An AI-based human acupoint recognition method and system thereof

Abstract

The application provides an AI-based human body acupoint recognition method and system, and relates to the technical field of artificial intelligence, wherein the method comprises: performing real-time human body three-dimensional scanning on a user to obtain a human body three-dimensional model; tracking key subject posture landmarks on the human body three-dimensional model; based on a pre-trained AI acupoint model, identifying acupoint distribution on the human body three-dimensional model according to the key subject posture landmarks; and outputting the acupoint distribution. The AI-based human body acupoint recognition method and system perform human body three-dimensional scanning on a user, identify acupoint distribution on a human body three-dimensional model obtained through human body three-dimensional scanning based on a pre-trained AI acupoint model, assist the user in taking acupoints based on the acupoint distribution, help the user accurately find acupoints, thereby improving the effectiveness of acupoint massage, improving the massage experience, and avoiding causing discomfort or injury.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to an AI-based human acupoint recognition method and system. BACKGROUND

[0002] At present, with the widespread popularity of health management concepts, more and more people begin to pay attention to self-care, and acupoint massage, as a traditional Chinese medicine physiotherapy method, has been widely concerned and applied.

[0003] However, accurate positioning of acupoints requires high requirements, and the distribution of acupoints on the human body is complex and varies due to individual differences (such as body shape, height, posture, etc.). Many users lack sufficient professional knowledge and experience, making it difficult to accurately find target acupoints when performing self-massage. This not only reduces the effectiveness of massage, but also may affect the user's massage experience, and even cause discomfort or injury.

[0004] Therefore, there is an urgent need for a solution. SUMMARY

[0005] One of the purposes of the present application is to provide an AI-based human acupoint recognition method, which performs human three-dimensional scanning on a user, identifies the acupoint distribution on the human three-dimensional model obtained by human three-dimensional scanning based on a pre-trained AI acupoint model, and assists the user in taking acupoints based on the acupoint distribution, helping the user to accurately find acupoints, thereby improving the effectiveness of acupoint massage, improving the massage experience, and avoiding causing discomfort or injury.

[0006] The AI-based human acupoint recognition method provided by the embodiments of the present application comprises:

[0007] Performing real-time human three-dimensional scanning on a user to obtain a human three-dimensional model;

[0008] Tracking key subject posture landmarks on the human three-dimensional model;

[0009] Based on a pre-trained AI acupoint model, identifying the acupoint distribution on the human three-dimensional model according to the key subject posture landmarks;

[0010] Outputting the acupoint distribution.

[0011] Optionally, the real-time human three-dimensional scanning on the user to obtain the human three-dimensional model comprises:

[0012] Performing real-time human three-dimensional scanning on the user by a three-dimensional scanner to obtain the human three-dimensional model.

[0013] Optionally, the tracking of the key subject posture landmarks on the human three-dimensional model comprises:

[0014] Tracking key body pose landmarks on the human three-dimensional model through the Pose Landmark model.

[0015] Optionally, the Pose Landmark model at least comprises MediaPipe, Hailo, Yolo, PoseNet or BlazePose Pose Landmark model.

[0016] Optionally, the key body pose landmarks at least comprise nose, left eye inner, left eye, left eye outer, right eye inner, right eye, right eye outer, left ear, right ear, mouth left, mouth right, left shoulder, right shoulder, left elbow, right elbow, left wrist, right wrist, left pinky, right pinky, left fingers, right fingers, left thumb, right thumb, left hip, right hip, left knee, right knee, left ankle, right ankle, left heel, right heel, left toe, right toe.

[0017] Optionally, the pre-trained AI acupoint model is used to identify the acupoint distribution on the human three-dimensional model according to the key body pose landmarks.

[0018] The AI acupoint model is used to query the human acupoint meridian library according to the key body pose landmarks, to determine the relationship between the human acupoint position and the key body pose landmark.

[0019] The relationship between the human acupoint position and the key body pose landmark is used to identify the acupoint distribution on the human three-dimensional model.

[0020] Optionally, the AI-based human acupoint identification method further comprises:

[0021] The acupoint distribution is used to assist the user in acupoint massage practice.

