Virtual private fitness coach service method and device based on AI large model and computer equipment

Abstract

The embodiment of the invention discloses a virtual private fitness coach service method and device based on an AI large model and computer equipment. The method comprises the steps that user portrait data are acquired, and the user portrait data are portraits formed through analysis after relevant information of the health state of a user is acquired through a question and answer mode when the user exercises for the first time; acquiring training field information and training equipment information; analyzing the training field information and the training equipment information to obtain an analysis result; inputting the analysis result and the user portrait data into a fitness expert large model to generate a corresponding fitness plan; outputting the fitness plan to the user terminal; the subjective fatigue degree and self-feeling information of the user in the training process according to the fitness plan are obtained, and feedback information is obtained; adjusting the fitness plan according to the feedback information. By implementing the method provided by the embodiment of the invention, real-time interaction with the user can be realized, and the training content suitable for the current environment and the user state can be given according to the environment of the gymnasium where the user is located and the personal feeling.

Description

Technical Field

[0001] The present invention relates to artificial intelligence, and more specifically to a virtual personal fitness coach service method, device and computer equipment based on an AI big model. Background Art

[0002] With the popularity of fitness awareness, more and more people are starting to participate in fitness activities, resulting in a rapid increase in the number of gyms. For many beginners, their first step in getting in touch with fitness is usually to find a professional and excellent fitness coach to get effective training guidance.

[0003] However, there are many problems in the current fitness coach industry, such as the lack of standardization, which makes it difficult to guarantee professional level, the scarcity of truly high-level coaches, and the high cost of private coaches. Chinese patent CN202310640387.6 discloses a video generation method, electronic device and storage medium for a virtual fitness coach, specifically, based on the action description information input by the configuration personnel, the action sequence of the virtual fitness coach is determined; each of the action description information includes one or more of text information, picture information and video information; the action sequence includes at least one first action and an action sequence; the action information corresponding to each of the first actions is obtained in the action library; in response to the action combination instruction, the fitness video information of the virtual fitness coach is generated according to the action sequence and the action information corresponding to each of the first actions. However, this method cannot interact with the user in real time, cannot provide training content suitable for the current environment and user status according to the gym environment and personal feelings of the user, and cannot effectively help fitness people.

[0004] Therefore, it is necessary to design a new method to achieve real-time interaction with users and provide training content suitable for the current environment and user status based on the user's gym environment and personal feelings. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the prior art and provide a virtual personal fitness coach service method, device and computer equipment based on an AI big model.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a virtual personal fitness coach service method based on an AI big model, comprising:

[0007] Obtaining user portrait data, wherein the user portrait data is a portrait formed by analyzing the user's health status information obtained through question-and-answer format when the user first exercises;

[0008] Obtain training venue information and training equipment information;

[0009] Analyze the training venue information and training equipment information to obtain analysis results;

[0010] Input the analysis results and the user portrait data into the fitness expert model to generate a corresponding fitness plan;

[0011] Outputting the fitness plan to a user terminal;

[0012] Obtaining the user's subjective fatigue and self-perception information during the training process according to the fitness plan to obtain feedback information;

[0013] The fitness plan is adjusted according to the feedback information.

[0014] Its further technical solution is: the fitness expert big model is formed by collecting and cleaning fitness knowledge data, combining it with an open source big language model for fine-tuning training, and integrating relevant knowledge base documents using retrieval enhancement generation technology.

[0015] The further technical solution is: the training process of the fitness expert large model includes:

[0016] Acquire fitness knowledge data, wherein the fitness knowledge includes fitness action library, exercise physiology, nutrition, and official guidance and suggestion related data;

[0017] Performing data cleaning on the fitness knowledge data to remove noise and duplicate content, so as to obtain cleaned data;

[0018] Decompose and extract the cleaned data, construct data in question-and-answer format, and filter and annotate the data in question-and-answer format in fitness scenarios to construct a dialogue dataset;

[0019] Using the conversation dataset, fine-tuning the pre-trained large model base to obtain a fine-tuned model;

[0020] Obtain relevant knowledge base documents and integrate them into the fine-tuned model to obtain a large fitness expert model.

[0021] A further technical solution is: analyzing the training venue information and the training equipment information to obtain analysis results, including:

[0022] Using scene recognition and target detection algorithms to analyze the training venue information and training equipment information, identify the size of the space, the type of equipment, and the degree of crowding, so as to obtain key information;

[0023] The key information is preprocessed and combined with preset prompt words to form an analysis result.

[0024] Its further technical solution is: the fitness plan includes exercise duration, exercise goals, plan decomposition, action names, action points, precautions, weight, number of times, number of sets, and intervals.

