Intelligent exercise guidance system

Abstract

The invention belongs to the field of human motion monitoring, and particularly relates to an intelligent motion guidance system. The invention provides an intelligent exercise guidance system. The intelligent exercise guidance system comprises a sensor unit, a data transmission unit, a data processing unit, an exercise guidance unit and a data storage unit. The data transmission unit transmits motion data collected by at least two sensors to the data processing unit in real time. The data processing unit is used for processing the motion data acquired by the at least two sensors and transmitting the processed data to the motion guidance unit through the data transmission unit. And the motion guidance unit is used for providing real-time motion guidance and generating a corresponding training plan. By adopting the intelligent exercise guidance system, the drift error of the sensor can be automatically corrected, progressive guidance from basic action specifications to special skill optimization is supported, and professional requirements of different training stages are met.

Description

Technical Field

[0001] The present application relates to the field of human motion monitoring, and in particular to an intelligent motion guidance system. Background Art

[0002] Currently, there are two main technical solutions in the field of motion monitoring: professional optical motion capture systems and inertial sensor systems. Professional optical motion capture systems use multiple high-speed cameras (usually ≥6) arranged in a fixed space and reflective markers to reconstruct three-dimensional motion trajectories. However, their disadvantage is their high price. Inertial sensor systems, based on microelectromechanical systems (MEMS) inertial measurement units (IMUs), collect motion data through wearable devices. However, their disadvantage is that inertial sensor systems can accumulate errors due to drift.

[0003] The above two technical solutions are difficult to synchronously obtain detection data such as joint kinematics, dynamics and electromyographic signals. Therefore, they generally remain at the primary functional stage such as action counting and simple posture alarms, and cannot provide real-time and personalized guidance suggestions based on the long-term and short-term training data of the target object (i.e., the user). Summary of the Invention

[0004] One of the purposes of this application is to provide an intelligent sports guidance system that can provide real-time and personalized guidance suggestions.

[0005] To achieve the above-mentioned and other related purposes, the present application provides an intelligent sports guidance system, which includes a sensor unit, a data transmission unit, a data processing unit, a sports guidance unit, and a data storage unit;

[0006] The sensor unit is used to collect the user's motion data; the sensor unit includes at least two types of sensors;

[0007] The data transmission unit transmits the motion data collected by at least two sensors to the data processing unit in real time;

[0008] The data storage unit is used to store standard action data;

[0009] The data processing unit processes the motion data collected by the at least two sensors, and transmits the processed data to the motion guidance unit through the data transmission unit;

[0010] The data processing includes fusing motion data collected by at least two sensors to obtain fused data; the data processing also includes at least one of the following steps: comparing the fused data with standard motion data to determine the deviation between the user's motion and the standard motion; predicting possible dangerous motions based on the fused data;

[0011] The motion guidance unit includes an action guidance module and a decision analysis module. The action guidance module is used to provide real-time action guidance based on the action deviation between the user's action and the standard action and the possible dangerous action; the decision analysis module is used to generate a corresponding training plan based on the action deviation between the user's action and the standard action and the possible dangerous action.

[0012] This application has at least the following beneficial effects:

[0013] The intelligent motion guidance system of the present application has low training costs and a wide range of applicable scenarios; with multimodal data fusion technology, the system can achieve millisecond-level motion guidance. Compared with single data processing technology, this system is more accurate and reliable; the liquid metal flexible sensor can accurately capture the changes in electrical signals generated by muscle micro-strain, realize millimeter-level motion recognition, and the liquid metal flexible sensor has small drift and cumulative errors; it can achieve millimeter-level joint displacement monitoring accuracy and three-dimensional spatial posture reconstruction capabilities, and can generate ergonomic motion correction suggestions within millisecond-level delays; the liquid metal flexible sensor can automatically correct the drift error of the inertial measurement unit; it supports progressive guidance from basic motion specifications to special skill optimization to meet the professional needs of different training stages; it can recognize complex motion patterns and generate long-term training plans, taking into account both real-time and analysis depth; it can flexibly adjust the topology of the sensor communication network according to the characteristics of the sports project. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 Schematic diagram of the structure of an intelligent sports guidance system according to an embodiment of the present application. DETAILED DESCRIPTION

[0016] The following describes the embodiments of the present application through specific embodiments. Those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments. The details in the present application can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.

