Motion posture recognition method and device based on large model

Abstract

The invention provides a motion posture recognition method and device based on a large model, and the method comprises the following steps: collecting human motion data through an energy button, and obtaining a motion data set; based on the motion data set, training to obtain an identification large model; and recognizing the human body motion posture by using the large recognition model to obtain a posture recognition result. According to the technical scheme, the large recognition model can process and analyze a large amount of motion data, and precise recognition of the motion posture is achieved.

Description

Background Art

[0002] With the popularization of smart wearable devices and the rapid development of big data technology, the accurate recognition of human motion postures has become an important requirement in fields such as health monitoring, sports performance analysis, and human-computer interaction. Traditional motion posture recognition methods mainly rely on image analysis or sensor data such as accelerometers and gyroscopes, but these methods have problems such as complex equipment, poor universality, and insufficient recognition accuracy. Summary of the Invention

[0003] This application provides a motion posture recognition method and device based on a large model to improve the accuracy of motion posture recognition.

[0004] In a first aspect, a motion posture recognition method based on a large model is provided, including the following steps:

[0005] Collect human motion data using an energy button to obtain a motion data set;

[0006] Train a recognition large model based on the motion data set;

[0007] Use the recognition large model to recognize human motion postures to obtain a posture recognition result.

[0008] In the above technical solution, by collecting human motion data using an energy button to obtain a motion data set; training a recognition large model based on the motion data set; using the recognition large model to recognize human motion postures to obtain a posture recognition result; the large model can process and analyze a large amount of motion data to achieve accurate recognition of motion postures; at the same time, the large model can be continuously optimized and upgraded according to actual needs to improve recognition accuracy and stability.

[0009] In a specific feasible implementation, it further includes:

[0010] Optimize the recognition accuracy and stability of the recognition large model;

[0011] Maintain and update the energy button.

[0012] In a specific feasible implementation, the energy button includes a button body, and the following are arranged inside the button body:

[0013] A position sensor for collecting limb position information during human motion;

[0014] A wireless communication module for real-time transmission of the collected data to a cloud server or a smart device;

[0015] A control module for controlling data collection and transmission.

[0016] In a specific feasible implementation, the interior of the button body is provided with:

[0017] A storage module for temporarily storing the collected data;

[0018] A power module for powering the energy button.

[0019] In a specific feasible implementation, the recognition large model adopts a deep learning algorithm.

[0020] In a second aspect, a motion posture recognition device based on a large model is provided, including:

[0021] A data set module for collecting human motion data using an energy button to obtain a motion data set;

[0022] A recognition large model module for training a recognition large model based on the motion data set;

[0023] A posture recognition module for using the recognition large model to recognize the human motion posture to obtain a posture recognition result.

[0024] In the above technical solution, by collecting human motion data using an energy button, a motion data set is obtained; based on the motion data set, a recognition large model is trained; the recognition large model is used to recognize the human motion posture to obtain a posture recognition result; the large model can process and analyze a large amount of motion data to achieve accurate recognition of the motion posture; at the same time, the large model can be continuously optimized and upgraded according to actual needs to improve the recognition accuracy and stability.

[0025] In a specific feasible implementation, it further includes an optimization module for:

[0026] Optimizing the recognition accuracy and stability of the recognition large model;

[0027] Maintaining and updating the energy button.

[0028] In a specific feasible implementation, the energy button includes a button body, and the interior of the button body is provided with:

[0029] A position sensor for collecting limb position information during human motion;

[0030] A wireless communication module for real-time transmitting the collected data to a cloud server or a smart device;

[0031] A control module for controlling the data collection and transmission.

[0032] In a specific feasible implementation, the interior of the button body is provided with:

[0033] A storage module for temporarily storing the collected data;

[0034] A power supply module for powering the energy button. Brief Description of the Drawings

[0035] Figure 1 It is a flowchart of the motion pose recognition method based on a large model provided by an embodiment of the present application;

[0036] Figure 2 It is a structural block diagram of the motion pose recognition device based on a large model provided by an embodiment of the present application. Detailed Embodiments

[0037] The present application will be further described in detail below with reference to the drawings and embodiments. Through these descriptions, the features and advantages of the present application will become more clearly defined.

[0038] The special term "exemplary" here means "serving as an example, embodiment, or illustrative". Any embodiment described as "exemplary" here does not necessarily have to be interpreted as superior to or better than other embodiments. Although various aspects of the embodiments are shown in the drawings, the drawings do not have to be drawn to scale unless otherwise specified.

