to complete the finger-to-nose action in a standardized manner using the left and right hands respectively.

[0014] In step S2, a common camera or a camera built into a mobile device is used to record a finger-to-nose action video.

[0015] In step S3, the video stream uploaded to the cloud platform is processed by frame division, and the video is decomposed into separate frames.

[0016] In step S6, the judgment conclusion obtained by the cloud platform analysis is returned to the mini program page, providing a detailed explanation.

[0017] Preferably, in step S4, an open source posture estimation algorithm framework is used to extract key points from each frame of the 2D image, and the two-dimensional coordinates (x, y) of these key points in the image are output, and the hand displacement s hand .

[0018] Analyze the relative position relationship between the face and the hand and the relationship between the distance changes over time in adjacent time periods. Through geometric calculation and time series analysis, the relative position relationship of key points is converted into the corresponding motion pattern to form a time series function, the Euclidean distance

[0019] Hand movement speed

[0020] Among them, x face (t) represents the x-axis coordinate of the facial key point at time t;

[0021] y face (t) represents the y-axis coordinate of the facial key point at time t;

[0022] x hand (t) represents the x-axis coordinate of the key point of the hand at time t;

[0023] y hand (t) represents the y-axis coordinate of the key point of the hand at time t;

[0024] t represents time.

[0025] Preferably, the judgment in step S5 includes accuracy judgment and consistency judgment;

[0026] In the accuracy judgment, if the Euclidean distance D(t) between the hand and the nose tip is less than 10 cm, and the finger can accurately touch the nose tip, and the position changes smoothly during the entire action, then it is determined that the accuracy requirement is met;

[0027] If the Euclidean distance D(t) between the hand and the nose tip is greater than or equal to 10 cm, or the finger cannot accurately touch the nose tip, or the position change is not stable during the entire action, it is determined that the accuracy requirement is not met.

[0028] In the continuity judgment, if the hand movement speed v is within 0.5 m / s and there is no sudden pause or rapid retreat, it means that the movement is smooth and it is determined that the continuity requirement is met;

[0029] If the hand movement speed v is greater than or equal to 0.5 m / s, or the fingers fail to touch the tip of the nose and there is obvious tremor or pause in the hand movement, it indicates that the movement is not smooth and is judged to not meet the continuity requirement.

[0030] Preferably, a cloud platform architecture is adopted, and the finger-nose movement information collected by the front-end applet is uploaded to the cloud platform. The cloud platform processes the deep learning algorithm and adopts the ataxia diagnosis algorithm to extract the position information of the key points of the face and hands in each frame of the picture, and uses the ataxia judgment rule model data analysis to obtain the judgment conclusion and then return it to the front end.

[0031] According to the present invention, a system for determining ataxia based on image information is provided, comprising:

[0032] Module M1: Display instructional videos on the front end of the mini program to guide actions;

[0033] Module M2: Upload the finger-to-nose action video recorded according to the instruction video;

[0034] Module M3: Process the finger-to-nose action video frame by frame and extract each frame of the image;

[0035] Module M4: locate and analyze key point information for each frame to obtain positional relationship;

[0036] Module M5: Based on the positional relationship, determine the accuracy and continuity of the finger-nose movement and obtain the judgment conclusion;

[0037] Module M6: Return the judgment conclusion to the mini program page and explain it in detail.

[0038] Preferably, the instruction video is an instruction video explaining the finger-to-nose test action requirements and precautions, guiding the user to complete the finger-to-nose action in a standardized manner using the left and right hands respectively.

[0039] The module M2 uses a common camera or a camera built into a mobile device to record the finger-to-nose action video.

[0040] The module M3 processes the video stream uploaded to the cloud platform in frames, and decomposes the video into separate frames.

[0041] In the module M6, the judgment conclusion obtained by the cloud platform analysis is returned to the mini program page, and a detailed explanation is provided.

[0042] Preferably, the module M4 uses an open source posture estimation algorithm framework to extract key points from each frame of the 2D image and output the two-dimensional coordinates (x, y) of these key points in the image, the hand displacement s hand .

[0043] Analyze the relative position relationship between the face and the hand and the relationship between the distance changes over time in adjacent time periods. Through geometric calculation and time series analysis, the relative position relationship of key points is converted into the corresponding motion pattern to form a time series function, the Euclidean distance

[0044] Hand movement speed

[0045] Among them, x face (t) represents the x-axis coordinate of the facial key point at time t;

[0046] y face (t) represents the y-axis coordinate of the facial key point at time t;

[0047] x hand (t) represents the x-axis coordinate of the key point of the hand at time t;

[0048] y hand (t) represents the y-axis coordinate of the key point of the hand at time t;

[0049] t represents time.