[0017] This embodiment provides an intelligent exercise guidance system that collects user exercise data through sensors, processes and analyzes the received data, and generates a personalized exercise guidance plan.

[0018] The intelligent motion guidance system includes a sensor unit, a data transmission unit, a data processing unit, a motion guidance unit, and a data storage unit.

[0019] The sensor unit is used to collect the user's motion data. The data transmission unit transmits the motion data collected by the sensor to the data processing unit in real time. The data processing unit processes the motion data collected by the sensor unit and transmits the processed data to the motion guidance unit via the data transmission unit. The motion guidance unit performs decision analysis based on the processed data and provides real-time movement guidance, generates training plans, or outputs reports based on the decision analysis. The data storage unit is used to store the user's historical movement data and standard movement data, facilitating data processing by the data processing unit and convenient access by the motion guidance unit for decision analysis.

[0020] The sensor unit can be built into sportswear to collect sports data, such as knee pads, wrist guards, tights, and leggings.

[0021] Movement data includes joint movement data and muscle activity data. The joint movement data is used to record various parameters of the joints during movement, while the muscle activity data is used to record the activity of the muscles during movement.

[0022] Joint motion data includes joint position and joint angle. Joint angle includes joint first angle and joint second angle.

[0023] The sensor unit includes a flexible sensor that can capture the first angle of the joint. Flexible sensors have physical properties such as softness, bendability, and stretchability. Multiple flexible sensors can be built into sportswear, using a serpentine routing layout to form a strain sensing array and arranged in the joint activity area. When the human joint moves, the muscles near the joint will produce micro-strain, and the flexible sensor can generate a changing electrical signal based on the micro-strain to infer the first angle of the joint. The forward kinematics algorithm can be used to calculate the position of the human body end and the user's position and displacement based on the human joint angle. The position of the human body end can be the position of the toes. For example, the hip joint angle and knee joint angle are obtained using flexible sensors, and the user's position and displacement are calculated using the forward kinematics algorithm.

[0024] In some embodiments, the flexible sensor used in the present application can be made of strain sensing material, and the strain sensing material can be liquid metal such as gallium-based liquid metal. The flexible sensor made of liquid metal is a liquid metal flexible sensor, such as a gallium-based liquid metal flexible sensor; compared with sensors made of other materials, the liquid metal flexible sensor has excellent conductive properties, can accurately capture the changes in electrical signals caused by muscle micro-strain, realize millimeter-level motion recognition, and the sensor drift and cumulative error are small, and can more sensitively collect the electrical signals caused by micro-strain changes, thereby obtaining a more accurate first angle of the joint, the position of the end of the human body, and the position and displacement of the user.

[0025] The sensor unit includes an inertial measurement unit (IMU) capable of collecting joint positions and joint second angles. The IMU includes an accelerometer and a gyroscope, wherein the accelerometer is used to collect joint positions, and the gyroscope is used to collect joint second angles.

[0026] The sampling frequency of the inertial measurement unit can be greater than or equal to 60Hz to meet the needs of accurately capturing rapid human movements. The inertial measurement unit can be built into sportswear and placed in the joint movement area.

[0027] The sensor unit includes a millimeter-wave radar, which can detect joint position and angle. It transmits high-frequency electromagnetic waves and receives reflected signals. When a user bends a limb, the surfaces on either side of the joint produce relative displacement. This relative displacement can be used to calculate the joint angle. The millimeter-wave radar can be placed on sportswear.

[0028] Muscle activity data includes muscle electrical signals (EMG) and muscle strength data.