[0039] In addition, the technical features involved in different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0040] To facilitate the understanding of the motion pose recognition method and device based on a large model provided by the embodiments of the present application, its application scenario will be described first. The motion pose recognition method and device based on a large model provided by the embodiments of the present application are used to improve the accuracy of motion pose recognition. With the popularization of smart wearable devices and the rapid development of big data technology, the accurate recognition of human motion poses has become an important requirement in the fields of health monitoring, motion analysis, human-computer interaction, etc. Traditional motion pose recognition methods mainly rely on image analysis or sensor data such as accelerometers and gyroscopes, but these methods have problems such as complex devices, poor universality, and insufficient recognition accuracy. Therefore, the embodiments of the present application provide a motion pose recognition method and device based on a large model to improve the accuracy of motion pose recognition. It will be described in detail below with reference to specific drawings and embodiments.

[0041] Refer to Figure 1 and Figure 2 , Figure 1 It is a flowchart of the motion pose recognition method based on a large model provided by an embodiment of the present application; Figure 2 It is a structural block diagram of the motion pose recognition device based on a large model provided by an embodiment of the present application.

[0042] In Figure 1In the embodiment of the present application, a motion posture recognition method based on a large model is provided, comprising the following steps:

[0043] Use the energy button to collect human motion data and obtain a motion data set;

[0044] Based on the motion data set, a large recognition model is trained;

[0045] The large recognition model is used to recognize the human body's motion posture and obtain the posture recognition result.

[0046] In the above technical solution, the energy button is used to collect human motion data to obtain a motion data set; based on the motion data set, a large recognition model is trained; the human motion posture is recognized using the large recognition model to obtain a posture recognition result; the large model can process and analyze a large amount of motion data to achieve accurate recognition of motion posture; at the same time, the large model can be continuously optimized and upgraded according to actual needs to improve recognition accuracy and stability.

[0047] Specifically, by using the energy button to collect human motion data and then training the recognition model to recognize human motion postures, the beneficial effects include:

[0048] Improve recognition accuracy: By collecting a large amount of human motion data, the trained recognition model can learn rich motion features, thereby achieving accurate recognition of human motion postures. This big data-based training method can significantly improve recognition accuracy and reliability compared to traditional methods.

[0049] Enhanced stability: When processing and analyzing large amounts of motion data, large models can learn the inherent laws and changing trends of the data, thereby enhancing their ability to adapt to complex motion scenes. This helps improve the stability of the recognition system in different environments and conditions, and reduces misidentification or missed recognition.

[0050] Achieve continuous optimization and upgrade: The large recognition model has strong learning ability and adaptability, and can be continuously optimized and upgraded according to actual needs. This means that with the continuous advancement of technology and the continuous increase of human motion data, the performance of the recognition system will continue to improve, maintaining its advancement and practicality.

[0051] Expand application scenarios: Accurate human motion posture recognition technology can be widely used in sports training, medical rehabilitation, human-computer interaction, virtual reality and other fields. The large recognition model obtained through training can provide reliable technical support for these fields and promote the development and innovation of related technologies.

[0052] Enhancing user experience: In fields such as sports training and medical rehabilitation, precise human motion posture recognition technology can help users better understand their own motion states and health conditions, thus formulating more scientific and reasonable training plans and treatment programs. This will greatly enhance the user experience and satisfaction.

[0053] Promoting the development of data science and artificial intelligence technologies: The collection and analysis of human motion data are important research directions in the fields of data science and artificial intelligence. By training a large recognition model to recognize human motion postures, it can not only promote the technological development in related fields but also provide useful references for other similar data analysis and recognition tasks.

[0054] In summary, the process of using energy buttons to collect human motion data and training a large recognition model for human motion posture recognition has significant beneficial effects in aspects such as improving recognition accuracy, enhancing stability, achieving continuous optimization and upgrading, expanding application scenarios, enhancing user experience, and promoting the development of data science and artificial intelligence technologies.

[0055] In a specific feasible implementation, the large recognition model adopts a recurrent neural network (RNN) with a Transformer architecture, including:

[0056] 1. Data collection and preprocessing: Use wearable devices such as energy buttons to collect human motion data, including information such as acceleration, angular velocity, and position.