[0022] The AI-based human acupoint identification system provided by the embodiment of the application comprises:

[0023] The human three-dimensional scanning module is used to perform real-time human three-dimensional scanning on the user, to obtain a human three-dimensional model.

[0024] The key body pose landmark tracking module is used to track key body pose landmarks on the human three-dimensional model.

[0025] The acupoint distribution identification module is used to identify the acupoint distribution on the human three-dimensional model according to the key body pose landmarks based on the pre-trained AI acupoint model.

[0026] The acupoint distribution output module is used to output the acupoint distribution.

[0027] Optionally, the human three-dimensional scanning module performs real-time human three-dimensional scanning on the user, to obtain a human three-dimensional model, which comprises:

[0028] The user is scanned in real time by a three-dimensional scanner to obtain a three-dimensional model of the human body.

[0029] Optionally, the key subject posture landmark tracking module tracks key subject posture landmarks on the three-dimensional model of the human body, including:

[0030] The key subject posture landmarks on the three-dimensional model of the human body are tracked by a Pose Landmark model.

[0031] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the application. The objects and other advantages of the present application can be realized and attained by means of the instrumentalities particularly pointed out in the written description and claims hereof.

[0032] The technical solutions of the present application will be further described in detail below with the help of the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0033] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, illustrate the present application and explain the principles of the present application, and are not intended to limit the present application. In the drawings:

[0034] Figure 1 A flowchart of an AI-based human acupoint recognition method in an embodiment of the present application;

[0035] Figure 2 A schematic diagram of key subject posture landmarks in an embodiment of the present application;

[0036] Figure 3 A schematic diagram of recognizing acupoint distribution in an embodiment of the present application;

[0037] Figure 4 A process schematic diagram of an AI-based human acupoint recognition method in an embodiment of the present application;

[0038] Figure 5 A schematic diagram of an AI-based human acupoint recognition system in an embodiment of the present application. DETAILED DESCRIPTION

[0039] The preferred embodiments of the present application will be described below with the help of the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and are not intended to limit the present application.

[0040] The present application provides an AI-based human acupoint recognition method, as shown in Figure 1 including:

[0041] S1, scanning the user in real time to obtain a three-dimensional model of the human body;

[0042] S2, track key subject posture landmarks on the human three-dimensional model;

[0043] S3, based on the pre-trained AI acupoint model, identify the acupoint distribution on the human three-dimensional model according to the key subject posture landmarks;

[0044] S4, output the acupoint distribution;

[0045] The real-time human three-dimensional scanning of the user obtains a human three-dimensional model, comprising:

[0046] The real-time human three-dimensional scanning of the user is obtained by a three-dimensional scanner, and a human three-dimensional model is obtained;

[0047] The key subject posture landmarks on the human three-dimensional model are tracked, comprising:

[0048] The key subject posture landmarks on the human three-dimensional model are tracked by a Pose Landmark model;

[0049] The Pose Landmark model at least includes: MediaPipe, Hailo, Yolo, PoseNet or BlazePose Pose Landmark model;

[0050] As shown in Figure 2 The key subject posture landmarks at least include: nose, left eye, left eye, left eye, right eye, right eye, right eye, left ear, right ear, mouth left, mouth right, left shoulder, right shoulder, left elbow, right elbow, left wrist, right wrist, left little finger, right little finger, left hand, right hand, left thumb, right thumb, left hip, right hip, left knee, right knee, left ankle, right ankle, left heel, right heel, left toe, right toe;

[0051] When the user is scanned for human three-dimensional scanning, it can be performed by a three-dimensional scanner, and after human three-dimensional scanning, a human three-dimensional model of the user's body three-dimensional simulation can be obtained; the AI acupoint model can identify the acupoint distribution on the human three-dimensional model, and the acupoint distribution refers to the distribution position of multiple human acupoints; based on the acupoint distribution, the user is assisted to take acupoints, and the acupoint refers to the selection of acupoints.