[0029] Muscle electrical signals are electrical activity signals generated by muscles during contraction and relaxation. By collecting electrical activity signals, the degree of muscle activation and activity pattern can be understood. Muscle electrical signals can be collected using flexible sensors or surface electromyography sensors. In some embodiments, the flexible sensor is a liquid metal flexible sensor. Flexible sensors and surface electromyography (sEMG) sensors can be built into sportswear and arranged in muscle activity areas. Muscle activation can be calculated from muscle electrical signals, and the three-dimensional motion trajectory of the joint can be reconstructed using the joint angle and muscle activation status to obtain the user's movements in three-dimensional space.

[0030] Muscle strength data records the magnitude and direction of the force generated by muscles during exercise. This data can be collected using force sensors. These sensors are built into sportswear and positioned over muscle activity areas. These force sensors can be muscle strength testers.

[0031] By combining muscle electrical signals, muscle force data, and joint angles, we can derive user motion characteristics. Motion characteristics refer to the characteristics of a user during exercise, such as different muscle contraction sequences and joint force characteristics.

[0032] The data transmission unit transmits the motion data collected by the sensor unit to the data processing unit in real time. The data processing unit can be located on the user or in a remote server. When the processing unit is located on the user, wired transmission can be used. When the processing unit is located in a remote server, wireless transmission can be used, such as Bluetooth transmission, Zigbee communication protocol transmission, ultra-wideband technology (UWB) transmission, or a proprietary radio frequency protocol.

[0033] The data processing unit receives the motion data collected by the sensor unit and performs data processing. Before data processing, the data processing unit may also perform data preprocessing to ensure the quality and consistency of the motion data collected by the sensor unit for data processing and improve the reliability of the analysis results obtained by data processing.

[0034] Data preprocessing includes noise reduction and data fusion. Noise reduction is used to remove noise and interference signals from motion data, while data fusion is used to synchronize the motion data collected by different sensor units in time and align them in the spatial coordinate system.

[0035] Noise reduction can be performed using adaptive filtering, Kalman filtering or wavelet transform noise reduction processing algorithms.

[0036] Spatiotemporal alignment technology can be used for data fusion, synchronizing motion data collected from different sensor units in time and aligning them in spatial coordinates, thereby obtaining more accurate and reliable acquisition results than a single sensor. The intelligent motion guidance system of this application uses data fusion technology to obtain fused data and provides motion guidance based on this fused data. Compared with methods that obtain data and provide motion guidance through inertial or optical systems, this system is more accurate and reliable.

[0037] The data processing unit can use data fusion combined with the first joint angle obtained by the flexible sensor to correct the drift error of the inertial measurement unit. The specific process includes: the flexible sensor collects the first joint angle; the inertial measurement unit collects the second joint angle; the first joint angle and the second joint angle are synchronized in time and aligned in the spatial coordinate system by data fusion; the second joint angle after synchronization in time and alignment in the spatial coordinate system is corrected by the first joint angle after synchronization in time and alignment in the spatial coordinate system, and then the inertial measurement unit is calibrated, so that the inertial measurement unit of the intelligent motion guidance system can continuously obtain accurate data. The intelligent motion guidance system of the present application, the sensor unit includes at least two sensors, a flexible sensor and an inertial sensor, and the drift error of the inertial sensor is corrected by the flexible sensor, which can make the system more stable.

[0038] In some embodiments, the flexible sensor is made of strain sensing material, which can be liquid metal such as gallium-based liquid metal. The flexible sensor made of liquid metal is a liquid metal flexible sensor, which has excellent conductive properties and can more sensitively collect electrical signals caused by micro-strain changes to obtain a more accurate first angle of the joint, thereby improving the accuracy of calibrating the inertial measurement unit and enhancing the performance of the intelligent motion guidance system to continuously obtain accurate data.

[0039] Through modeling, analysis, or calculations performed by the data processing unit, it is possible to calculate the three-dimensional motion trajectory of the joints, identify the user's current motion phase, determine the deviation between the user's motion and the standard motion, extract the user's motion characteristics, or predict possible dangerous movements. The following details the calculation of the three-dimensional motion trajectory of the joints, identification of the user's current motion phase, determination of the deviation between the user's motion and the standard motion, extraction of the user's motion characteristics, or prediction of possible dangerous movements.

