Real-time intelligent fitness exercise guidance method, system and device
By obtaining the user's exercise video and heart rate information in real time, combining the mediapipe framework and neural network to calculate the movement standard, personalized fitness guidance is provided, which solves the problems of high cost and time-consuming in the existing technology, and achieves convenient and scientific fitness guidance.
Patent Information
- Application Number
- CN202510432711.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-08
- Publication Date
- 2025-07-11
AI Technical Summary
Existing fitness exercise guidance requires on-site guidance from professional coaches, which is expensive and time-consuming, and cannot achieve convenient and personalized fitness guidance.
By obtaining the user's exercise video, heart rate information and experience information in real time, combining the mediapipe framework and dynamic time regularization algorithm to calculate the action standard, and using preset dual-stream input neural networks to fusion features to provide personalized fitness exercise guidance.
Convenient and personalized fitness exercise guidance has been achieved, the costs have been reduced, the scientificity and accuracy of guidance have been improved, and the intensity and content of exercise have been dynamically adjusted.
Smart Images
Figure CN120299615A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of exercise guidance, and in particular to a real-time intelligent fitness exercise guidance method, system and device. Background Art
[0002] Existing fitness exercise guidance is generally provided by professional coaches on-site, and usually occurs in professional venues such as gyms and sports fields. The benefits of professional fitness guidance by coaches are: one-on-one and personalized; the disadvantages are: high cost, need to make an appointment in advance, and long overall duration. Summary of the Invention
[0003] The purpose of this application is to provide a real-time intelligent fitness exercise guidance method, system and device, which can achieve more convenient and more comprehensive fitness exercise guidance.
[0004] To achieve the above purpose, this application provides the following solutions:
[0005] In the first aspect, this application provides a real-time intelligent fitness exercise guidance method, including:
[0006] Determine the current fitness action according to the target fitness plan; the current fitness action is used to be shown to the user for the user to exercise according to the current fitness action; the target fitness plan includes stopping exercise and multiple fitness actions;
[0007] Obtain the exercise video, heart rate information and experience information of the user during exercise;
[0008] Based on the exercise video and the current fitness action, combined with the mediapipe framework and the dynamic time warping algorithm, calculate the action standard degree;
[0009] Calculate the energy consumption based on the heart rate information;
[0010] Based on the action standard degree, the energy consumption, the experience information, the current fitness action and the target fitness plan, use a preset two-stream input neural network for feature fusion and action guidance to obtain fitness exercise guidance information; the fitness exercise guidance information is the next fitness action or stopping exercise;
[0011] When the fitness exercise guidance information is the next fitness action, update the current fitness action to the next fitness action, and then return to the step of obtaining the exercise video, heart rate information and experience information of the user during exercise.
[0012] In the second aspect, this application provides a real-time intelligent fitness exercise guidance system, including:
[0013] A fitness movement determination module for determining the current fitness movement according to a target fitness plan; the current fitness movement is used to be shown to the user for the user to exercise according to the current fitness movement; the target fitness plan includes stopping exercise and multiple fitness movements;
[0014] A data acquisition module for acquiring the exercise video, heart rate information and experience information of the user during exercise;
[0015] An action standard degree calculation module for calculating the action standard degree based on the exercise video and the current fitness movement, in combination with the mediapipe framework and the dynamic time warping algorithm;
[0016] An energy consumption calculation module for calculating the energy consumption based on the heart rate information;
[0017] A guidance analysis module for performing feature fusion and action guidance using a preset two-stream input neural network based on the action standard degree, the energy consumption, the experience information, the current fitness movement and the target fitness plan to obtain fitness exercise guidance information; the fitness exercise guidance information is the next fitness movement or stopping exercise;
[0018] A fitness exercise update module for, when the fitness exercise guidance information is the next fitness movement, updating the current fitness movement to the next fitness movement and then returning to the data acquisition module.
[0019] In a third aspect, the present application provides a real-time intelligent fitness exercise guidance device, including a visualization device, a shooting device, a heart rate sensing device and a controller;
[0020] The controller is used to: execute the steps of the real-time intelligent fitness exercise guidance method;
[0021] The visualization device, the shooting device and the heart rate sensing device are all connected to the controller;
[0022] The visualization device is used to: show the current fitness movement to the user and provide a follow-up prompt voice for the user to exercise according to the current fitness movement;
[0023] The shooting device is used to: acquire the exercise video of the user during exercise and transmit it to the controller in real time;
[0024] The heart rate sensing device is used to: acquire the heart rate information of the user during exercise and transmit it to the controller in real time.