[0025] A further technical solution is: the feedback information includes the current breathing heart rate, muscle fatigue and subjective feedback fatigue detected by the fitness peripherals worn by the user.

[0026] A further technical solution is: adjusting the fitness plan according to the feedback information includes:

[0027] Evaluate the user's training effort and fatigue using the RPE method according to the feedback information to obtain an RPE value;

[0028] When the RPE value is lower than a set threshold, the fitness plan is dynamically adjusted.

[0029] The present invention also provides a virtual personal fitness coach service device based on an AI big model, comprising:

[0030] A data acquisition unit, used to acquire user portrait data, wherein the user portrait data is a portrait formed by analyzing the user's health status related information obtained through question-and-answer format when the user first exercises;

[0031] An information acquisition unit, used to acquire training venue information and training equipment information;

[0032] An analysis unit, used for analyzing the training venue information and the training equipment information to obtain an analysis result;

[0033] A plan generating unit, used for inputting the analysis result and the user portrait data into a fitness expert macro model to generate a corresponding fitness plan;

[0034] An output unit, used for outputting the fitness plan to a user terminal;

[0035] A feedback acquisition unit, used to acquire the user's subjective fatigue and self-perception information during the training process according to the fitness plan to obtain feedback information;

[0036] An adjustment unit is used to adjust the fitness plan according to the feedback information.

[0037] The present invention further provides a computer device, which includes a memory and a processor. The memory stores a computer program, and the processor implements the above method when executing the computer program.

[0038] The present invention also provides a storage medium, wherein the storage medium stores a computer program, and the computer program implements the above method when executed by a processor.

[0039] Compared with the prior art, the present invention has the following beneficial effects: the present invention obtains health information and analyzes the training environment through user questions and answers, generates a personalized fitness plan, and outputs the plan to the user's device. The feedback information includes the user's fatigue and feelings, which are used to dynamically adjust the fitness plan to optimize the user's training effect and experience, realize real-time interaction with the user, and provide training content suitable for the current environment and user status according to the user's gym environment and personal feelings.

[0040] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0042] Figure 1 A schematic diagram of an application scenario of a virtual personal fitness coach service method based on an AI big model provided in an embodiment of the present invention;

[0043] Figure 2 A flowchart of a virtual personal fitness coach service method based on an AI big model provided in an embodiment of the present invention;

[0044] Figure 3 A schematic diagram of a sub-process of a virtual personal fitness coach service method based on an AI big model provided in an embodiment of the present invention;

[0045] Figure 4 A schematic diagram of a sub-process of a virtual personal fitness coach service method based on an AI big model provided in an embodiment of the present invention;

[0046] Figure 5 A schematic block diagram of a virtual personal fitness coach service device based on an AI big model provided in an embodiment of the present invention; Figure 6 A schematic block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0049] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0050] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0051] See also Figure 1 and Figure 2 , Figure 1 A schematic diagram of an application scenario of a virtual personal fitness coach service method based on an AI big model is provided for an embodiment of the present invention. Figure 2 A schematic flow chart of a virtual personal fitness coach service method based on an AI big model provided in an embodiment of the present invention. The virtual personal fitness coach service method based on an AI big model is applied to a server. The server interacts with the terminal for data, and the user answers preset open questions through the terminal to provide personal information, health status, sports background, and fitness goals. This information is vectorized to form specific user portrait data, and the current fitness environment photo is uploaded through the terminal. The system performs preprocessing and constructs a fitness environment data set to provide basic data for subsequent personalized services. The server combines the user's personal portrait data and fitness environment data, and the fitness expert big model provides a personalized fitness plan. Using adaptive feedback technology, the system continuously updates and optimizes the user's training plan to ensure that the user obtains the latest and most effective fitness guidance and professional knowledge; and by analyzing the user's subjective fatigue and self-perception, the RPE (Rating of Perceived Exertion) value is comprehensively evaluated, and the fitness plan is dynamically adjusted accordingly to ensure the optimization of the training effect, to achieve real-time interaction with the user, and to provide training content suitable for the current environment and user status according to the user's gym environment and personal feelings.

[0052] Figure 2 FIG. 1 is a flow chart of a virtual personal fitness coach service method based on an AI big model provided by an embodiment of the present invention. Figure 2As shown, the method includes the following steps S110 to S170.

[0053] S110. Obtain user portrait data, wherein the user portrait data is a portrait formed by analyzing the user's health status related information obtained through question-and-answer format when the user exercises for the first time.