[0040] Calculating the three-dimensional motion trajectory of a joint includes calculating the joint's three-dimensional motion trajectory based on joint angles and muscle activation. Muscle activation can be calculated using muscle electrical signals. The three-dimensional motion trajectories of multiple joints can be used to reconstruct the user's movements in three-dimensional space. Reconstructing the user's movements in three-dimensional space includes calculating joint angles and muscle electrical signals using multimodal fusion technology based on data collected by flexible sensors and an inertial measurement unit. Using the joint angles and muscle electrical signals in combination with a kinematic algorithm, the user's current displacement and joint torque are inferred. The user's movements in three-dimensional space are reconstructed based on the user's current position and relative displacement, as well as the joint torque. Joint torque refers to the magnitude and direction of the force acting on a joint that causes it to rotate. In some embodiments, the flexible sensor is made of a strain sensing material, which can be a liquid metal such as gallium-based liquid metal. Liquid metal flexible sensors have excellent electrical conductivity and can more sensitively capture electrical signals generated by microstrain changes, thereby obtaining a more accurate first joint angle, thereby improving the accuracy of the calculated three-dimensional motion trajectory of the joint.

[0041] Kinematic algorithms include forward kinematics and inverse kinematics. Forward kinematics can calculate the position of the human body's extremities, as well as the user's position and displacement, based on the body's joint angles. The position of the human body's extremities can be the position of the toes. For example, the hip and knee joint angles are obtained using flexible sensors and / or inertial measurement units, and the user's position and displacement are calculated using forward kinematics. Inverse kinematics can calculate the body's joint angles based on the position of the human body's extremities. For example, the knee and hip joint angles can be inferred from the three-dimensional coordinates of the toes.

[0042] Identifying the user's current motion stage includes using a statistical model, such as a hidden Markov model, to divide the motion into multiple motion cycles based on the periodicity of the motion data, and determining the user's current motion stage based on the motion cycles. The periodicity of the motion data can be the periodicity of joint positions, joint angles, or muscle strength data.

[0043] Determining the deviation between a user's motion and a standard motion includes: comparing the motion data collected by the sensor unit with the standard motion data to obtain a difference value; presetting an allowable deviation range based on the requirements of the standard motion data; and identifying a motion deviation when the difference value exceeds the allowable deviation range. The collected motion data can include joint position, joint angle, or muscle strength data.

[0044] In some embodiments, the standard action may be a historical motion data benchmark value, which is obtained by calculating an average value of the user's historical motion data.

[0045] Extracting user motion features involves building a joint model using the OpenSim simulation engine or a simplified rigid body model, and calculating joint torques based on joint angle and muscle force data. Joint torque refers to the magnitude and direction of the force acting on a joint to cause it to rotate. Machine learning methods are used to derive muscle activation timing from electromyographic (EMG) data. Finally, user motion feature vectors and features are derived from the joint torques and muscle activation timing. User motion features include both correct and incorrect motion features.

[0046] The user motion feature vector and user motion features obtained from the joint torque and muscle activation timing can be obtained using an attention mechanism, a traditional machine learning model (e.g., SVM, random forest) or a three-dimensional convolutional neural network (3D-CNN).

[0047] An incremental learning framework, such as parameter isolation technology, can be used to ensure that the extracted new user motion features do not destroy or overwrite the extracted original user motion features.

[0048] Predicting potentially dangerous maneuvers involves building and training a deep learning model based on a long short-term memory (LSTM) network. The deep learning model predicts movement data a predetermined time later based on input movement data; compares the predicted movement data a predetermined time later with standard movement data to determine a difference; sets a predefined tolerance range based on the standard movement data; and identifies a potentially dangerous maneuver when the difference exceeds the tolerance range. The input movement data can be collected joint position, joint angle, or muscle strength data.

