Human posture auxiliary correction method and system based on deep learning

Abstract

The present application provides a method and system for assisting human posture correction based on deep learning, which relates to the technical field of posture correction, including: receiving a target posture correction scheme of a target user; generating a first real-time posture sequence when the target user performs posture correction training; calling a correction limit reminder to perform correction abnormality pattern recognition on the first real-time posture sequence, and generating a posture abnormality pattern; if the posture abnormality pattern is the first abnormality pattern, connecting a surface electromyography sensor, and receiving electromyography sensing data; performing muscle tension recognition based on the electromyography sensing data, and performing posture correction abnormality warning according to the muscle tension recognition result. This application can solve the technical problem in the prior art that it is difficult to provide intelligent feedback based on the user's real-time posture, which makes it difficult to achieve personalized and accurate posture correction, and provide users with more accurate posture correction feedback, thereby achieving the technical effect of ensuring the safety and effectiveness of posture correction training.

Description

Technical Field

[0001] The present application relates to the technical field of posture correction, and in particular to a method and system for assisting human posture correction based on deep learning. Background Art

[0002] As poor posture becomes more common in modern people's daily life and work, long-term incorrect posture may lead to health problems such as muscle strain and bone deformation. Especially in the workplace, such as office workers working at a desk for a long time, or lack of correct posture awareness in daily life, the risk of chronic pain and injury will increase. Therefore, posture correction training has become an important means for many people to maintain their health.

[0003] At present, traditional posture correction methods mainly rely on manual correction or reminders and feedback provided by simple devices, which have many limitations. For example, manual correction relies on the guidance of coaches or professionals, which is costly and cannot provide real-time and continuous correction supervision. Simple posture correction devices can usually only provide basic reminder functions, and it is difficult to provide intelligent feedback based on the user's real-time posture, nor can they identify the user's muscle tension status, making it difficult to achieve personalized and accurate posture correction.

[0004] In summary, the prior art has a technical problem that it is difficult to provide intelligent feedback based on the user's real-time posture, which makes it difficult to achieve personalized and accurate posture correction. Summary of the invention

[0005] The purpose of this application is to provide a method and system for assisting human posture correction based on deep learning, so as to solve the technical problem in the prior art that it is difficult to provide intelligent feedback based on the user's real-time posture, resulting in difficulty in achieving personalized and accurate posture correction.

[0006] In view of the above problems, the present application provides a method and system for assisting human posture correction based on deep learning.

[0007] In the first aspect, the present application provides a human posture auxiliary correction method based on deep learning, and the human posture auxiliary correction method based on deep learning is implemented by a human posture auxiliary correction system based on deep learning, wherein the human posture auxiliary correction method based on deep learning includes: receiving a target posture correction scheme of a target user, wherein the target posture correction scheme includes a preset correction starting posture and a preset correction end posture; when the target user performs posture correction training, connecting a posture capture network to perform posture monitoring on the target user to generate a first real-time posture time series; based on the preset correction starting posture and the preset correction end posture, calling a correction limit reminder to perform correction abnormality pattern recognition on the first real-time posture time series to generate a posture abnormality pattern; if the posture abnormality pattern is the first abnormal mode, connecting a surface electromyography sensor to receive electromyography sensing data; performing muscle tension recognition based on the electromyography sensing data, and issuing a posture correction abnormality warning according to the muscle tension recognition result.

[0008] In the second aspect, the present application also provides a human posture auxiliary correction system based on deep learning, which is used to execute the human posture auxiliary correction method based on deep learning as described in the first aspect, wherein the human posture auxiliary correction system based on deep learning includes: a correction scheme receiving module, which is used to receive the target posture correction scheme of the target user, wherein the target posture correction scheme includes a preset correction starting posture and a preset correction end posture; a posture monitoring module, which is used to connect to a posture capture network to monitor the posture of the target user when the target user performs posture correction training, and generate a first real-time posture time series; a correction abnormal pattern recognition module, which is used to call a correction limit reminder to perform correction abnormal pattern recognition on the first real-time posture time series based on the preset correction starting posture and the preset correction end posture, and generate a posture abnormal pattern; a surface electromyography sensing module, which is used to connect a surface electromyography sensor and receive electromyography sensing data if the posture abnormal pattern is the first abnormal pattern; a muscle tension recognition module, which is used to recognize muscle tension based on the electromyography sensing data, and issue a posture correction abnormality warning according to the muscle tension recognition result.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] Receive the target posture correction scheme of the target user, wherein the target posture correction scheme includes a preset correction starting posture and a preset correction end posture; when the target user performs posture correction training, connect the posture capture network to monitor the posture of the target user and generate a first real-time posture sequence; based on the preset correction starting posture and the preset correction end posture, call the correction limit reminder to identify the correction abnormality mode of the first real-time posture sequence and generate a posture abnormality mode; if the posture abnormality mode is the first abnormal mode, connect the surface electromyography sensor to receive electromyography sensing data; perform muscle tension recognition based on the electromyography sensing data, and perform posture correction abnormality warning according to the muscle tension recognition result. Through the posture capture network and electromyography sensing technology, the user's posture changes and muscle tension state are monitored in real time, and abnormality recognition and warning are performed based on the deep learning algorithm, so as to provide the user with more accurate posture correction feedback, and achieve the technical effect of ensuring the safety and effectiveness of posture correction training.