[0052] The present application scans the user for human three-dimensional scanning, identifies the acupoint distribution on the human three-dimensional model obtained by human three-dimensional scanning based on the pre-trained AI acupoint model, assists the user to take acupoints based on the acupoint distribution, helps the user to accurately find acupoints, and thus improves the effectiveness of the acupoint massage, improves the massage experience, and avoids causing discomfort or injury.

[0053] In one embodiment, based on the pre-trained AI acupoint model, the distribution of acupoints on the human body three-dimensional model is identified according to the key subject posture landmarks, including:

[0054] Based on the AI acupoint model, the relationship between the human acupoint position and the key subject posture landmark is determined by querying the human acupoint meridian library according to the key subject posture landmark.

[0055] Based on the relationship between the human acupoint position and the key subject posture landmark, the distribution of acupoints on the human body three-dimensional model is identified.

[0056] Firstly, the system uses the AI acupoint model trained by deep learning algorithm, combined with human three-dimensional scanning data, to extract the key subject posture landmarks of human body (such as shoulders, elbows, knees, etc.). These landmarks serve as reference points, providing posture information of human body in three-dimensional space.

[0057] Then, the AI model queries the pre-established human acupoint meridian library based on these key landmarks. The library contains the position data of each meridian and acupoint of human body. The AI model calculates the specific position of related meridians and acupoints according to the spatial coordinates of the landmarks, and determines the change relationship of these acupoints under specific posture.

[0058] Then, the system dynamically determines the specific position of acupoints by identifying the relative position of key landmarks on the human body three-dimensional model, combined with the meridian library data. Considering the change of posture (such as bending, stretching, etc.), the system will adjust the position of acupoints in real time, so as to accurately calibrate each acupoint.

[0059] Finally, the identified acupoint distribution will be visualized on the three-dimensional human model, helping users intuitively understand the acupoint position under different postures of human body, ensuring accurate acupoint identification and application.

[0060] As shown in Figure 3 The direction of the rigid body part object is identified by the positions of the nose, the thumb, the little finger, and the foot index finger. Based on the above landmarks and the identified orientation, the 670 acupoints of the 12 meridians can be calculated by interpolation and optimization according to the traditional acupoint atlas, as follows:

[0061] Cun is a unit of measurement for locating acupoints, which is "body inch". The measurement is always taken from the patient's hand. The width of the thumb is 1 cun, the width of two fingers is 1.5 cun, and the width of four fingers is 3 cun.

[0062] As shown in Figure 4 In specific application, the present application first determines the key subject posture landmarks of human body, and then identifies the distribution of acupoints.

[0063] In one embodiment, the AI-based human acupoint identification method further includes:

[0064] Based on the acupoint distribution, assist the user in acupoint massage practice.

[0065] In one embodiment, the pre-training step of the AI acupoint model is as follows:

[0066] Based on a machine learning algorithm, train the AI acupoint model according to the training samples; wherein the training samples at least include: a large number of test human three-dimensional models with labeled standard acupoint distribution, acupoint selection criteria, and expert acupoint selection experience.

[0067] When collecting training samples, a large number of test subjects can be scanned for human three-dimensional scanning to obtain test human three-dimensional models, and then acupoint experts can mark acupoints on the test human three-dimensional models to form standard acupoint distribution; acupoint selection criteria and expert acupoint selection experience can also be collected. After collecting the training samples, the model is trained through the machine learning algorithm to obtain the AI acupoint model, so that the AI acupoint model can independently determine the acupoint distribution based on the human three-dimensional model.

[0068] In one embodiment, the AI-based human acupoint recognition method further comprises:

[0069] Based on the acupoint distribution, assist the user in acupoint selection.

[0070] In one embodiment, the AI-based human acupoint recognition method further comprises:

[0071] When the expert remotely teaches the user acupoint massage, identify the first target acupoint selected by the user and the corresponding indication time indicated by the expert;

[0072] Generate a first time period and a second time period; wherein the first time period is a time period of a first duration after the indication time; the second time period is a second time period of a second duration after the first time period and adjacent to the first time period; the second duration is less than the first duration;

[0073] In the first time period, based on the acupoint distribution, assist the user in selecting the first target acupoint, and identify whether the user successfully selects the first target acupoint;

[0074] If the user has not successfully selected the first target acupoint at the end of the first time period, the user is assisted in selecting the first target acupoint; wherein the second assistance is more direct than the first assistance.