[0049] The standard action data is obtained by calling the data storage unit, specifically including: calling the historical motion data of the data storage unit; calculating the average value of the user's historical motion data to obtain the historical motion data baseline value; using the historical motion data baseline value as the standard action data for predicting possible dangerous actions.

[0050] The data processing unit can be arranged on the user or located in a remote server. When the processing unit is arranged on the user, it can be integrated with the sensor unit.

[0051] The data processing unit may include multiple data processing modules, each of which is responsible for one of data preprocessing or data processing, and the content of data preprocessing or data processing that each data processing module is responsible for is different.

[0052] Any data processing module can be located in a data processing unit arranged on the user or in a data processing unit located in a remote server. The arrangement of the data processing modules can be flexibly selected according to the sports event and the data transmission capacity of the data transmission unit to improve the efficiency of data processing.

[0053] When the data processing module is located in a data processing unit placed on the user, a lightweight model built into the data processing unit facilitates data preprocessing or data processing by the data processing module. The lightweight model can adopt the TinyML architecture.

[0054] When the data processing module is located in the data processing unit of a remote server, blockchain technology can be used to implement multi-device collaborative computing to facilitate the data processing module to perform data preprocessing or data processing.

[0055] The processed data includes the three-dimensional motion trajectory of the joints, the user's current motion stage, the deviation between the user's motion and the standard motion, possible dangerous motions, and the user's motion characteristics. The motion guidance unit includes a motion guidance module and a decision analysis module. The motion guidance module is used to provide real-time motion guidance for the user's current motion stage based on the deviation between the user's motion and the standard motion, possible dangerous motions, or incorrect motion characteristics. The decision analysis module is used to build and update the user's ability model, generate training plans, or output reports.

[0056] The following is a detailed analysis of the working principles of the action guidance module and decision analysis module.

[0057] The action guidance module is deployed on the user's device, which can be a mobile phone, wristband, or other device. It provides guidance through the user's various sensory channels, helping them understand and execute correct actions. The action guidance module includes at least one of the following: a tactile guidance module, an auditory guidance module, and a visual guidance module.

[0058] The tactile guidance module is placed on the user's body. The tactile guidance module is equipped with a vibration device that can generate vibrations on the user's body to provide movement guidance. In some embodiments, the tactile guidance module can also use electrical stimulation, thermal guidance, or pneumatic tactile guidance for movement guidance.

[0059] The auditory guidance module can be an earphone that sends prompts or suggestions to the user in the form of voice to guide him to adjust his movements. In some embodiments, the voice can be generated using natural language generation technology.

[0060] The visual guidance module can be AR glasses, which generate animations through the AR glasses to provide action guidance.

[0061] The tactile, auditory, and visual guidance modules work together to enable users to receive information from multiple sensory channels. For example, the auditory guidance module uses natural language generation technology to issue voice prompts, such as "When squatting, move your center of gravity back 2 cm." Simultaneously with the voice prompts, the tactile guidance module uses vibrations or other tactile signals to provide tactile prompts at specific parts of the user's body (such as the waist or legs), helping the user perceive the actual position change of their center of gravity.

[0062] The action guidance module can issue an alarm to the user through one or more of the tactile guidance module, the auditory guidance module, or the visual guidance module based on the deviation between the user's action and the standard action determined by the data processing unit, the predicted possible dangerous action, or the extracted incorrect movement characteristics. For example, when a possible dangerous action is predicted, the tactile feedback module can be used to issue a vibration alarm.

[0063] In addition to alarm reminders, the system can also provide action guidance with content, including adjustments to the intensity, rhythm, or posture of the movement. The intelligent motion guidance system has a large language model interface, which can be used to call the large language model to generate action guidance with content.

[0064] For example, the three-dimensional motion trajectory of the joint is combined with the deviation between the user's action and the standard action, and the predicted dangerous or incorrect action features are input into the large language model, and the large language model outputs action guidance with guidance content. It should be noted that the large language model mentioned in this application refers to a large language model trained with the user's historical motion data on the basis of a general large language model such as Deepseek or Chatgpt. It can combine the input action deviation from the standard action, possible dangerous or incorrect action features, generate action guidance suggestions, and help users correct incorrect actions.