[0054] In this embodiment, the user first needs to answer open questions pre-programmed by the system, which are designed to assess the user's health status and personal background information. These questions cover the following four aspects:

[0055] Presence of diagnosed diseases: Users are asked to indicate whether they have any current medically diagnosed diseases or health problems;

[0056] Whether there are potential risk factors for certain diseases: Users need to indicate whether there are factors that may increase the risk of certain diseases, such as family history or other potential risk factors;

[0057] Presence of certain disease symptoms: Users are asked to describe whether they have any potential health symptoms or signs;

[0058] Whether there are any lifestyle or behavioral habits that are harmful to health: Users are required to report whether they have any lifestyle or behavioral habits, such as smoking, drinking, unhealthy diet, etc., which may have a negative impact on health.

[0059] In addition, basic personal information is required, such as gender, age, height, weight, etc. In addition, users can also describe their fitness goals, such as muscle gain, fat loss, and physical fitness.

[0060] Based on the information provided by the user during the question-and-answer process, the server vectorizes this data and forms a personal portrait of the specific user. These vectorized personal data are stored in the database, providing the basis for the personalization of subsequent fitness plans and recommendations.

[0061] This Q&A warm-up session not only helps users understand their health status and potential risks, but also ensures that subsequent fitness services can be more personalized and effective.

[0062] Specifically, the collected data may come from the user's text description or option selection, and needs to be preprocessed and normalized to ensure the consistency and comparability of the data. For example, height and weight are converted into standard units (such as meters and kilograms), and medical history information is classified and coded.

[0063] For each user, key features are extracted from the various data collected. These features may include whether there is a specific disease, an assessment of lifestyle habits (such as smoking, drinking frequency), physical indicators (such as BMI index), fitness goals, etc. Each feature may need to be converted into digital form or binary (for example, yes / no). The extracted features are combined into a vector to represent the user's personal portrait. This vector can be an array of numbers, with each position representing a specific feature. For example, a vector containing disease information, fitness goals, lifestyle habits, etc. The formed user data is usually stored in a vector database so that the system can quickly access and use it when needed. This database management usually involves secure storage of data and permission management to ensure user privacy and data protection. The formed personal portrait can be used to generate personalized fitness plans, nutritional advice, or health risk assessments. The system can recommend suitable exercise programs, diet plans, or lifestyle recommendations based on the user's personal portrait to better meet the user's health needs and goals.

[0064] Only when using the virtual personal fitness trainer service for the first time or when there is a change in health status, is it necessary to re-execute the question-and-answer method to obtain information related to the user's health status and analyze the resulting portrait.

[0065] S120: Obtain training venue information and training equipment information.

[0066] In this embodiment, the training venue information refers to the photos and / or textual description of the training venue, and the textual description may include the available time period, etc. The training equipment information refers to the photos and / or textual description of the training equipment.

[0067] S130: Analyze the training venue information and training equipment information to obtain analysis results.

[0068] In this embodiment, the analysis results include feature vectors such as space size, crowding level, and available equipment types.

[0069] In one embodiment, see Figure 3 , the above-mentioned step S130 may include steps S131 to S132.

[0070] S131. Use scene recognition and target detection algorithms to analyze the training venue information and training equipment information, identify the size of the space, the type of equipment and the degree of crowding, so as to obtain key information.

[0071] In this embodiment, the key information includes information about the size of the space, the type of equipment, and the degree of crowding.

[0072] Specifically, first, the layout and structure of the training venue are analyzed using scene recognition algorithms. This involves image processing or sensor data analysis to identify the shape, size, and layout features of the room or venue; then, the placement of the training equipment is analyzed using object detection algorithms. Object detection algorithms can identify and locate the type and location of training equipment, thereby determining the type and distribution of equipment.

[0073] S132: Preprocess the key information and form an analysis result in combination with preset prompt words.

[0074] In this embodiment, the key information obtained is preprocessed. This includes data cleaning, denoising or further data processing to ensure the accuracy and consistency of the information; according to the preset prompt words or rules, the system converts the preprocessed information into a specific feature vector or descriptive information. For example, the size of the space can be expressed in square meters, the type of equipment can be converted into a categorical variable through coding, and the degree of crowding can be expressed in numerical values, such as density or utilization rate.

[0075] In this embodiment, by analyzing the photos of fitness venues taken by users, whether they are gyms or other places that meet exercise needs, the current exercise environment and available fitness equipment or tools (such as kettlebells, balance balls, pullers, sandbags, etc.) are quickly identified. After processing by scene recognition and target detection algorithms, the system will pre-process the data and build a detailed fitness environment data set, including space size, crowding level, and available equipment types.

[0076] Based on these data, the system uses built-in prompts to input the fitness expert model and provide alternative personalized exercise plans. These plans can flexibly meet the needs of different users, taking into account changes in venues and equipment, ensuring that users can achieve the expected results when implementing exercise plans and avoiding environmental restrictions that affect exercise quality and results.