[0075] When the first target acupoint and the corresponding indication moment indicated by the expert are selected by the user, the expert's speech semantics or teaching actions can be used for identification, for example: the speech semantics of the expert represents that the expert instructs the user to select Baihui acupoint, and the first target acupoint is Baihui acupoint, and the indication moment is the moment when the speech semantics belongs to the speech. The first time length can be set to 25 seconds, and the second time length can be set to 8 seconds. The auxiliary direct degree represents the direct degree of prompting the first target acupoint on the user's body, for example: the auxiliary direct degree of projecting the first target acupoint on the user's body is higher than guiding the user to find the first target acupoint by himself. In a longer time after the indication moment, the first auxiliary with a relatively lower auxiliary direct degree is executed, and the user finds the first target acupoint by himself under the guidance in this process, and if it is found, the impression of the position where the first target acupoint is located is deepened, and in the next time of finding the same first target acupoint, it can be more independent and autonomous to find it. When it is identified whether the user selects the first target acupoint successfully, it can be identified whether the hand of the user presses the first target acupoint. When the first time period ends, the user still fails to select the first target acupoint, and the second auxiliary with a relatively higher auxiliary direct degree is executed, which finds the first target acupoint by himself under the guidance in the first time period and by means of direct projection marking in the second time period, and also achieves the effect of deepening the impression of the position where the first target acupoint is located, and avoids that the first target acupoint is found for too long time, causing the expert to wait for a long time, and improving the teaching experience.

[0076] In one embodiment, the first auxiliary for the user to select the first target acupoint comprises:

[0077] determining the first position of the first target acupoint from the acupoint distribution;

[0078] marking the first position on the three-dimensional model of the human body;

[0079] displaying the three-dimensional model of the human body with the marked first position to the user;

[0080] determining a plurality of reference parts from the three-dimensional model of the human body, which are within a distance threshold from the first position;

[0081] determining the second positions of the reference parts on the user's body based on the position mapping relationship between the three-dimensional model of the human body and the user's body;

[0082] obtaining the hand movement trajectory of the user;

[0083] selecting a third position to be used for reference from the second positions of the reference parts based on the hand movement trajectory;

[0084] prompting the user to select the first target acupoint based on the relative position relationship between the reference part corresponding to the third position on the three-dimensional model of the human body and the first position.

[0085] After the user is displayed the human body three-dimensional model with the first position marked, the user can start to try to select by checking the approximate position of the first target acupoint on his body. The distance threshold can be set to 30 cm. The reference part refers to a common sense part with a straight-line distance from the first position not exceeding the distance threshold, for example, the tip of the nose, the palm, the center of the eyebrows, the corner of the eye, the center of the chest, etc. The position mapping relationship between the human body three-dimensional model and the user's body refers to the real-time body position on the user's body currently corresponding to different positions on the human body three-dimensional model. When the user is scanned for the human body three-dimensional model, the position mapping relationship is generated. The hand movement trajectory of the user refers to the spatial movement trajectory continuously generated by the movement of the user's hand after the user starts to try to select. When the user is prompted to select, the reference part can be prompted, for example, the acupoint is selected at a distance of 2 fingers above the center of the eyebrows. The hand movement trajectory reflects the selection of the user, and the third position most suitable for reference can be selected based on the hand movement trajectory. Then, the user can be prompted to select the first target acupoint based on the relative position relationship between the reference part corresponding to the third position on the human body three-dimensional model and the first position, that is, the user is prompted to where the first target acupoint is in the reference part corresponding to the third position.

[0086] In the embodiment of the present application, when the first target acupoint selected by the user is assisted, the human body three-dimensional model with the first position marked is first displayed to the user, so that the user starts to try to select. The third position most suitable for reference is selected based on the hand movement trajectory of the user, and the user is prompted based on the relative position relationship between the reference part corresponding to the third position on the human body three-dimensional model and the first position. The directness of the assistance before and after is relatively large to small, which ensures that the user has a sufficient process to find the first target acupoint by himself, and the system applicability is improved by implementing appropriate assistance at different times.