[0065] When there are multiple pieces of guidance, such as adjusting posture or changing the way of exerting force, the guidance can be sorted according to the importance of the guidance content and then provided in sequence.

[0066] In some embodiments, the intelligent sports guidance system can provide a unified large language model interface for multiple users, or it can set up an independent large language model interface for each individual user. The independent large language model interface can be connected to the large language model trained with the user's historical sports data to output targeted action guidance with guidance content for the user.

[0067] Action guidance with instructional content can also include sports medicine evidence corresponding to adjustments to the intensity, rhythm, or posture of the action. For example, sports medicine evidence recommends reducing the ankle dorsiflexion angle to reduce patellofemoral joint pressure. The sports medicine evidence can be output by a large language model, or a sports medicine rule library can be set up in the data storage unit. By calling the sports medicine rule library, corresponding sports medicine evidence can be generated for adjustments to the intensity, rhythm, or posture of the action to enhance user trust.

[0068] The decision analysis module is used to build and update user capability models, generate training plans, and output reports. The following is a detailed explanation of building and updating user capability models, generating training plans, and generating reports.

[0069] Build a user capability model: Historical motion data includes the user's historical joint position, historical joint angle, and historical muscle strength data; calculate the joint mobility based on the user's historical joint position and historical joint angle, and the joint mobility is used to evaluate the range of motion of each joint of the user; draw a force output curve based on the user's historical muscle strength data, and the force output curve is used to reflect the user's strength performance under different movements or loads; build a user capability model based on the joint mobility and force output curve.

[0070] Update the user ability model: recalculate the joint range of motion; redraw the force output curve; update the user ability model based on the recalculated joint range of motion and the redrawn force output curve.

[0071] Recalculating joint range of motion can be done by collecting joint angles using flexible sensors, inertial measurement units, or millimeter-wave radar, or by using the joint's three-dimensional motion trajectory. Redrawing the force output curve can be done by drawing muscle force data collected by force sensors.

[0072] A training plan can be generated in the following ways. The training plan includes specific training movements and corresponding load intensity:

[0073] The first method is to identify problems with movement strength and joint mobility in historical movements by using the built user capability model. Based on these problems, a corresponding training plan is generated by combining the large language model with the movement strength and joint mobility problems in historical movements.

[0074] The second method is to identify problems with the user's movement strength and joint mobility in the current movement stage from the updated user ability model. Based on these problems, the training plan is updated in combination with the large language model.

[0075] The third method is to generate a corresponding training plan based on the deviation between the user's action and the standard action, predict possible dangerous actions or incorrect action features, and combine the large language model to generate a corresponding training plan.

[0076] Generate a report: Compare the user's movements in 3D space with the standard movements to find out the differences between the two movements.

[0077] The generated report may be a visual report, which may be provided to the user through a visual guidance module, so that the user can clearly and intuitively understand the differences between his or her own actions and the standard actions in the current action stage of the user.

[0078] The generated report can be generated after the user finishes exercising, or it can be generated according to the exercise cycle.

[0079] In some embodiments, the flexible sensor can be made of an optical sensing material, which can be a flexible optical fiber combined with a fiber Bragg grating (FBG), and detects strain using changes in optical signals.

[0080] In some embodiments, the flexible sensor used in the present application includes a conductive composite material, which can be one or more of carbon nanotubes, silicone composite materials, or silver paste printed circuits.

[0081] In some embodiments, the flexible sensor used in this application that contacts the user's skin can be made of a biocompatible material, such as medical-grade silicone or thermoplastic polyurethane. In other embodiments, the flexible sensor used in this application that contacts the user's skin can be made of a smart-responsive material, such as a temperature-sensitive shape memory polymer (SMP) that changes shape based on temperature, or a dielectric elastomer that deforms under the action of an electric field.