[0057] Clean and preprocess the collected data to remove noise and outliers to ensure data quality.

[0058] Label the data according to task requirements to generate a training set, a validation set, and a test set.

[0059] 2. Model design and selection:

[0060] According to the characteristics of human motion posture recognition, select the Transformer model architecture;

[0061] Design the input and output formats of the model to ensure that the model can process the collected motion data.

[0062] Adjust the parameters and number of layers of the model according to requirements to balance the complexity and performance of the model.

[0063] 3. Model training and optimization:

[0064] Use the preprocessed training set to train the model and update the model parameters through the backpropagation algorithm.

[0065] During the training process, use the validation set to evaluate the model and adjust the hyperparameters and learning rate to optimize the model performance.

[0066] Data augmentation techniques are adopted, including random cropping, rotation, and flipping, to increase the diversity of training data and improve the generalization ability of the model.

[0067] 4. Model evaluation and testing:

[0068] Use the test set to evaluate the trained model and calculate accuracy, recall, and F1-score metrics.

[0069] Conduct visual analysis on the recognition results of the model.

[0070] Optimize the model according to the evaluation results, including increasing the data volume and adjusting the model structure.

[0071] 5. Model deployment and application:

[0072] Deploy the trained model to actual application scenarios, such as sports training and medical rehabilitation.

[0073] Optimize the performance of the model according to actual needs, including reducing the inference time and resource consumption.

[0074] Provide user-friendly interfaces and tools to facilitate the use and customization of the model by users.

[0075] In a specific implementable embodiment, it further includes:

[0076] Optimize the recognition accuracy and stability of the recognition large model;

[0077] Maintain and update the energy button.

[0078] In a specific implementable embodiment, the energy button includes a button body, and inside the button body are provided:

[0079] A position sensor for collecting limb position information during human movement;

[0080] A wireless communication module for real-time transmission of the collected data to a cloud server or a smart device;

[0081] A control module for managing data collection and transmission. Specifically, when implemented, the energy button has a size of a cylinder with a radius of 0.5 mm and a height of 0.5 mm, and is worn by means of detachable bioadhesive.

[0082] In a specific implementable embodiment, inside the button body are provided:

[0083] A storage module for temporarily storing the collected data;

[0084] A power module for powering the energy button.

[0085] In a specific feasible embodiment, the recognition large model adopts a deep learning algorithm.

[0086] Specifically, the large model-based motion posture recognition method includes:

[0087] 1. Energy button design:

[0088] The energy button includes a button body, and a position sensor, a wireless communication module, a control module, a storage module, and a power module are arranged inside the button body.

[0089] The position sensor is used to collect limb position information during human movement, such as acceleration, angular velocity, etc.

[0090] The wireless communication module is used to transmit the collected data to the cloud or intelligent device in real time.

[0091] The control module is responsible for coordinating the work of each module to ensure the smooth progress of data collection and transmission.

[0092] The storage module is used to temporarily store the collected data to prevent data loss.

[0093] The power module provides a stable power supply for the energy button.

[0094] 2. Data collection and transmission:

[0095] The user wears the energy button to exercise, and the position sensor collects human motion data in real time.

[0096] The collected data is transmitted to the cloud server or intelligent device in real time through the wireless communication module.

[0097] 3. Large model construction and training:

[0098] A large model is constructed on the cloud server. This model is based on a deep learning algorithm and can process and analyze a large amount of motion data.

[0099] A large number of labeled motion data is used to train the model so that it can accurately identify various motion postures.

[0100] 4. Motion posture recognition:

[0101] After the cloud server receives the data transmitted by the energy button, it uses the large model for data analysis and posture recognition.

[0102] The recognition result is fed back to the user in real time through the wireless communication module, or displayed to the user through the cloud platform.

[0103] 5. Optimization and feedback:

[0104] Continuously optimize the recognition accuracy and stability of the large model according to user feedback and actual needs.

[0105] Regularly maintain and update the energy button to ensure its long-term stable operation.

[0106] In the above technical solution, the beneficial effects include:

[0107] Portability: The energy button is small in size and light in weight, making it convenient for users to wear and carry.

[0108] Accuracy: The large model can process and analyze a large amount of motion data to achieve accurate recognition of motion postures.

[0109] Universality: This method is applicable to various motion scenarios and populations, with wide applicability.

[0110] Real-time: Data collection and transmission are carried out in real time, enabling users to timely understand their own motion states.