[0087] In one embodiment, the second assistance to the user to select the first target acupoint comprises:

[0088] Based on the position mapping relationship between the human body three-dimensional model and the user's body, the fourth position corresponding to the first target acupoint on the user's body is determined;

[0089] When the fourth position does not fall into the non-directly visible area on the user's body, the acupoint selection prompt mark is projected to the fourth position; otherwise, the user is prompted to approach the reflection object, and the environment three-dimensional model is obtained by scanning the environment three-dimensional model of the user;

[0090] Based on the environment three-dimensional model and the human body three-dimensional model, the best standing posture of the user to view the non-directly visible area with the help of the reflection object is determined;

[0091] Based on the best standing posture, the user is prompted to view the non-directly visible area with the help of the reflection object.

[0092] project the acupoint selection prompt mark to the fourth position.

[0093] Correspondingly, the fourth position corresponding to the first target acupoint on the user's body can also be determined based on the position mapping relationship. The non-directly visible area refers to a body part that is usually not directly visible, such as the back, the back of the head, etc. The fourth position does not fall into the non-directly visible area on the user's body, which means that the user can directly see it. The acupoint selection prompt mark can be projected to the fourth position. When projecting, the light beam projection technology can be used. When performing three-dimensional scanning of the environment, a three-dimensional scanner can be used to perform three-dimensional scanning of the user's surrounding environment, and a three-dimensional model of the environment can be obtained accordingly. The reflector refers to an object that can be used to view the user's body part through reflection, such as a mirror, a reflective screen, etc. When the fourth position falls into the non-directly visible area, the user needs to use the reflector to find the first target acupoint. However, the reflector has limited reflection capability (e.g., limited by the size of the mirror), and the user may need to adjust the standing posture multiple times to be able to use the reflector to find the first target acupoint, which is not very convenient and reduces the efficiency of the user in finding the first target acupoint. Therefore, based on the three-dimensional model of the environment and the three-dimensional model of the human body, the user can be helped to determine the optimal standing posture, and based on the optimal standing posture, the user can be prompted to use the reflector to view the non-directly visible area. At this time, the acupoint selection prompt mark is projected to the fourth position again. The user can quickly find the first target acupoint in the non-directly visible area with the help of the reflector under the system prompt, which improves the convenience and the efficiency of finding the first target acupoint, and is also more humanized and intelligent.

[0094] In one embodiment, the third position to be used for reference is selected from the second positions of the reference parts based on the hand movement trajectory, comprising:

[0095] A distance curve of each second position of each reference part is generated; wherein a plurality of coordinate points are determined in a target coordinate system based on the distance between the trajectory points generated at different times of the hand movement trajectory and the same second position, and the distance curve is obtained by sequentially connecting the coordinate points; the horizontal coordinate of the target coordinate system is time, and the vertical coordinate is distance;

[0096] The time instants when the first trough value below the trough threshold on different distance curves are generated are time-sequentially sorted to obtain a generation time instant sequence;

[0097] When the time difference between the first generation time instant in the generation time instant sequence and the starting time instant of the hand movement trajectory does not exceed a first time difference threshold, and the time difference between the first generation time instant in the generation time instant sequence and the second generation time instant in the generation time instant sequence exceeds a second time difference threshold, the second position corresponding to the distance curve of the first trough value generated at the first generation time instant in the generation time instant sequence is selected as the third position;

[0098] Otherwise, the feature description processing is performed on each distance curve respectively to obtain the feature description vector of each distance curve;

[0099] The similarity between each feature description vector and the standard feature description vector is calculated.

[0100] The second position corresponding to the distance curve of the feature description vector with the largest similarity to the standard feature description vector is taken as the third position.