[0082] In some embodiments, the circuit of the flexible sensor used in the present application can be a circuit embroidered with conductive yarn or a circuit made of liquid metal-filled microchannels.

[0083] In some embodiments, the data storage unit of the present application stores a user's motion patterns, common error types, and change trends. Motion patterns refer to the habitual motion patterns formed by a user during exercise, common error types refer to the errors frequently made by a user during exercise, and change trends refer to the changing trends of the user's motion data over different time periods. The user's motion patterns, common error types, and change trends can be provided to the decision analysis module for building and updating the user's ability model, generating training plans, and outputting reports.

[0084] The trend of change can be determined by establishing a graph-structured database in the data storage unit to correlate data across time periods. Specifically, the graph-structured database stores users as nodes, and their exercise records and related information as node attributes. By querying the graph-structured database, correlations can be established between exercise data from different time periods. By correlating short-term, medium-term, and long-term exercise data, trend of change can be determined.

[0085] In some embodiments, the data storage unit of the present application stores a sports science knowledge graph, which includes human anatomical structure, training adaptability principles, etc.

[0086] The content of the action instructions, such as the intensity, rhythm, or posture adjustments of the movements, can be verified through the sports science knowledge graph to ensure the safety of the action instructions. For example, the human anatomy structure in the sports science knowledge graph can be referenced to confirm whether the action instructions comply with anatomical constraints; and the principles of adaptive training can be referenced to confirm whether the action instructions comply with the principles of adaptive training.

[0087] In some embodiments, the present application provides an intelligent sports guidance system, including:

[0088] The intelligent motion guidance system includes a sensor unit, a data transmission unit, a data processing unit, a motion guidance unit, and a data storage unit;

[0089] The sensor unit is used to collect the user's motion data; the sensor unit includes at least two types of sensors;

[0090] The data transmission unit transmits the motion data collected by at least two sensors to the data processing unit in real time;

[0091] The data storage unit is used to store standard action data;

[0092] The data processing unit processes the motion data collected by the at least two sensors, and transmits the processed data to the motion guidance unit through the data transmission unit;

[0093] The data processing includes fusing motion data collected by at least two sensors to obtain fused data; the data processing also includes at least one of the following steps: comparing the fused data with standard motion data to determine the deviation between the user's motion and the standard motion; predicting possible dangerous motions based on the fused data;

[0094] The motion guidance unit includes an action guidance module and a decision analysis module. The action guidance module is used to provide real-time action guidance based on the action deviation between the user's action and the standard action and the possible dangerous action; the decision analysis module is used to generate a corresponding training plan based on the action deviation between the user's action and the standard action and the possible dangerous action.

[0095] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and replacements can be made without departing from the technical principles of the present application. These improvements and replacements should also be regarded as the scope of protection of the present application.

Claims

1. An intelligent sports guidance system, characterized by: The intelligent motion guidance system includes a sensor unit, a data transmission unit, a data processing unit, a motion guidance unit, and a data storage unit; The sensor unit is used to collect the user's motion data; the sensor unit includes at least two types of sensors; The data transmission unit transmits the motion data collected by at least two sensors to the data processing unit in real time; The data storage unit is used to store standard action data; The data processing unit processes the motion data collected by the at least two sensors, and transmits the processed data to the motion guidance unit through the data transmission unit; The data processing includes fusing motion data collected by at least two sensors to obtain fused data; the data processing also includes at least one of the following steps: comparing the fused data with standard motion data to determine the deviation between the user's motion and the standard motion; predicting possible dangerous motions based on the fused data; The motion guidance unit includes an action guidance module and a decision analysis module. The action guidance module is used to provide real-time action guidance based on the action deviation between the user's action and the standard action and the possible dangerous action; the decision analysis module is used to generate a corresponding training plan based on the action deviation between the user's action and the standard action and the possible dangerous action.

2. The intelligent sports guidance system according to claim 1, characterized in that: One of the at least two sensors is a flexible sensor.