[0025] According to the specific embodiments provided in this application, the following technical effects are achieved: This application provides a real-time intelligent fitness exercise guidance method, system, and device. Based on the exercise video and current fitness actions of the user during the exercise process, combined with the mediapipe framework and the dynamic time warping algorithm, the action standard degree is calculated, and the energy consumption is calculated based on the heart rate information. Then, by comprehensively considering these four items of data: action standard degree, energy consumption, and experience information, it comprehensively and truly reflects the user's physical condition and exercise situation during the exercise process, providing a data basis for subsequent processing. Then, combined with the current fitness action and the target fitness plan, a preset dual-stream input neural network is used for feature fusion and action guidance to obtain more accurate and more user-specific fitness exercise guidance information. When the fitness exercise guidance information is the next fitness action, the current fitness action is updated to the next fitness action, and then the step of obtaining data is returned, and thus the guidance for the next exercise is entered. Otherwise, the exercise is stopped. Thus, this application takes into account multiple situations of the user during the exercise process, quantifies them in the form of data, and then conducts more accurate and scientific fitness exercise guidance based on this quantitative data; in addition, this application does not require the follow-up and participation of a fitness coach, which is more convenient. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] In order to more clearly illustrate the technical solutions in the embodiments of this application or in the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of this application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.
[0027] Figure 1 FIG. is a schematic flowchart of a real-time intelligent fitness exercise guidance method provided by an embodiment of this application.
[0028] Figure 2 FIG. is a schematic flowchart of determining fitness exercise guidance information provided by an embodiment of this application.
[0029] Figure 3 FIG. is a schematic diagram of a real-time intelligent fitness exercise guidance device provided by an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0030] The following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the drawings in the embodiments of this application. Obviously, the described embodiments are only some, but not all, of the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of this application without creative efforts fall within the scope of protection of this application.
[0031] To make the objectives, features, and advantages of this application more apparent and understandable, the following provides a more detailed description of this application in conjunction with the accompanying drawings and specific embodiments.
[0032] In an exemplary embodiment, as Figure 1 shown, a real-time intelligent fitness exercise guidance method is provided. This method can be executed by a computer device, specifically, it can be executed independently by a computer device such as a terminal or a server, or jointly executed by a terminal and a server. In the embodiments of this application, this method includes the following steps 101 to 106.
[0033] Step 101: Determine the current fitness exercise according to the target fitness plan; the current fitness exercise is used to be shown to the user for the user to exercise according to the current fitness exercise; the target fitness plan includes stopping exercise and multiple fitness exercises.
[0034] In practical applications, the user determines the corresponding target fitness plan according to their own fitness needs, such as muscle gain, fat loss, etc. A target fitness plan includes multiple fitness exercises. For each fitness exercise, the user can follow along and imitate by combining follow-along prompt voices and follow-along prompt videos. For example, for one fitness exercise, follow along for one or two eight-beat cycles.
[0035] Step 102: Obtain the exercise video, heart rate information, and experience information of the user during the exercise process. Among them, corresponding to the example of following along for one eight-beat cycle for one fitness exercise in step 101, the exercise process of the user corresponds to an exercise duration. Therefore, an exercise video can be obtained by taking exercise images during this exercise duration. During the shooting process, in order to be accurate to the individual and exclude other influences, only the user is continuously photographed for their exercise movements. In a specific practical application, the process of obtaining the experience information of the user during the exercise process includes the following steps (21)-(23).
[0036] (21) Every preset duration, obtain the exercise feeling voice information of the user; for example, for one fitness exercise, at intervals of five or ten minutes, and this preset duration can vary according to different target fitness plans selected by the user. The process of obtaining the exercise feeling voice information of the user is as follows: Send an exercise feeling inquiry voice through a voice device, and receive and record the exercise feeling voice information replied by the user. Among them, the exercise feeling inquiry voice includes the user's fatigue level and other usage feelings.
[0037] (22) Convert the exercise feeling voice information into exercise feeling text information; specifically, voice text conversion technology can be used to convert the exercise feeling voice information into exercise feeling text information.
[0038] (23) The UPRISE (Universal Prompt Retrieval for Improving zero-Shot Evaluation) model is adopted to generate experience information according to the motion perception text information. Among them, the UPRISE model is a classification model that can understand the user's answer and retrieve the corresponding user experience to obtain the user's experience information, which is actually a numerical value.