[0077] This technology not only improves the level of intelligence in fitness venue management, but also provides users with a more personalized and effective exercise experience, fully utilizes existing resources, and optimizes the implementation process of exercise plans.

[0078] S140, inputting the analysis results and the user portrait data into a fitness expert model to generate a corresponding fitness plan.

[0079] In this embodiment, the fitness plan includes exercise duration, exercise goals, plan breakdown, action names, action points, precautions, weight, number of times, number of sets, and intervals.

[0080] In one embodiment, the fitness expert big model is a model formed by collecting and cleaning fitness knowledge data, fine-tuning and training with an open source big language model, and integrating relevant knowledge base documents using retrieval enhancement generation technology.

[0081] Specifically, the training process of the fitness expert model includes:

[0082] Acquire fitness knowledge data, wherein the fitness knowledge includes fitness action library, exercise physiology, nutrition, official guidance and other related data;

[0083] The fitness knowledge data is cleaned to remove noise and duplicate content to obtain cleaned data.

[0084] In this embodiment, fitness knowledge data is collected and deduplication and noise processing is performed, and crawlers are used to obtain more data or data enhancement technology is used to increase data diversity.

[0085] The cleaned data is decomposed and extracted to construct data in question-and-answer format. The data in question-and-answer format in fitness scenarios is filtered and labeled to construct a dialogue dataset.

[0086] In this embodiment, the preprocessed data is decomposed and annotated to construct training data in the question-answer format. These data are mainly used to train the generative model to answer the user's fitness-related questions in the conversation.

[0087] The pre-trained large model base is fine-tuned using the dialogue dataset to obtain a fine-tuned model.

[0088] In this embodiment, an open source large model base is used for fine-tuning, and a large amount of Q&A training data is input. Through supervised fine-tuning technology (SFT), it is ensured that the model reaches a certain accuracy on the validation set, thus becoming a fine-tuned model.

[0089] Obtain relevant knowledge base documents and integrate them into the fine-tuned model to obtain a large fitness expert model.

[0090] In this embodiment, the refined knowledge base documents are mounted on the model through the retrieval enhancement generation (RAG) technology. In this way, the model can prioritize the retrieval and use of these refined knowledge when interacting with the user, and answer in combination with its own memory; configure the model role, such as initializing Prompt as a fitness coach, specifically, "You are now a high-level fitness coach with professional and rich knowledge of fitness sports. You need to provide professional fitness plans and training guidance for your users, and encourage and praise users with positive and sunny language to arrange fitness plans reasonably and urge them to execute them." The model needs to provide users with fitness plans and guidance in professional and positive language, and encourage and praise them to promote users to execute fitness plans; adjust and optimize the user's fitness plan based on user feedback and historical interaction data. For example, adjust the weight or number of groups according to the user's physical condition and feedback to ensure the personalization and effectiveness of the fitness plan.

[0091] The Fitness Expert Big Model replaces the professionalism of coaches from data analysis to plan output. The Fitness Expert Big Model also provides fitness enthusiasts with high-quality emotional value, that is, positive feedback, including encouragement and praise, etc. Like a conversation between friends, the Fitness Expert Big Model can provide motivational feedback to fitness enthusiasts after training.

[0092] This approach combines advanced natural language processing technology with expertise in the fitness field to provide users with highly personalized and professional fitness guidance and recommendations.

[0093] In this embodiment, the fitness expert big model includes: Qwen2-7B, Llama-3.1-8B, Internlm2-7B. Qwen2 and Internlm2 both have quite good Chinese comprehension capabilities, and Llama-3.1, as the latest open source big model, has outstanding English conversation capabilities. The required vector databases can be selected from Qdrant and Milvus, both of which can store high-dimensional vector data of different user conversation history records or refined knowledge in the RAG scheme, and have excellent vector retrieval capabilities.

[0094] S150, outputting the fitness plan to a user terminal;

[0095] S160: Obtain the user's subjective fatigue and self-perception information during training according to the fitness plan to obtain feedback information.

[0096] In this embodiment, the feedback information includes the current breathing heart rate, muscle fatigue and subjective feedback fatigue detected by the fitness peripherals worn by the user.

[0097] S170: Adjust the fitness plan according to the feedback information.

[0098] In one embodiment, see Figure 4 , the above-mentioned step S170 may include steps S171 to S172.

[0099] S171, using RPE (Rating of Perceived Exertion) method to evaluate the user's training effort and fatigue according to the feedback information to obtain an RPE value;

[0100] S172: When the RPE value is lower than a set threshold, dynamically adjust the fitness plan.

[0101] In this embodiment, the adjustment may include modifying the weight, increasing or decreasing the number of sets, adjusting the interval time, etc., to better meet the user's physical condition and personal feelings.