[0101] One second position corresponds to one distance curve, the ordinate of each coordinate point in the same distance curve represents the straight-line distance between different trajectory points on the hand movement trajectory and the same second position, the abscissa of each coordinate point in the same distance curve represents the generation time of the different trajectory points on the hand movement trajectory, and the sequential connection means connecting in the order of time. The trough value threshold can be set to 5 cm. The trough value refers to the ordinate of each trough coordinate on the distance curve. When generating the generation time sequence, the generation time of the first trough value lower than the trough threshold is found on each distance curve, and then the trough values are sorted in the order of generation time to obtain the generation time sequence. The generation of the trough represents that the user's hand continuously approaches the second position and then continuously leaves. The time difference is a positive value. The first time difference threshold can be set to 5 seconds, and the second time difference threshold can be set to 3 seconds. When the time difference between the first generation time in the generation time sequence and the starting time of the hand movement trajectory does not exceed the first time difference threshold, and the time difference between the first generation time in the generation time sequence and the second generation time in the generation time sequence exceeds the second time difference threshold, it indicates that the user's hand subjective tendency / habit approaches a certain second position in the initial stage of attempting to select, and approaches other second positions relatively later. If the certain second position approached by the user's hand subjective tendency / habit in the initial stage is taken as the third position, the impression of finding the position of the first target acupoint can be maximized, and the possibility of independently finding the first target acupoint can be maximized. Therefore, the second position corresponding to the distance curve of the first trough value generated at the first generation time in the generation time sequence is taken as the third position.

[0102] When the characteristic description processing is performed on each distance curve, the characteristics of the distance curve are extracted, and the characteristics at least include: the number of wave troughs, the average of wave trough values, etc. The characteristics are represented in the form of a vector to obtain a characteristic description vector. The standard characteristic description vector is composed of characteristics representing that the second position is suitable as the third position (for example, the number of wave troughs is 3, and the average of wave trough values is 1 cm, which represents that the user approaches the second position multiple times and the approaching distance is very close). Therefore, the second position corresponding to the distance curve of the characteristic description vector with the greatest similarity to the standard characteristic description vector is taken as the third position, which further improves the accuracy, suitability and comprehensiveness of the selection of the third position. Specifically, the standard characteristic description vector can also be set by the technician according to the actual needs in advance.

[0103] In one embodiment, the best standing posture of the user to view the non-directly visible area by means of the reflector is determined based on the environmental three-dimensional model and the human body three-dimensional model, comprising:

[0104] The standing posture of the human body three-dimensional model in the environmental three-dimensional model is simulated until the optimal constraint condition is reached;

[0105] The best standing posture is determined based on the standing posture of the human body three-dimensional model after the optimal constraint condition is reached;

[0106] The optimal constraint condition includes:

[0107] Constraint one: the complete visible model area in the reflection area of the reflector model; wherein the reflector model corresponds to the reflector, and the model area corresponds to the non-directly visible area;

[0108] Constraint two: the minimum moving cost of the viewing angle model in the human body three-dimensional model to view the reflection area is minimum.

[0109] After the optimal constraint condition is reached, the standing posture of the human body three-dimensional model is generated, i.e. the current standing posture of the human body three-dimensional model, based on which the best standing posture can be determined. In the optimal constraint condition, constraint one is used to ensure that the complete visible model area is in the reflection area of the reflector model. The reflection area refers to the area that the reflector can reflect to make the model area visible, for example, the reflecting surface of a mirror. The viewing angle model refers to the part of the human body three-dimensional model that needs to be moved to view a specific area, for example, the head of the human body. The minimum moving cost refers to the degree of change of the viewing angle model from the start of movement to the stop of movement when the viewing angle model can view the reflection area. It can be quantified as the total length of the trajectory generated by the movement of the viewing angle model during this process. Further, constraint two is used to ensure that the minimum moving cost of the viewing angle model in the human body three-dimensional model to view the reflection area is minimum, so that the best standing posture can be determined. This greatly improves the accuracy of the determination of the best standing posture and improves the applicability of the system.

[0110] The embodiment of the present application provides a human body acupoint recognition system based on AI, which comprises Figure 5 as shown, comprising:

[0111] A human body three-dimensional scanning module 1 is used for performing real-time human body three-dimensional scanning on a user to obtain a human body three-dimensional model.

[0112] A key subject posture landmark tracking module 2 is used for tracking key subject posture landmarks on the human body three-dimensional model.