[0039] Step 103: Based on the motion video and the current fitness action, combined with the mediapipe framework and the Dynamic Time Warping (DTW) algorithm, calculate the action standard degree.
[0040] In an application example, step 103 includes the following steps (31)-(32).
[0041] (31) For any frame of motion image in the motion video, use the mediapipe framework to calculate the spatial angles of the human body bone points in each frame of the motion image. Among them, by using the mediapipe framework to analyze the three-dimensional actions of each frame of motion image in the motion video with the human body bone key point algorithm, the spatial angles of the human body bone points can be calculated.
[0042] In a specific application, as shown in Table 1 below, the included angle types of the spatial angles of the human body bone points are: the left elbow angle is: left shoulder joint - left elbow joint - left wrist joint, the left shoulder angle is: left elbow joint - left shoulder joint - left hip joint, the left hip angle is: left shoulder joint - left hip joint - left knee joint, the left knee angle is: left hip joint - left knee joint - left ankle joint, the right elbow angle is: right shoulder joint - right elbow joint - right wrist joint, the right shoulder angle is: right elbow joint - right shoulder joint - right hip joint, the right hip angle is: right shoulder joint - right hip joint - right knee joint, the right knee angle is: right hip joint - right knee joint - right ankle joint.
[0043] Table 1
[0044]
[0045] On this basis, the left / right wrist angle and the left / right ankle angle can also be selected. Taking the calculation of the left elbow angle as an example, the left wrist coordinates (x1, y1, z1), the left elbow coordinates (x2, y2, z2), and the left shoulder coordinates (x3, y3, z3). Thus, the vector is (x1 - x2, y1 - y2, z1 - z2), and the vector is (x3 - x2, y3 - y2, z3 - z2). The corresponding calculation formula for the left elbow angle is:
[0046]
[0047] (32) The dynamic time warping algorithm is adopted to align the spatial angles of the human body bone points in each frame of the motion image with the current fitness action to calculate the action standard degree. Specifically, according to the DTW algorithm, the time series joint angles of the user's action (i.e., the spatial angles of the human body bone points in each motion image) are aligned with the time series joint angles of the action standard template (i.e., the current fitness action) (or called matching in the time series); the differences at each joint after alignment, that is, the differences in the cosine angle values of the above formula at each joint at each aligned moment, can represent the difference between the user's action and the standard action at that moment and that joint, that is, the action standard degree.
[0048] Step 104, calculate the energy consumption based on the heart rate information; wherein, the exercise duration includes multiple time windows. On this basis, step 104 includes:
[0049] (41) Input the heart rate information corresponding to one of the time windows into a preset energy consumption estimation model to obtain the window energy consumption; the preset energy consumption estimation model is obtained by training a preset deep neural network model with a heart rate energy training sample set.
[0050] Wherein, any training sample in the heart rate energy training sample set includes heart rate sample information and corresponding window energy consumption sample information as a label. The window energy consumption sample information can come from measurement values within a corresponding time period of a portable energy consumption test device based on carbon dioxide exhalation volume, estimated values of a sports watch, etc. The preset deep neural network model includes an input layer, a one-dimensional convolutional layer, a fully connected layer, a linear regression layer, and an output layer arranged in sequence.
[0051] In a practical application, the time window generally takes 10 seconds, that is, the window energy consumption for this 10 - second period is calculated every 10 seconds.
[0052] (42) Perform a summation process on the window energy consumptions corresponding to all the time windows to obtain the final energy consumption.
[0053] Step 105, based on the action standard degree, the energy consumption, the experience information, the current fitness action, and the target fitness plan, adopt a preset two - stream input neural network for feature fusion and action guidance to obtain fitness exercise guidance information; the fitness exercise guidance information is the next fitness action or stop exercising. In a specific application example, as Figure 2 shown, step 105 includes the following steps (51) - step (55).
[0054] (51) Perform adaptive pooling on the action standardization degree, the energy consumption, the experience information, and the type of the current fitness action to obtain the action standardization degree feature, the energy consumption feature, the experience information feature, and the fitness action type feature with a unified feature dimension.
[0055] In practical applications, the type of the current fitness action is a value, the action standardization degree has a dimension of several hundred, the energy consumption is a value, and the experience information has a dimension of single digits (if there is no experience information, it is padded with 0 to this dimension). First, copy the smaller of the above four values and expand it to a high dimension by itself, and then use adaptive pooling to convert it to a unified dimension.