[0050] Preferably, the judgment in the module M5 includes accuracy judgment and coherence judgment;

[0051] In the accuracy judgment, if the Euclidean distance D(t) between the hand and the nose tip is less than 10 cm, and the finger can accurately touch the nose tip, and the position changes smoothly during the entire action, then it is determined that the accuracy requirement is met;

[0052] If the Euclidean distance D(t) between the hand and the nose tip is greater than or equal to 10 cm, or the finger cannot accurately touch the nose tip, or the position change is not stable during the entire action, it is determined that the accuracy requirement is not met.

[0053] In the continuity judgment, if the hand movement speed v is within 0.5 m / s and there is no sudden pause or rapid retreat, it means that the movement is smooth and it is determined that the continuity requirement is met;

[0054] If the hand movement speed v is greater than or equal to 0.5 m / s, or the fingers fail to touch the tip of the nose and there is obvious tremor or pause in the hand movement, it indicates that the movement is not smooth and is judged to not meet the continuity requirement.

[0055] Preferably, a cloud platform architecture is adopted, and the finger-nose movement information collected by the front-end applet is uploaded to the cloud platform. The cloud platform processes the deep learning algorithm and adopts the ataxia diagnosis algorithm to extract the position information of the key points of the face and hands in each frame of the picture, and uses the ataxia judgment rule model data analysis to obtain the judgment conclusion and then return it to the front end.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] 1. The present invention adopts a cloud platform architecture to upload the action information collected by the front-end applet to the cloud platform, and draws conclusions through cloud platform algorithm analysis and then returns them to the front end, thereby simplifying hardware requirements, greatly reducing hardware costs and usage thresholds, and facilitating popularization.

[0058] 2. The present invention adopts an ataxia diagnosis algorithm to extract the position information of key points of the face and hands in the image, analyzes the relationship between the distance between the face and the hands changing over time, obtains the ataxia judgment conclusion, eliminates the influence of subjective factors, and makes the diagnosis more accurate.

[0059] 3. The present invention uses a deep learning algorithm deployed on a cloud platform to analyze the accuracy and fluency of the patient's finger-nose movements, assist in the determination of ataxia, reduce the burden on local computing devices, and achieve real-time analysis and instant feedback. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0061] Figure 1 Schematic diagram of the flow of the ataxia judgment method based on image information. DETAILED DESCRIPTION

[0062] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0063] According to the present invention, a method for determining ataxia based on image information is provided. Figure 1 For example, including:

[0064] Step S1: On the front end of the mini program, a video explaining the finger-to-nose test action requirements and precautions is displayed to guide the user to complete the finger-to-nose action with the left and right hands respectively.

[0065] Step S2: According to the instruction video, the user independently completes the recording of the finger-to-nose action process through the front camera and uploads the video to the cloud platform.

[0066] Currently, most methods use non-contact Kinect or similar special sensors to collect patient hand motion data. Although they can capture fine movements, they require high hardware configuration and technical support. For home users or primary medical institutions, the cost is high and the operation is complicated, so it is not popular. By using ordinary cameras or mobile devices to record finger-nose motion videos, without the need for professional sensor devices like Kinect, this method greatly reduces hardware costs and usage barriers. Any device with a camera function can participate in the detection, and the user operation is simple and convenient.

[0067] Step S3: Process the video stream uploaded to the cloud platform in frames, extracting each frame for separate analysis.

[0068] Directly processing videos requires high computing resources, especially when processing long videos or high frame rate videos, the computing overhead is huge. Therefore, through frame processing, the video is decomposed into individual frames, thereby reducing the processing complexity, facilitating distributed processing or parallel computing, further speeding up the processing speed and improving the responsiveness of the system.

[0069] At the same time, frame processing can better cope with videos with different frame rates. Because each frame of the picture will be processed and analyzed independently, the system can maintain consistent analysis accuracy regardless of how the frame rate changes during video recording. Direct video processing may face compatibility issues under different frame rates or video formats.

[0070] Step S4: Use a deep learning algorithm to locate the key points of the face and hands in each frame of the image, analyze the relative position relationship between the two, and the changes in the position relationship between the two in adjacent time periods.

[0071] By adopting the ataxia diagnosis algorithm, the position information of the key points of the face and hands in the image is extracted, and the relationship between the distance between the face and the hands changing with time is analyzed.