[0113] An acupoint distribution recognition module 3 is used for recognizing acupoint distribution on the human body three-dimensional model according to the key subject posture landmarks based on a pre-trained AI acupoint model.

[0114] An acupoint distribution output module 4 is used for outputting the acupoint distribution.

[0115] The human body three-dimensional scanning module performs real-time human body three-dimensional scanning on a user to obtain a human body three-dimensional model, and comprises:

[0116] A three-dimensional scanner is used for performing real-time human body three-dimensional scanning on a user to obtain a human body three-dimensional model.

[0117] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the claims of the present application and the equivalent technologies thereof, the present application also intends to include these modifications and variations.

Claims

1. An AI-based method for identifying human acupoints, characterized in that, include: Real-time 3D human body scanning of users to obtain 3D human body models; Track key subject pose landmarks on a 3D human body model; Based on a pre-trained AI acupoint model, the distribution of acupoints on a 3D human body model is identified according to key subject pose landmarks. Output acupoint distribution; When an expert provides remote acupressure instruction to a user, the system identifies the first target acupoint selected by the expert and the corresponding timing. Within the first time period, based on the distribution of acupoints, the system provides initial assistance to the user in selecting the first target acupoint and identifies whether the user's selection of the first target acupoint was successful. If the user fails to select the first target acupoint by the end of the first time period, a second assistance will be provided to help the user select the first target acupoint. Determine the first location of the first target acupoint from the distribution of acupoints; Mark the first position on the 3D human body model; Display the human body 3D model with the first position marked to the user; From the 3D human body model, determine multiple reference locations whose distance from the first location does not exceed a distance threshold; Based on the positional mapping relationship between the human body 3D model and the user's body, the second position corresponding to each reference part on the user's body is determined. Obtain the user's hand movement trajectory; Based on the hand movement trajectory, a third position to be used for reference is selected from the second positions of each reference part; Based on the relative positional relationship between the reference part corresponding to the third position on the human body 3D model and the first position, prompts are given to the user to select the first target acupoint; The process of selecting a third reference position from the second positions of various reference parts based on the hand movement trajectory includes: Generate the spacing curves for the second positions of each reference part; wherein, based on the distance between the trajectory points generated by the hand movement trajectory at different times and the same second position, multiple coordinate points are determined in the target coordinate system, and the spacing curves are obtained by connecting the coordinate points in sequence; the horizontal axis of the target coordinate system is time, and the vertical axis is distance; The generation times of the first trough value below the trough threshold on different spacing curves are sorted in time sequence to obtain the generation time sequence. When the time difference between the first generation time in the generation time sequence and the start time of the hand movement trajectory does not exceed the first time difference threshold, and the time difference between the first generation time in the generation time sequence and the second generation time in the generation time sequence exceeds the second time difference threshold, the second position corresponding to the spacing curve of the first trough value generated at the first generation time in the generation time sequence will be used as the third position. Otherwise, feature description processing is performed on each spacing curve separately to obtain the feature description vector of each spacing curve. Calculate the similarity between each feature description vector and the standard feature description vector; The second position corresponding to the spacing curve of the feature description vector that has the greatest similarity to the standard feature description vector is taken as the third position.

2. The AI-based acupoint recognition method as described in claim 1, characterized in that, The process of performing real-time 3D human body scanning on the user to obtain a 3D human body model includes: A 3D scanner is used to perform real-time 3D scanning of the user's body to obtain a 3D model of the human body.

3. The AI-based human acupoint recognition method as described in claim 1, characterized in that, The key subject pose landmarks on the human 3D model being tracked include: Track key subject pose landmarks on a 3D human body model using the Pose Landmark model.

4. The AI-based human acupoint recognition method as described in claim 3, characterized in that, The Pose Landmark model includes at least the MediaPipe Pose Landmark model.

5. The AI-based human acupoint recognition method as described in claim 3, characterized in that, The key subject posture landmarks include at least the following: nose, left inner eye, left eye, left outer eye, right inner eye, right eye, right outer eye, left ear, right ear, left side of mouth, right side of mouth, left shoulder, right shoulder, left elbow, right elbow, left wrist, right wrist, left little finger, right little finger, left fingers, right fingers, left thumb, right thumb, left hip, right hip, left knee, right knee, left ankle, right ankle, left heel, right heel, left toe, and right toe.