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0013] Figure 1 This is a flowchart of the human body posture auxiliary correction method based on deep learning in this application;

[0014] Figure 2 This is a structural diagram of the human posture auxiliary correction system based on deep learning in this application.

[0015] Description of reference numerals:

[0016] Correction scheme receiving module 11, posture monitoring module 12, correction abnormal pattern recognition module 13, surface electromyography sensing module 14, muscle tension recognition module 15. DETAILED DESCRIPTION

[0017] This application solves the technical problem in the prior art that it is difficult to provide intelligent feedback based on the user's real-time posture, which makes it difficult to achieve personalized and accurate posture correction by providing a human posture auxiliary correction method and system based on deep learning. Through posture capture network and electromyography sensing technology, the user's posture changes and muscle tension state are monitored in real time, and abnormal identification and early warning are performed based on deep learning algorithms, thereby providing users with more accurate posture correction feedback, achieving the technical effect of ensuring the safety and effectiveness of posture correction training.

[0018] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0019] For example, please refer to the attached Figure 1 The present application provides a human body posture auxiliary correction method based on deep learning, wherein the human body posture auxiliary correction method based on deep learning is applied to a human body posture auxiliary correction system based on deep learning, and the human body posture auxiliary correction method based on deep learning specifically includes the following steps:

[0020] Step 1: Receive a target posture correction scheme of a target user, wherein the target posture correction scheme includes a preset correction starting posture and a preset correction end posture.

[0021] Specifically, the target posture correction scheme is a posture adjustment plan tailored for the target user by professional and technical personnel in this field, with the aim of helping the target user gradually adjust from the current posture to the ideal correction posture. Specifically, the target posture correction scheme includes two key postures: a preset correction starting posture and a preset correction end posture. Among them, the preset correction starting posture refers to the initial body posture when the user starts the correction, such as an undesirable posture caused by bad habits or long-term incorrect posture. For example, a user who works at a desk for a long time may have a starting posture with a bent back and forward shoulders. The preset correction end posture is the target posture, that is, the ideal body posture that the user should achieve. For example, the end posture of this user may be with a straight back and naturally drooping shoulders.

[0022] Step 2: When the target user is performing posture correction training, a posture capture network is connected to monitor the posture of the target user to generate a first real-time posture time series.

[0023] Specifically, the posture capture network consists of multiple sensors or cameras, which are used to capture the movements and postures of the target user during training. These sensors are usually installed in the user's training environment or worn directly on the user to ensure accurate capture of the user's body joints, limb movements, and posture changes. In this process, the user's first real-time posture sequence is generated, which records the user's posture changes from the beginning of training to the current moment. The first real-time posture sequence refers to the continuous record of posture changes in the time dimension, similar to a dynamic video record, which can capture the user's posture changes at every moment. For example, if the user is undergoing spinal correction training, the posture capture network will monitor the user's entire process from standing to bending over, and then standing upright, capturing the details of the body posture at each moment, thereby continuously monitoring the user's posture changes to ensure the effectiveness and safety of the correction training.

[0024] Step 3: Based on the preset correction starting point posture and the preset correction end point posture, a correction limit reminder is called to perform correction abnormality pattern recognition on the first real-time posture sequence to generate a posture abnormality pattern.

[0025] Specifically, when the user is doing correction training, the correction limit reminder is called to analyze the user's posture data in real time. The correction limit reminder is used to detect whether the user's posture deviates from the preset correction route and identify abnormal patterns that may occur during the posture adjustment process. It mainly determines whether the user's actual posture meets expectations by analyzing the first real-time posture time series, that is, the time series of the recorded user posture during the training process.

[0026] In this process, the user's first real-time posture timing is compared with the preset correction starting posture and the preset correction end posture to detect whether there is an abnormal correction mode. The abnormal correction mode refers to the situation where the user's posture deviates from the normal training trajectory during the correction process. For example, suppose the user is performing correction training to straighten the waist, the starting posture is "bent waist" and the end posture is "straight waist". If during the training process, it is detected that the user's waist posture has not been effectively adjusted to the target position, the correction limit reminder will recognize that this is the first abnormal mode. If a reverse bend occurs during the adjustment process (for example, too much tilt in the opposite direction), the correction limit reminder will recognize that this is the second abnormal mode.

[0027] Step 4: If the abnormal posture mode is the first abnormal mode, connect the surface electromyography sensor to receive electromyography sensing data.