[0102] Users start exercising according to the generated exercise plan. During the exercise, the interactive module interacts with users in real time through algorithms and natural language processing technology, understands the context and emotional color, analyzes the user's current emotional state, and makes personalized encouragement and praise responses based on the subjective fatigue and self-perception described by the user.

[0103] The RPE value is assessed for the user based on the subjective feelings described by the user. When the RPE value exceeds or falls below the preset threshold range, the current training action or overall exercise plan is dynamically adjusted. For example, suppose a user is doing a barbell squat with a single-sided weight of 10 kg, and describes his feelings as "My heart rate is off the charts, I'm almost exhausted!" After comprehensive analysis, the fitness expert model determines that the user's current RPE value is 17, while the set threshold range is 12 to 15. Based on this judgment, it may be recommended to reduce the weight to 8 kg and increase the rest time between groups to 30 seconds to better suit the user's current physical condition and feelings.

[0104] Specifically, when users exercise according to their exercise plans, their training efforts and fatigue are evaluated by describing their feelings during the communication process. The RPE method is used to quantify the user's training intensity. The large model combines the data obtained from the fitness peripherals worn by the user (such as sports watches or bracelets), including the current respiratory heart rate, muscle fatigue and overall fatigue, and the user's health information stored in the vector database, and comprehensively considers the user's physiological and psychological state to identify the current training status.

[0105] When the RPE value is below or above a specific threshold range, the big model will dynamically adjust the user's exercise plan, increasing or reducing the exercise load to ensure the effectiveness and safety of training.

[0106] The server also built an interactive module that uses natural language processing technology to understand the context and emotional color through the subjective fatigue and emotional state described by the user. The large model uses its contextual reasoning ability to enter the corresponding situation according to the user's state, and provides encouragement and praise in the form of dialogue, providing high-quality emotional support.

[0107] After each interaction with the user, the conversation data is vectorized and stored in a vector database. Through continuous interaction, the model can gradually deepen its understanding of the user's physiological and psychological state, thereby improving the implementation effect of the fitness plan.

[0108] For example, for a newly joined user A, obtaining user information through question-and-answer format includes but is not limited to:

[0109] Personal information: Male, 30 years old, working in IT, with high work pressure, can exercise three times a week;

[0110] Lifestyle or habits: staying up late, sitting for long periods of time, smoking 10 to 19 cigarettes a day;

[0111] Sports foundation: Have running foundation, no resistance training foundation, and have not participated in any sports in the past two years;

[0112] Personal medical history: moderate fatty liver, no hereditary diseases in the family;

[0113] Coronary heart disease (CHD) risk factor assessment: 2 risk factors, medium-risk group;

[0114] Basic indicators: height, weight, body circumference, composition, resting heart rate, blood pressure, etc.;

[0115] Physical fitness: cardiopulmonary endurance - number of step tests, muscular endurance - number of sit-ups and push-ups;

[0116] Fitness demands: adjustment of overwork obesity, shaping, and body shape improvement;

[0117] A's common sports venue is the fitness room provided by his company. He takes photos of the sports venue and the training equipment in the venue and uploads them;

[0118] The expert model provides a fitness plan three times a week (including movement explanations and demonstrations) and corresponding dietary recommendations, and reminds A to complete the corresponding plan in time on the day of exercise;

[0119] When A is completing a certain fitness plan, there are more people in the gym than usual, and the equipment used in the training plan is occupied. A takes a real-time photo of the venue and has a text conversation: "There are so many people working out today! This equipment is occupied" (the plan includes the "equipment seated press" action, and someone is using the seated shoulder press in the photo). At this time, the expert model may give the following answer: "It doesn't matter! The free weight area is relatively empty. Let's change the equipment seated press in today's plan to dumbbell seated shoulder press. According to your current foundation, you can choose 10 kg on one side. Let me demonstrate it first", and play the dumbbell seated shoulder press demonstration video.

[0120] After completing the second set, A felt a little sore in his shoulders. At this time, the expert model combined the previous training foundation and intensity, and comprehensively determined that the RPE value did not exceed the threshold range, and might give the following answer: "I can totally understand how you feel! Two sets of 10 kg shoulder presses are great, and you are really working hard. Imagine that your deltoid muscles are getting stronger. Rest for a minute, take a few deep breaths, and then tell yourself that you can complete the remaining two sets. Believe me, you will feel super accomplished when you do it! Keep it up, I believe in you!"