6. The AI-based human acupoint recognition method as described in claim 1, characterized in that, The aforementioned AI acupoint model, based on pre-trained technology, identifies the distribution of acupoints on a 3D human body model according to key subject pose landmarks, including: Based on the AI ​​acupoint model, the relationship between the location of human acupoints and key subject posture landmarks is determined by querying the human acupoint and meridian database. Based on the relationship between the location of acupoints on the human body and key subject posture landmarks, the distribution of acupoints on the three-dimensional human body model is identified.

7. The AI-based human acupoint recognition method as described in claim 1, characterized in that, Also includes: Based on the distribution of acupoints, it assists users in practicing acupoint massage.

8. An AI-based human acupoint recognition system, characterized in that, include: The human body 3D scanning module is used to perform real-time human body 3D scanning on users to obtain a human body 3D model; The key subject pose landmark tracking module is used to track key subject pose landmarks on a 3D human body model; The acupoint distribution recognition module is used to identify the distribution of acupoints on a 3D human body model based on a pre-trained AI acupoint model and key subject pose landmarks. The acupoint distribution output module is used to output the acupoint distribution. The acupoint selection assistance module is used for: When an expert provides remote acupressure instruction to a user, the system identifies the first target acupoint selected by the expert and the corresponding timing. Within the first time period, based on the distribution of acupoints, the system provides initial assistance to the user in selecting the first target acupoint and identifies whether the user's selection of the first target acupoint was successful. If the user fails to select the first target acupoint by the end of the first time period, a second assistance will be provided to help the user select the first target acupoint. Determine the first location of the first target acupoint from the distribution of acupoints; Mark the first position on the 3D human body model; Display the human body 3D model with the first position marked to the user; From the 3D human body model, determine multiple reference locations whose distance from the first location does not exceed a distance threshold; Based on the positional mapping relationship between the human body 3D model and the user's body, the second position corresponding to each reference part on the user's body is determined. Obtain the user's hand movement trajectory; Based on the hand movement trajectory, a third position to be used for reference is selected from the second positions of each reference part; Based on the relative positional relationship between the reference part corresponding to the third position on the human body 3D model and the first position, prompts are given to the user to select the first target acupoint; The process of selecting a third reference position from the second positions of various reference parts based on the hand movement trajectory includes: Generate the spacing curves for the second positions of each reference part; wherein, based on the distance between the trajectory points generated by the hand movement trajectory at different times and the same second position, multiple coordinate points are determined in the target coordinate system, and the spacing curves are obtained by connecting the coordinate points in sequence; the horizontal axis of the target coordinate system is time, and the vertical axis is distance; The generation times of the first trough value below the trough threshold on different spacing curves are sorted in time sequence to obtain the generation time sequence. When the time difference between the first generation time in the generation time sequence and the start time of the hand movement trajectory does not exceed the first time difference threshold, and the time difference between the first generation time in the generation time sequence and the second generation time in the generation time sequence exceeds the second time difference threshold, the second position corresponding to the spacing curve of the first trough value generated at the first generation time in the generation time sequence will be used as the third position. Otherwise, feature description processing is performed on each spacing curve separately to obtain the feature description vector of each spacing curve. Calculate the similarity between each feature description vector and the standard feature description vector; The second position corresponding to the spacing curve of the feature description vector that has the greatest similarity to the standard feature description vector is taken as the third position.

9. The AI-based human acupoint recognition system as described in claim 8, characterized in that, The human body 3D scanning module performs real-time human body 3D scanning on the user to obtain a human body 3D model, including: A 3D scanner is used to perform real-time 3D scanning of the user's body to obtain a 3D model of the human body.

10. The AI-based human acupoint recognition system as described in claim 8, characterized in that, The key subject pose landmark tracking module tracks key subject pose landmarks on the human 3D model, including: Track key subject pose landmarks on a 3D human body model using the Pose Landmark model.

Images