[0072] Using an open source posture estimation algorithm framework, human key points are extracted from 2D images. Facial feature points on the face, including eyes, nose, mouth, etc., and hand key points, including finger joints, palms, etc., can be identified, and the 2D coordinates (x, y) of these key points in the image and the hand displacement s can be output. hand .

[0073] After extracting the key point position information, the relative position relationship of the key points is converted into the corresponding motion pattern through geometric calculation and time series analysis.

[0074] In each frame, the relative position between the face and hand key points can be described by calculating the Euclidean distance between them. Assume that the coordinates of the facial key points are (x face ,y face ), the coordinates of the key points of the hand are (x hand ,y hand ), over time, the Euclidean distance D(t) between the key points of the hand and the face will change with different frames. By recording these changes, a time series function D(t) is formed:

[0075]

[0076] Among them, x face (t) represents the x-axis coordinate of the facial key point at time t;

[0077] y face (t) represents the y-axis coordinate of the facial key point at time t;

[0078] x hand (t) represents the x-axis coordinate of the key point of the hand at time t;

[0079] y hand (t) represents the y-axis coordinate of the key point of the hand at time t;

[0080] t represents time.

[0081] Hand movement speed

[0082] Step S5: According to the positional relationship between the hand and the face, the accuracy and continuity of the user's finger-to-nose action are judged using the ataxia judgment rule model to obtain a judgment conclusion on whether the user has ataxia.

[0083] The ataxia judgment rule model includes:

[0084] Accuracy judgment: If the positional relationship between the hand and the face remains within a certain distance range, that is, the Euclidean distance D(t) between the hand and the tip of the nose is less than 10 cm, and the fingers can accurately touch the tip of the nose, and the position changes smoothly during the entire action, then it is judged that the accuracy requirements are met. If any point is not achieved, it is judged that the accuracy requirements are not met.

[0085] Continuity judgment: If the position change between the hand and the face is smooth and the speed change is not large, that is, the hand movement speed v is within 0.5m / s, and there is no sudden pause or rapid retreat, then the movement is smooth and it is judged to meet the continuity requirements. If the distance between the hand and the face changes greatly, or the fingers fail to touch the tip of the nose and there is obvious tremor or pause in the hand movement, it indicates that the movement is not smooth and it is judged to not meet the continuity requirements.

[0086] Step S6: Return the conclusions obtained from the cloud platform analysis to the mini program page, and provide detailed explanations to help users understand their physical condition.

[0087] By adopting the cloud platform architecture, the motion information collected by the front-end applet is uploaded to the cloud platform, and the cloud platform processes the deep learning algorithm and data analysis, and the conclusion is returned to the front-end, thereby achieving the purpose of real-time diagnosis of ataxia, which greatly reduces the computing requirements of local devices. Patients use the applet on the front end to upload videos, and the back-end cloud platform processes and provides results in real time. This design not only improves the diagnosis speed, but also reduces the burden on local computing devices, and can achieve real-time analysis and instant feedback.

[0088] The present invention also provides an ataxia judgment system based on image information, and the ataxia judgment system based on image information can be realized by executing the process steps of the ataxia judgment method based on image information, that is, those skilled in the art can understand the ataxia judgment method based on image information as a preferred implementation of the ataxia judgment system based on image information.

[0089] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0090] According to the present invention, a system for determining ataxia based on image information is provided, comprising:

[0091] Module M1: Display instructional videos on the front end of the mini program to guide actions;

[0092] Module M2: Upload the finger-to-nose action video recorded according to the instruction video;

[0093] Module M3: Process the finger-to-nose action video frame by frame and extract each frame of the image;

[0094] Module M4: locate and analyze key point information for each frame to obtain positional relationship;

[0095] Module M5: Based on the positional relationship, determine the accuracy and continuity of the finger-nose movement and obtain the judgment conclusion;

[0096] Module M6: Return the judgment conclusion to the mini program page and explain it in detail.

[0097] In more preferred examples, the instruction video is an instruction video explaining the finger-to-nose test action requirements and precautions, guiding the user to complete the finger-to-nose action with the left and right hands respectively.

[0098] The module M2 uses a common camera or a camera built into a mobile device to record the finger-to-nose action video.

[0099] The module M3 processes the video stream uploaded to the cloud platform in frames, and decomposes the video into separate frames.

[0100] In the module M6, the judgment conclusion obtained by the cloud platform analysis is returned to the mini program page, and a detailed explanation is provided.