[0121] After three months of regular exercise, A's health and fitness level has improved, his physical quality has been significantly improved, and he has initially mastered basic sports knowledge. The above-mentioned virtual personal fitness coach service method based on the AI ​​large model obtains health information and analyzes the training environment through user questions and answers, generates a personalized fitness plan, and outputs the plan to the user's device. Feedback information includes the user's fatigue and feelings, which is used to dynamically adjust the fitness plan to optimize the user's training effect and experience, realize real-time interaction with the user, and provide training content suitable for the current environment and user status based on the user's gym environment and personal feelings.

[0122] Figure 5 is a schematic block diagram of a virtual personal fitness coach service device 300 based on an AI big model provided by an embodiment of the present invention. Figure 5 As shown, corresponding to the above virtual personal fitness coach service method based on AI big model, the present invention also provides a virtual personal fitness coach service device 300 based on AI big model. The virtual personal fitness coach service device 300 based on AI big model includes a unit for executing the above virtual personal fitness coach service method based on AI big model, and the device can be configured in a server. Specifically, please refer to Figure 5 The virtual personal fitness trainer service device 300 based on the AI ​​big model includes a data acquisition unit 301, an information acquisition unit 302, an analysis unit 303, a plan generation unit 304, an output unit 305, a feedback acquisition unit 306 and an adjustment unit 307.

[0123] The data acquisition unit 301 is used to acquire user portrait data, wherein the user portrait data is a portrait formed by analyzing the user's health status related information obtained through question-and-answer format when the user exercises for the first time; the information acquisition unit 302 is used to acquire training venue information and training equipment information; the analysis unit 303 is used to analyze the training venue information and training equipment information to obtain analysis results; the plan generation unit 304 is used to input the analysis results and the user portrait data into a fitness expert model to generate a corresponding fitness plan; the output unit 305 is used to output the fitness plan to the user terminal; the feedback acquisition unit 306 is used to acquire the user's subjective fatigue and self-perception information during the training according to the fitness plan to obtain feedback information; the adjustment unit 307 is used to adjust the fitness plan according to the feedback information.

[0124] Among them, the fitness expert big model is formed by collecting and cleaning fitness knowledge data, combining it with an open source big language model for fine-tuning training, and integrating relevant knowledge base documents using retrieval enhancement generation technology.

[0125] The training process of the fitness expert model includes:

[0126] Obtain fitness knowledge data, wherein the fitness knowledge includes fitness movement library, exercise physiology, nutrition, official guidance and other related data; perform data cleaning on the fitness knowledge data to remove noise and duplicate content to obtain cleaned data; decompose and extract the cleaned data to construct data in question and answer format, and filter and annotate the data in question and answer format in fitness scenarios to construct a dialogue data set; use the dialogue data set to fine-tune the pre-trained large model base to obtain a fine-tuned model; obtain relevant knowledge base documents and integrate them into the fine-tuned model to obtain a fitness expert large model.

[0127] In one embodiment, the analysis unit 303 includes an identification subunit and a preprocessing subunit.

[0128] The recognition subunit is used to analyze the training venue information and training equipment information using scene recognition and target detection algorithms, identify the space size, equipment type and crowding level to obtain key information; the preprocessing subunit is used to preprocess the key information and form an analysis result in combination with preset prompt words.

[0129] In one embodiment, the adjustment unit 307 includes an RPE value calculation subunit and a dynamic adjustment subunit.

[0130] The RPE value calculation subunit is used to evaluate the user's training effort and fatigue using the RPE method according to the feedback information to obtain the RPE value; the dynamic adjustment subunit is used to dynamically adjust the fitness plan when the RPE value is lower than a set threshold.

[0131] It should be noted that technical personnel in the relevant field can clearly understand that the specific implementation process of the above-mentioned virtual personal fitness trainer service device 300 based on the AI ​​big model and each unit can refer to the corresponding description in the aforementioned method embodiment, and for the convenience and conciseness of the description, it will not be repeated here.

[0132] The above-mentioned virtual personal fitness coach service device 300 based on the AI ​​big model can be implemented in the form of a computer program, which can be used in Figure 6 Runs on the computer device shown.

[0133] See also Figure 6 , Figure 6 5 is a schematic block diagram of a computer device provided in an embodiment of the present application. The computer device 500 may be a server, wherein the server may be an independent server or a server cluster composed of multiple servers.

[0134] See also Figure 6 The computer device 500 includes a processor 502 , a memory and a network interface 505 connected via a system bus 501 , wherein the memory may include a non-volatile storage medium 503 and an internal memory 504 .

[0135] The non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions, which, when executed, can enable the processor 502 to execute a virtual personal fitness coach service method based on an AI big model.

[0136] The processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500 .

[0137] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a virtual personal fitness coach service method based on an AI big model.