[0056] (52) Use the collaborative filtering algorithm to perform feature matching on the action standardization degree feature, the energy consumption feature, the experience information feature, and the fitness action type feature with the fitness plan features in the target fitness plan respectively to obtain different candidate fitness plans; the candidate fitness plans are to stop exercising or any fitness action. Generally, the number of candidate fitness plans is four.
[0057] (53) Take the union of all the candidate fitness plans to obtain the final candidate plan; in addition, the above process is called "multi-channel recall" to generate the final candidate plan.
[0058] (54) Take the action standardization degree, the energy consumption, the experience information, and the type of the current fitness action as the first input data and input it into the preset two-stream input neural network, and take the final candidate plan as the second input data and input it into the preset two-stream input neural network to obtain the matching degree score of any candidate fitness plan in the final candidate plan, such as a score between 0 and 1.
[0059] In a specific application, step (54) includes the following steps (541)-(546).
[0060] (541) Perform embedding encoding on the first input data to obtain user features.
[0061] (542) Fuse the user features with the second input data to obtain the first feature.
[0062] (543) Perform bilinear interpolation processing on the first feature to obtain the second feature.
[0063] (544) Input the first feature into the SeNet module for squeeze-and-excitation processing to obtain the third feature.
[0064] (545)Perform bilinear interpolation processing on the third feature to obtain a fourth feature.
[0065] (546)Input the second feature and the fourth feature into a fully connected layer to obtain an output feature; the output feature is any candidate fitness plan and its corresponding matching degree score in the final candidate plan.
[0066] (55)Sort the matching degree scores of all the candidate fitness plans to obtain fitness exercise guidance information. In an application, the top k candidate fitness plans can be recommended according to the sorting of the matching degree scores, and k can be set as needed. For example, k can be 3, 5, 10, etc.
[0067] Step 106, when the fitness exercise guidance information is the next fitness action, update the current fitness action to the next fitness action, and then return to the step of obtaining the exercise video, heart rate information, and experience information of the user during the exercise.
[0068] In summary, this application first obtains the exercise video, heart rate information, and experience information of the user during the exercise, providing a data basis for subsequent understanding of what action the user is doing, whether the action is standard, the user's feelings, and the user's physical condition during the exercise. Then, a preset two-stream input neural network is used for feature fusion and action guidance to obtain the next fitness action or stop exercising, that is, to guide the user on which action to practice next, or that no more practice is needed. Thus, fitness exercise guidance and recommendation are realized. Through the above steps, this application can dynamically adjust the exercise intensity and content, and more comprehensive information is considered during this process, including user actions, heart rate, and user training experience. Correspondingly, the next fitness action that can be recommended will be more scientific and more in line with the user's personal physical fitness.
[0069] Based on the same inventive concept, an embodiment of this application also provides a real-time intelligent fitness exercise guidance system. The implementation solutions provided by this system to solve problems are similar to the implementation solutions described in the above method. Therefore, the specific limitations in one or more embodiments of the real-time intelligent fitness exercise guidance system provided below can refer to the limitations on the method in the above text, and will not be repeated here. In an exemplary embodiment, the real-time intelligent fitness exercise guidance system of this application includes the following multiple modules.
[0070] The fitness action determination module is used to determine the current fitness action according to the target fitness plan; the current fitness action is used to be shown to the user for the user to exercise according to the current fitness action; the target fitness plan includes stopping exercising and multiple fitness actions.
[0071] The data acquisition module is used to acquire the exercise video, heart rate information, and experience information of the user during the exercise.
[0072] An action standard calculation module, configured to calculate the action standard based on the motion video and the current fitness action, in combination with the mediapipe framework and the dynamic time warping algorithm.
[0073] An energy consumption calculation module, configured to calculate the energy consumption based on the heart rate information.
[0074] A guidance analysis module, configured to perform feature fusion and action guidance using a preset two-stream input neural network based on the action standard, the energy consumption, the experience information, the current fitness action, and the target fitness plan, so as to obtain fitness exercise guidance information; the fitness exercise guidance information is the next fitness action or stop exercising.
[0075] A fitness exercise update module, configured to, when the fitness exercise guidance information is the next fitness action, update the current fitness action to the next fitness action, and then return to the data acquisition module.
[0076] Based on the same inventive concept, an embodiment of the present application further provides a real-time intelligent fitness exercise guidance device. The implementation solutions provided by the device for solving problems are similar to the implementation solutions described in the above method. Therefore, the specific limitations in one or more embodiments of the real-time intelligent fitness exercise guidance device provided below can refer to the limitations on the method in the above text, and will not be repeated here.