3. The intelligent sports guidance system according to claim 2, characterized in that: The flexible sensor is a gallium-based liquid metal flexible sensor.

4. The intelligent sports guidance system according to claim 2, characterized in that: Another of the at least two sensors is an inertial measurement unit.

5. The intelligent sports guidance system according to claim 4, characterized in that: The motion data includes joint angles; the joint angles include a first joint angle and a second joint angle; the first joint angle, muscle electrical signals, and position information of the human body end can be obtained by the flexible sensor; The joint position and the second joint angle are collected by an inertial measurement unit.

6. The intelligent sports guidance system according to claim 5, characterized in that: The data processing unit can compensate for the drift error of the inertial measurement unit by utilizing data fusion combined with the first joint angle acquired by the flexible sensor.

7. The intelligent sports guidance system according to claim 5, characterized in that: The motion data includes muscle strength data; the sensor unit includes a force sensor, and the force sensor is used to collect muscle strength data.

8. The intelligent sports guidance system according to claim 7, characterized in that: The data storage unit further stores historical motion data, the historical motion data including historical joint positions, historical joint angles, and historical muscle strength data of the user, and the decision analysis module can calculate the joint range of motion based on the historical joint positions and historical joint angles of the user; Draw a strength output curve based on the user's historical muscle strength data; Build a user ability model based on joint range of motion and force output curves.

9. The intelligent sports guidance system according to claim 8, characterized in that: The decision analysis module can calculate the updated joint range of motion based on the collected joint positions and joint angles; and draw the updated force output curve based on the collected muscle strength data; The user capability model is updated according to the updated joint range of motion and the updated force output curve.

10. The intelligent sports guidance system according to claim 7, characterized in that: The motion data includes muscle electrical signals; the sensor unit includes an electromyographic sensor, and the electromyographic sensor is used to collect muscle electrical signals.

11. The intelligent sports guidance system according to claim 10, characterized in that: The data processing unit can calculate the three-dimensional motion trajectory of the joint based on the first joint angle and the muscle electrical signal.

12. The intelligent sports guidance system according to claim 11, characterized in that: The data processing unit can calculate the joint torque based on the preset joint model and the collected joint angle and muscle strength data; Using a machine learning method to obtain a muscle activation timing sequence from the muscle electrical signal; A user motion feature vector and a user motion feature are obtained from the joint torque and muscle activation timing.

13. The intelligent sports guidance system according to any one of claims 1 to 3, characterized in that: The action guidance module includes at least one guidance module among a tactile guidance module, an auditory guidance module, and a visual guidance module.

14. The intelligent sports guidance system according to any one of claims 1 to 3, characterized in that: The intelligent sports guidance system is provided with a large language model interface, and the action guidance module is used to generate action guidance with guidance content by calling the large language model through the large language model interface according to the deviation between the user's action and the standard action and the possible dangerous action; The decision analysis module is used to generate a corresponding training plan by calling the large language model through the large language model interface according to the action deviation between the user action and the standard action and the possible dangerous action.

15. The intelligent sports guidance system according to any one of claims 1 to 3, characterized in that: The data processing unit can utilize a hidden Markov model to divide the motion into a plurality of motion cycles according to the periodicity of the motion data, and determine the current motion stage of the user according to the motion cycles.

16. The intelligent sports guidance system according to any one of claims 1 to 3, characterized in that: The data processing unit can compare the collected motion data with the standard action data to obtain a difference value; preset an allowable deviation range according to the requirements of the standard action data; and identify it as an action deviation when the difference value exceeds the allowable deviation range.

17. The intelligent sports guidance system according to any one of claims 1 to 3, characterized in that: The data processing unit is capable of inputting the collected motion data into a preset deep learning model, and the preset deep learning model outputs the predicted motion data after a predetermined time; the predicted motion data after the predetermined time is compared with the standard motion data to obtain a difference value; The allowable deviation range is preset according to the requirements of the standard action data; when the difference value exceeds the allowable deviation range, it is identified as a possible dangerous action.

Images