[0138] The network interface 505 is used to communicate with other devices over the network. Figure 6The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present application, and does not constitute a limitation on the computer device 500 to which the solution of the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0139] The processor 502 is used to run the computer program 5032 stored in the memory to implement the following steps:

[0140] Obtain user portrait data, wherein the user portrait data is a portrait formed by analyzing the user's health status related information obtained through question-and-answer format when the user exercises for the first time; obtain training venue information and training equipment information; analyze the training venue information and training equipment information to obtain analysis results; input the analysis results and the user portrait data into a fitness expert model to generate a corresponding fitness plan; output the fitness plan to a user terminal; obtain the user's subjective fatigue and self-perception information during training according to the fitness plan to obtain feedback information; and adjust the fitness plan according to the feedback information.

[0141] Among them, the fitness expert big model is formed by collecting and cleaning fitness knowledge data, combining it with an open source big language model for fine-tuning training, and integrating relevant knowledge base documents using retrieval enhancement generation technology.

[0142] The training process of the fitness expert model includes:

[0143] Obtain fitness knowledge data, wherein the fitness knowledge includes fitness movement library, exercise physiology, nutrition, official guidance and other related data; perform data cleaning on the fitness knowledge data to remove noise and duplicate content to obtain cleaned data; decompose and extract the cleaned data to construct data in question and answer format, and filter and annotate the data in question and answer format in fitness scenarios to construct a dialogue data set; use the dialogue data set to fine-tune the pre-trained large model base to obtain a fine-tuned model; obtain relevant knowledge base documents and integrate them into the fine-tuned model to obtain a fitness expert large model.

[0144] The fitness plan includes exercise duration, exercise goals, plan breakdown, action names, action points, precautions, weight, number of times, number of sets, and intervals.

[0145] The feedback information includes the current breathing heart rate, muscle fatigue and subjective feedback fatigue detected by the fitness peripherals worn by the user.

[0146] In one embodiment, when the processor 502 implements the step of analyzing the training venue information and the training equipment information to obtain the analysis result, the processor 502 specifically implements the following steps:

[0147] The training venue information and training equipment information are analyzed using scene recognition and target detection algorithms to identify the size of the space, the type of equipment and the degree of crowding to obtain key information; the key information is preprocessed and combined with preset prompt words to form an analysis result.

[0148] In one embodiment, when the processor 502 implements the step of adjusting the fitness plan according to the feedback information, the processor 502 specifically implements the following steps:

[0149] The RPE method is used to evaluate the user's training effort and fatigue according to the feedback information to obtain an RPE value; when the RPE value is lower than a set threshold, the fitness plan is dynamically adjusted.

[0150] It should be understood that in the embodiment of the present application, the processor 502 may be a central processing unit (CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0151] It can be understood by those skilled in the art that all or part of the processes in the method for implementing the above embodiment can be completed by instructing the relevant hardware through a computer program. The computer program includes program instructions, and the computer program can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiment of the above method.

[0152] Therefore, the present invention also provides a storage medium. The storage medium may be a computer-readable storage medium. The storage medium stores a computer program, wherein when the computer program is executed by a processor, the processor executes the following steps:

[0153] Obtain user portrait data, wherein the user portrait data is a portrait formed by analyzing the user's health status related information obtained through question-and-answer format when the user exercises for the first time; obtain training venue information and training equipment information; analyze the training venue information and training equipment information to obtain analysis results; input the analysis results and the user portrait data into a fitness expert model to generate a corresponding fitness plan; output the fitness plan to a user terminal; obtain the user's subjective fatigue and self-perception information during training according to the fitness plan to obtain feedback information; and adjust the fitness plan according to the feedback information.

[0154] Among them, the fitness expert big model is formed by collecting and cleaning fitness knowledge data, combining it with an open source big language model for fine-tuning training, and integrating relevant knowledge base documents using retrieval enhancement generation technology.

[0155] The training process of the fitness expert model includes:

[0156] Obtain fitness knowledge data, wherein the fitness knowledge includes fitness movement library, exercise physiology, nutrition, official guidance and other related data; perform data cleaning on the fitness knowledge data to remove noise and duplicate content to obtain cleaned data; decompose and extract the cleaned data to construct data in question and answer format, and filter and annotate the data in question and answer format in fitness scenarios to construct a dialogue data set; use the dialogue data set to fine-tune the pre-trained large model base to obtain a fine-tuned model; obtain relevant knowledge base documents and integrate them into the fine-tuned model to obtain a fitness expert large model.

[0157] The fitness plan includes exercise duration, exercise goals, plan breakdown, action names, action points, precautions, weight, number of times, number of sets, and intervals.

[0158] The feedback information includes the current breathing heart rate, muscle fatigue and subjective feedback fatigue detected by the fitness peripherals worn by the user.