[0077] In an exemplary embodiment, the real-time intelligent fitness exercise guidance device of the present application includes a visualization device, a shooting device, a heart rate sensing device, and a controller. The visualization device, the shooting device, and the heart rate sensing device are all connected to the controller.
[0078] As Figure 3 shown, the shooting device is configured to: acquire the motion video of the user during the exercise process and transmit it to the controller in real time. The heart rate sensing device is configured to: acquire the heart rate information of the user during the exercise process and transmit it to the controller in real time. The visualization device is configured to: display the current fitness action to the user (such as providing a follow-along hint video) and provide a follow-along hint voice for the user to exercise according to the current fitness action. The controller is configured to: execute the steps of the above real-time intelligent fitness exercise guidance method to implement the processing of the motion video and the processing of the heart rate information.
[0079] In addition, the device further includes a voice device, and the voice device is configured to: emit a voice for asking about the exercise feeling, receive and record the voice information of the exercise feeling replied by the user, and then process it by the controller.
[0080] In a practical application, the photographing device is a single camera. The heart rate sensing device is a sports bracelet. The controller can be a desktop computer or a notebook. The controller performs corresponding data processing operations by running system programs and various algorithms. The controller is connected to the camera via USB and to the sports bracelet via Bluetooth.
[0081] While maintaining the benefits of one-to-one, personalized, and professional, the device of the present application reduces costs and improves convenience.
[0082] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.
[0083] Specific examples are used in this article to illustrate the principle and implementation of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.
Claims
1. A real-time intelligent fitness exercise guidance method, characterized in that, The real-time intelligent fitness exercise guidance method includes: Determining a current fitness action according to a target fitness plan; the current fitness action is used to be shown to a user for the user to exercise according to the current fitness action; the target fitness plan includes stopping exercise and multiple fitness actions; Obtaining a motion video, heart rate information, and experience information of the user during exercise; Calculating an action standard degree based on the motion video and the current fitness action, in combination with the mediapipe framework and the dynamic time warping algorithm; Calculating energy consumption based on the heart rate information; Performing feature fusion and action guidance using a preset two-stream input neural network based on the action standard degree, the energy consumption, the experience information, the current fitness action, and the target fitness plan to obtain fitness exercise guidance information; the fitness exercise guidance information is the next fitness action or stopping exercise; When the fitness exercise guidance information is the next fitness action, updating the current fitness action to the next fitness action, and then returning to the step of obtaining the motion video, heart rate information, and experience information of the user during exercise.
2. The real-time intelligent fitness exercise guidance method according to claim 1, characterized in that Calculating the action standard degree based on the motion video and the current fitness action, in combination with the mediapipe framework and the dynamic time warping algorithm, includes: For any frame of motion image in the motion video, using the mediapipe framework to calculate the spatial angles of human body bone points in each frame of the motion image; Using the dynamic time warping algorithm to align the spatial angles of human body bone points in each frame of the motion image with the current fitness action to calculate the action standard degree.
3. The real-time intelligent fitness exercise guidance method according to claim 1, characterized in that The exercise process of the user corresponds to an exercise duration, and the exercise duration includes multiple time windows; Calculating energy consumption based on the heart rate information, includes: Inputting the heart rate information corresponding to one time window into a preset energy consumption estimation model to obtain window energy consumption; the preset energy consumption estimation model is obtained by training a preset deep neural network model using a heart rate energy training sample set; Performing a summation process on the window energy consumption corresponding to all the time windows to obtain the final energy consumption.
4. The real-time intelligent fitness exercise guidance method according to claim 3, characterized in that Any training sample in the heart rate energy training sample set includes heart rate sample information and corresponding window energy consumption sample information as a label; The preset deep neural network model includes an input layer, a one-dimensional convolutional layer, a fully connected layer, a linear regression layer, and an output layer arranged in sequence.
5. The real-time intelligent fitness exercise guidance method according to claim 1, wherein, The process of obtaining the experience information of the user during exercise includes: Obtaining the user's exercise feeling voice information at every preset time interval; Converting the exercise feeling voice information into exercise feeling text information; Using the UPRISE model to generate experience information according to the exercise feeling text information.