[0101] In more preferred embodiments, the module M4 uses an open source posture estimation algorithm framework to extract key points from each frame of the 2D image and output the two-dimensional coordinates (x, y) of these key points in the image, the hand displacement s hand .

[0102] Analyze the relative position relationship between the face and the hand and the relationship between the distance changes over time in adjacent time periods. Through geometric calculation and time series analysis, the relative position relationship of key points is converted into the corresponding motion pattern to form a time series function, the Euclidean distance

[0103] Hand movement speed

[0104] Among them, x face (t) represents the x-axis coordinate of the facial key point at time t;

[0105] y face (t) represents the y-axis coordinate of the facial key point at time t;

[0106] x hand (t) represents the x-axis coordinate of the key point of the hand at time t;

[0107] y hand (t) represents the y-axis coordinate of the key point of the hand at time t;

[0108] t represents time.

[0109] In more preferred embodiments, the judgment in the module M5 includes accuracy judgment and coherence judgment;

[0110] In the accuracy judgment, if the Euclidean distance D(t) between the hand and the nose tip is less than 10 cm, and the finger can accurately touch the nose tip, and the position changes smoothly during the entire action, then it is determined that the accuracy requirement is met;

[0111] If the Euclidean distance D(t) between the hand and the nose tip is greater than or equal to 10 cm, or the finger cannot accurately touch the nose tip, or the position change is not stable during the entire action, it is determined that the accuracy requirement is not met.

[0112] In the continuity judgment, if the hand movement speed v is within 0.5 m / s and there is no sudden pause or rapid retreat, it means that the movement is smooth and it is determined that the continuity requirement is met;

[0113] If the hand movement speed v is greater than or equal to 0.5 m / s, or the fingers fail to touch the tip of the nose and there is obvious tremor or pause in the hand movement, it indicates that the movement is not smooth and is judged to not meet the continuity requirement.

[0114] In more preferred examples, a cloud platform architecture is used to upload the finger-nose movement information collected by the front-end applet to the cloud platform. The cloud platform processes the deep learning algorithm and adopts the ataxia diagnosis algorithm to extract the position information of the key points of the face and hands in each frame of the image. The ataxia judgment rule model data is used for data analysis to obtain the judgment conclusion and then return it to the front end.

[0115] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A method for determining ataxia based on image information, characterized in that: include: Step S1: Displaying a guidance video on the front end of the mini program to guide the action; Step S2: Upload the finger-to-nose action video recorded according to the instruction video; Step S3: Process the finger-to-nose action video frame by frame and extract each frame of the image; Step S4: locate and analyze key point information for each frame of the image to obtain the position relationship; Step S5: judging the accuracy and continuity of the finger-to-nose action according to the position relationship, and obtaining a judgment conclusion; Step S6: Return the judgment conclusion to the mini program page and explain it in detail.

2. The method for determining ataxia based on image information according to claim 1, characterized in that: The instructional video is a video explaining the requirements and precautions of the finger-to-nose test, and instructs the user to complete the finger-to-nose action in a standardized manner with the left and right hands respectively; In step S2, a common camera or a camera built into a mobile device is used to record a finger-to-nose action video; In step S3, the video stream uploaded to the cloud platform is processed by frame division to decompose the video into separate frames; In step S6, the judgment conclusion obtained by the cloud platform analysis is returned to the mini program page, providing a detailed explanation.

3. The method for determining ataxia based on image information according to claim 1, characterized in that: In step S4, an open source posture estimation algorithm framework is used to extract key points from each frame of the 2D image, and the two-dimensional coordinates (x, y) of these key points in the image are output, and the hand displacement s hand ; Analyze the relative position relationship between the face and the hand and the relationship between the distance changes over time in adjacent time periods. Through geometric calculation and time series analysis, the relative position relationship of key points is converted into the corresponding motion pattern to form a time series function, the Euclidean distance Hand movement speed Among them, x face (t) represents the x-axis coordinate of the facial key point at time t; y face (t) represents the y-axis coordinate of the facial key point at time t; x hand (t) represents the x-axis coordinate of the key point of the hand at time t; y hand (t) represents the y-axis coordinate of the key point of the hand at time t; t represents time.