[0159] In one embodiment, when the processor executes the computer program to implement the step of analyzing the training venue information and the training equipment information to obtain the analysis result, the processor specifically implements the following steps:

[0160] The training venue information and training equipment information are analyzed using scene recognition and target detection algorithms to identify the size of the space, the type of equipment and the degree of crowding to obtain key information; the key information is preprocessed and combined with preset prompt words to form an analysis result.

[0161] In one embodiment, when the processor executes the computer program to implement the step of adjusting the fitness plan according to the feedback information, the processor specifically implements the following steps:

[0162] The RPE method is used to evaluate the user's training effort and fatigue according to the feedback information to obtain an RPE value; when the RPE value is lower than a set threshold, the fitness plan is dynamically adjusted.

[0163] The storage medium may be a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk, or an optical disk, etc., which are computer-readable storage media that can store program codes.

[0164] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0165] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of each unit is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0166] The steps in the method of the embodiment of the present invention can be adjusted in order, combined and deleted according to actual needs. The units in the device of the embodiment of the present invention can be combined, divided and deleted according to actual needs. In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0167] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, terminal, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention.

[0168] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A virtual personal fitness coach service method based on an AI big model, characterized in that: include: Obtaining user portrait data, wherein the user portrait data is a portrait formed by analyzing the user's health status information obtained through question-and-answer format when the user first exercises; Obtain training venue information and training equipment information; Analyze the training venue information and training equipment information to obtain analysis results; Input the analysis results and the user portrait data into the fitness expert model to generate a corresponding fitness plan; Outputting the fitness plan to a user terminal; Obtaining the user's subjective fatigue and self-perception information during the training process according to the fitness plan to obtain feedback information; The fitness plan is adjusted according to the feedback information.

2. The virtual personal fitness coach service method based on AI big model according to claim 1 is characterized in that: The fitness expert big model is formed by collecting and cleaning fitness knowledge data, combining it with an open source big language model for fine-tuning training, and integrating relevant knowledge base documents using retrieval enhancement generation technology.

3. The virtual personal fitness coach service method based on AI big model according to claim 2 is characterized in that: The training process of the fitness expert model includes: Acquire fitness knowledge data, wherein the fitness knowledge data includes movement library, exercise physiology, nutrition, and official guidance and suggestion related data; Performing data cleaning on the fitness knowledge data to remove noise and duplicate content, so as to obtain cleaned data; Decompose and extract the cleaned data, construct data in question-and-answer format, and filter and annotate the data in question-and-answer format in fitness scenarios to construct a dialogue dataset; Using the conversation dataset, fine-tuning the pre-trained large model base to obtain a fine-tuned model; Obtain relevant knowledge base documents and integrate them into the fine-tuned model to obtain a large fitness expert model.

4. The virtual personal fitness coach service method based on AI big model according to claim 1 is characterized in that: The analyzing the training venue information and the training equipment information to obtain analysis results includes: Using scene recognition and target detection algorithms to analyze the training venue information and training equipment information, identify the size of the space, the type of equipment, and the degree of crowding, so as to obtain key information; The key information is preprocessed and combined with preset prompt words to form an analysis result.

5. The virtual personal fitness coach service method based on AI big model according to claim 1 is characterized in that: The fitness plan includes exercise duration, exercise goals, plan breakdown, action names, action key points, precautions, weight, number of times, number of sets, and intervals.

6. The virtual personal fitness coach service method based on AI big model according to claim 1 is characterized in that: The feedback information includes the current breathing heart rate, muscle fatigue and subjective feedback fatigue detected by the fitness peripherals worn by the user.

7. The virtual personal fitness coach service method based on AI big model according to claim 6 is characterized in that: The step of adjusting the fitness plan according to the feedback information includes: Evaluate the user's training effort and fatigue using the RPE method according to the feedback information to obtain an RPE value; When the RPE value is lower than a set threshold, the fitness plan is dynamically adjusted.

8. A virtual personal fitness coach service device based on an AI big model, characterized in that: include: A data acquisition unit, used to acquire user portrait data, wherein the user portrait data is a portrait formed by analyzing the user's health status related information obtained through question-and-answer format when the user first exercises; An information acquisition unit, used to acquire training venue information and training equipment information; An analysis unit, used for analyzing the training venue information and the training equipment information to obtain an analysis result; A plan generating unit, used for inputting the analysis result and the user portrait data into a fitness expert macro model to generate a corresponding fitness plan; An output unit, used for outputting the fitness plan to a user terminal; A feedback acquisition unit, used to acquire the user's subjective fatigue and self-perception information during the training process according to the fitness plan to obtain feedback information; An adjustment unit is used to adjust the fitness plan according to the feedback information.

9. A computer device, characterized in that: The computer device comprises a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images