6. The real-time intelligent fitness exercise guidance method according to claim 1, wherein, Performing feature fusion and action guidance using a preset two-stream input neural network based on the action standard degree, the energy consumption, the experience information, the current fitness action, and the target fitness plan to obtain fitness exercise guidance information, includes: Perform adaptive pooling on the action standard degree, the energy consumption, the experience information, and the type of the current fitness action to obtain action standard degree features, energy consumption features, experience information features, and fitness action type features with a unified feature dimension; Adopt a collaborative filtering algorithm to perform feature matching on the action standard degree features, the energy consumption features, the experience information features, and the fitness action type features with the fitness plan features in the target fitness plan respectively to obtain different candidate fitness plans; the candidate fitness plans are to stop exercising or any fitness action; Take the union of all the candidate fitness plans to obtain the final candidate plan; Use the action standard degree, the energy consumption, the experience information, and the type of the current fitness action as the first input data and input it into a preset two-stream input neural network, and use the final candidate plan as the second input data and input it into the preset two-stream input neural network to obtain the matching degree score of any candidate fitness plan in the final candidate plan; Sort the matching degree scores of all the candidate fitness plans to obtain fitness exercise guidance information.
7. The real-time intelligent fitness exercise guidance method according to claim 6, wherein Use the action standard degree, the energy consumption, the experience information, and the type of the current fitness action as the first input data and input it into a preset two-stream input neural network, and use the final candidate plan as the second input data and input it into the preset two-stream input neural network to obtain the matching degree score of any candidate fitness plan in the final candidate plan, including: Perform embedding encoding on the first input data to obtain user features; Fuse the user features with the second input data to obtain the first feature; Perform bilinear interpolation processing on the first feature to obtain the second feature; Input the first feature into the SeNet module for squeeze-and-excitation processing to obtain the third feature; Perform bilinear interpolation processing on the third feature to obtain the fourth feature; Input the second feature and the fourth feature into a fully connected layer to obtain an output feature; the output feature is any candidate fitness plan in the final candidate plan and the corresponding matching degree score.
8. The real-time intelligent fitness exercise guidance method according to claim 2, characterized in that The included angle types of the spatial angles of the human body bone points include: the left elbow angle is: left shoulder joint - left elbow joint - left wrist joint, the left shoulder angle is: left elbow joint - left shoulder joint - left hip joint, the left hip angle is: left shoulder joint - left hip joint - left knee joint, the left knee angle is: left hip joint - left knee joint - left ankle joint, the right elbow angle is: right shoulder joint - right elbow joint - right wrist joint, the right shoulder angle is: right elbow joint - right shoulder joint - right hip joint, the right hip angle is: right shoulder joint - right hip joint - right knee joint, the right knee angle is: right hip joint - right knee joint - right ankle joint.
9. A real-time intelligent fitness exercise guidance system, characterized in that, The real-time intelligent fitness exercise guidance system includes: A fitness action determination module, configured to determine the current fitness action according to the target fitness plan; the current fitness action is used to be displayed to the user for the user to exercise according to the current fitness action; the target fitness plan includes stopping exercising and multiple fitness actions; A data acquisition module for acquiring the exercise video, heart rate information, and experience information of the user during exercise; An action standard degree calculation module for calculating the action standard degree based on the exercise video and the current fitness action, in combination with the mediapipe framework and the dynamic time warping algorithm; An energy consumption calculation module for calculating the energy consumption based on the heart rate information; A guidance analysis module for performing feature fusion and action guidance using a preset dual-stream input neural network based on the action standard degree, the energy consumption, the experience information, the current fitness action, and the target fitness plan to obtain fitness exercise guidance information; the fitness exercise guidance information is the next fitness action or stopping exercise; A fitness exercise update module for updating the current fitness action to the next fitness action when the fitness exercise guidance information is the next fitness action, and then returning to the data acquisition module.
10. A real-time intelligent fitness exercise guidance device, characterized in that, The real-time intelligent fitness exercise guidance device includes a visualization device, a shooting device, a heart rate sensing device, and a controller; The controller is configured to: execute the steps of the real-time intelligent fitness exercise guidance method according to any one of claims 1-8; The visualization device, the shooting device, and the heart rate sensing device are all connected to the controller; The visualization device is configured to: display the current fitness action to the user and provide a follow-up prompt voice for the user to exercise according to the current fitness action; The shooting device is configured to: acquire the exercise video of the user during exercise and transmit it to the controller in real time; The heart rate sensing device is configured to: acquire the heart rate information of the user during exercise and transmit it to the controller in real time.