4. The method for determining ataxia based on image information according to claim 1, characterized in that: The judgment in step S5 includes accuracy judgment and consistency judgment; In the accuracy judgment, if the Euclidean distance D(t) between the hand and the nose tip is less than 10 cm, and the finger can accurately touch the nose tip, and the position changes smoothly during the entire action, then it is determined that the accuracy requirement is met; If the Euclidean distance D(t) between the hand and the nose tip is greater than or equal to 10 cm, or the finger cannot accurately touch the nose tip, or the position changes unsteadily during the entire action, it is determined that the accuracy requirement is not met; In the continuity judgment, if the hand movement speed v is within 0.5 m / s and there is no sudden pause or rapid retreat, it means that the movement is smooth and it is determined that the continuity requirement is met; If the hand movement speed v is greater than or equal to 0.5 m / s, or the fingers fail to touch the tip of the nose and there is obvious tremor or pause in the hand movement, it indicates that the movement is not smooth and is judged to not meet the continuity requirement.

5. The method for determining ataxia based on image information according to claim 1, characterized in that: Using a cloud platform architecture, the finger-nose movement information collected by the front-end applet is uploaded to the cloud platform. The cloud platform processes the deep learning algorithm and adopts the ataxia diagnosis algorithm to extract the position information of the key points of the face and hands in each frame of the image. The ataxia judgment rule model data is used for data analysis to obtain the judgment conclusion and then return it to the front end.

6. An ataxia judgment system based on image information, characterized in that: include: Module M1: Display instructional videos on the front end of the mini program to guide actions; Module M2: Upload the finger-to-nose action video recorded according to the instruction video; Module M3: Process the finger-to-nose action video frame by frame and extract each frame of the image; Module M4: locate and analyze key point information of each frame to obtain positional relationship; Module M5: Based on the positional relationship, determine the accuracy and continuity of the finger-nose movement and obtain the judgment conclusion; Module M6: Return the judgment conclusion to the mini program page and explain it in detail.

7. The ataxia determination system based on image information according to claim 6, characterized in that: The instructional video is a video explaining the requirements and precautions of the finger-to-nose test, and instructs the user to complete the finger-to-nose action in a standardized manner with the left and right hands respectively; The module M2 uses a common camera or a camera built into a mobile device to record a finger-to-nose action video; The module M3 processes the video stream uploaded to the cloud platform by frame division, and decomposes the video into individual frames; In the module M6, the judgment conclusion obtained by the cloud platform analysis is returned to the mini program page, and a detailed explanation is provided.

8. The ataxia determination system based on image information according to claim 6, characterized in that: The module M4 uses an open source posture estimation algorithm framework to extract key points from each frame of the 2D image and output the two-dimensional coordinates (x, y) of these key points in the image and the hand displacement s. hand ; Analyze the relative position relationship between the face and the hand and the relationship between the distance changes over time in adjacent time periods. Through geometric calculation and time series analysis, the relative position relationship of key points is converted into the corresponding motion pattern to form a time series function, the Euclidean distance Hand movement speed Among them, x face (t) represents the x-axis coordinate of the facial key point at time t; y face (t) represents the y-axis coordinate of the facial key point at time t; x hand (t) represents the x-axis coordinate of the key point of the hand at time t; y hand (t) represents the y-axis coordinate of the key point of the hand at time t; t represents time.

9. The ataxia determination system based on image information according to claim 6, characterized in that: The judgment in the module M5 includes accuracy judgment and coherence judgment; In the accuracy judgment, if the Euclidean distance D(t) between the hand and the nose tip is less than 10 cm, and the finger can accurately touch the nose tip, and the position changes smoothly during the entire action, then it is determined that the accuracy requirement is met; If the Euclidean distance D(t) between the hand and the nose tip is greater than or equal to 10 cm, or the finger cannot accurately touch the nose tip, or the position changes unsteadily during the entire action, it is determined that the accuracy requirement is not met; In the continuity judgment, if the hand movement speed v is within 0.5 m / s and there is no sudden pause or rapid retreat, it means that the movement is smooth and it is determined that the continuity requirement is met; If the hand movement speed v is greater than or equal to 0.5 m / s, or the fingers fail to touch the tip of the nose and there is obvious tremor or pause in the hand movement, it indicates that the movement is not smooth and is judged to not meet the continuity requirement.

10. The ataxia determination system based on image information according to claim 6, characterized in that: Using a cloud platform architecture, the finger-nose movement information collected by the front-end applet is uploaded to the cloud platform. The cloud platform processes the deep learning algorithm and adopts the ataxia diagnosis algorithm to extract the position information of the key points of the face and hands in each frame of the image. The ataxia judgment rule model data is used for data analysis to obtain the judgment conclusion and then return it to the front end.

Images