[0111] Scalability: The large model can be continuously optimized and upgraded according to actual needs to improve recognition accuracy and stability.

[0112] In Figure 2 the embodiments of the present application provide a motion posture recognition device based on a large model, including:

[0113] A dataset module for collecting human motion data using an energy button to obtain a motion dataset;

[0114] An identification large model module for training an identification large model based on the motion dataset;

[0115] A posture recognition module for using the identification large model to recognize human motion postures to obtain posture recognition results.

[0116] In the above technical solution, by using an energy button to collect human motion data to obtain a motion dataset; training an identification large model based on the motion dataset; using the identification large model to recognize human motion postures to obtain posture recognition results; the large model can process and analyze a large amount of motion data to achieve accurate recognition of motion postures; at the same time, the large model can be continuously optimized and upgraded according to actual needs to improve recognition accuracy and stability.

[0117] In a specific feasible implementation, it further includes an optimization module for:

[0118] Optimizing the recognition accuracy and stability of the identification large model;

[0119] Maintaining and updating the energy button.

[0120] In a specific feasible embodiment, the energy button includes a button body, and inside the button body are provided:

[0121] A position sensor for collecting limb position information during human movement;

[0122] A wireless communication module for real-time transmitting the collected data to a cloud server or a smart device;

[0123] A control module for controlling data collection and transmission.

[0124] In a specific feasible embodiment, inside the button body are provided:

[0125] A storage module for temporarily storing the collected data;

[0126] A power module for powering the energy button.

[0127] Those skilled in the art of the present application know that the present application can be implemented as a system, a method, or a computer program product.

[0128] Therefore, the present disclosure can be specifically implemented in the following forms, that is: it can be completely hardware, can also be completely software (including firmware, resident software, microcode, etc.), and can also be a combination of hardware and software, generally referred to as "circuit", "module" or "system" in this article. In addition, in some embodiments, the present application can also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable media contains computer-readable program code.

[0129] Any combination of one or more computer-readable media can be adopted. The computer-readable media can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or component, or any combination of the above. More specific examples (non-exhaustive list) of the computer-readable storage medium include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, the computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, device, or component.

[0130] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limitations on the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application. On this basis, various substitutions and improvements can be made to the present application, and all of these fall within the protection scope of the present application.

Claims

1. A motion posture recognition method based on a large model, characterized in that Including the following steps: Using an energy button to collect human motion data to obtain a motion data set; Training a recognition large model based on the motion data set; Using the recognition large model to recognize human motion postures to obtain posture recognition results.

2. The method for recognizing motion postures based on a large model according to claim 1, wherein, It also includes: Optimizing the recognition accuracy and stability of the recognition large model; Maintaining and updating the energy button.

3. The method for recognizing motion postures based on a large model according to claim 2, wherein, The energy button includes a button body, and inside the button body are provided: A position sensor for collecting limb position information during human motion; A wireless communication module for transmitting the collected data to a cloud server or a smart device in real time; A control module for controlling data collection and transmission.

4. The method for recognizing motion postures based on a large model according to claim 3, wherein, Inside the button body are provided: A storage module for temporarily storing the collected data; A power module for supplying power to the energy button.

5. The method for recognizing motion postures based on a large model according to claim 4, wherein, The recognition large model adopts a deep learning algorithm.

6. A motion posture recognition device based on a large model, characterized in that, Including: A data set module for using an energy button to collect human motion data to obtain a motion data set; A recognition large model module for training a recognition large model based on the motion data set; A posture recognition module for using the recognition large model to recognize human motion postures to obtain posture recognition results.

7. The motion posture recognition device based on a large model according to claim 6, wherein It also includes an optimization module for: Optimizing the recognition accuracy and stability of the recognition large model; Maintaining and updating the energy button.

8. The motion posture recognition device based on a large model according to claim 7, wherein, The energy button includes a button body, and inside the button body are provided: A position sensor for collecting limb position information during human motion; A wireless communication module for transmitting the collected data to a cloud server or a smart device in real time; A control module for controlling data collection and transmission.

9. The motion posture recognition device based on a large model according to claim 8, wherein, Inside the button body are provided: A storage module for temporarily storing the collected data; A power module for supplying power to the energy button.

10. The motion posture recognition device based on a large model according to claim 9, characterized in that, The recognition large model adopts a deep learning algorithm.

Images