[0028] Specifically, during the posture correction process, when the target user's posture abnormality mode is detected as the first abnormal mode, an additional monitoring step is immediately initiated, that is, connecting to the surface electromyography sensor. The first abnormal mode usually refers to a slight deviation in the user's posture adjustment process, and the posture adjustment direction is basically correct. In order to more accurately assess the user's physical condition, a surface electromyography sensor is used to obtain the user's muscle activity data. The surface electromyography sensor is a non-invasive bioelectric signal detection device that measures the electrical signals generated by muscle contraction through electrodes attached to the surface of the user's skin. This step is very important because even if the user's posture looks correct, if the muscles are overly tense, it may cause potential muscle strain or discomfort. For example, when the user is doing back correction training, although the user's back posture is close to the preset target posture, if the muscle electrical activity is too high, it means that the user is trying hard to maintain the posture, which may cause excessive tension in the back muscles.

[0029] By connecting to the surface electromyography sensor, the user's electromyography sensing data can be received in real time. This data reflects the tension and activity pattern of the user's muscles. These electromyography data can be analyzed later and combined with the user's real-time posture data to comprehensively evaluate the user's physical condition and ensure that the user not only achieves the preset posture goals during the posture correction process, but also maintains the natural and relaxed state of the muscles.

[0030] Step 5: Perform muscle tension recognition based on the electromyographic sensing data, and issue a posture correction abnormality warning according to the muscle tension recognition result.

[0031] Specifically, electromyographic sensing data refers to the user's muscle activity data collected by the surface electromyographic sensor, which reflects the contraction strength and tension state of the user's muscles during the posture adjustment process. By analyzing these data, the user's muscle tension, that is, the tension of the muscles in a specific posture, can be evaluated. The pre-trained tension recognition network is used to process the electromyographic sensing data to identify the user's muscle state during the correction training. After identifying the user's muscle tension, these recognition results are used to determine whether there are abnormalities in the posture correction process. The role of the posture correction abnormality warning is to promptly remind the target user to prevent continuing to maintain an inappropriate posture or excessive force when the muscle tension is abnormal. For example, if the user is performing a posture correction training with a straight waist, if it is detected that the tension of the user's waist muscles is too high, it indicates that the user may be maintaining the posture by over-contracting the waist muscles, which may cause muscle overwork. At this time, an abnormal warning will be triggered to remind the user to adjust the posture and reduce the tension of the muscles.

[0032] This process ensures that the user not only achieves the desired posture goals, but also avoids potential injuries caused by excessive muscle tension or improper relaxation.

[0033] Further, step 2 of this application includes:

[0034] The posture capture network includes multiple visual sensors, which are arranged at local training joints of the target user; the data collected by the multiple visual sensors are processed by the data processing layer in the posture capture network to generate the first real-time posture time series; wherein the first real-time posture time series includes the continuously changing posture of the target user in the time dimension.

[0035] Specifically, the posture capture network is a core module used to monitor user posture changes in real time. It includes multiple visual sensors installed in key parts of the user, especially in local training joints during posture correction training. For example, in shoulder correction training, sensors may be installed around the user's shoulder joints, elbow joints, and shoulder blades to accurately capture shoulder posture changes. Similarly, in waist correction, sensors are deployed in key locations such as the spine and hip joints.

[0036] The main function of the visual sensor is to capture the user's posture changes during training and generate high-precision posture data. Through this data, the user's posture state can be understood in real time, analyzed and processed. In order to ensure the accuracy and consistency of the data, the data collected by multiple visual sensors are processed and integrated through the data processing layer in the posture capture network. The function of the data processing layer is to synchronize, filter and fuse the raw data of multiple sensors, remove redundant information and generate a coherent data stream. This process ensures that the posture information collected from different joints can be integrated into a complete posture time series, reflecting the user's posture changes at different time points.

[0037] After processing, the first real-time posture time series is generated. The first real-time posture time series refers to a sequence of user posture changes recorded continuously in the time dimension, which can reflect the user's posture evolution throughout the training process. For example, when the user is doing back straightening training, the first real-time posture time series records the entire process from the user's bent back at the beginning to the gradual straightening of the back. This time series data can help the system analyze whether the user's posture adjustment conforms to the preset correction plan, and identify deviations or anomalies in the user's posture. It provides accurate basic data for posture correction, ensuring that effective correction analysis and decision-making can be carried out based on the user's actual posture.

[0038] Further, step three of this application includes:

[0039] Generate a dynamic reciprocating posture with the preset correction starting point posture and the preset correction end point posture; send the dynamic reciprocating posture to the correction limit reminder; input the first real-time posture time sequence into the correction limit reminder, perform abnormal pattern recognition, and generate the posture abnormal pattern.

[0040] Specifically, dynamic reciprocating posture refers to the continuous posture change of the target user in the posture correction process, which gradually transitions from the starting posture to the end posture through repeated posture adjustments. The generation of this dynamic posture can help capture the user's complete adjustment process in training and ensure that each posture movement meets the correction goal. For example, in a back correction training, the starting posture may be "back bending" and the end posture is "back straightening". A series of continuous posture changes from bending to straightening will be generated as dynamic reciprocating postures to guide users in training. Next, the dynamic reciprocating posture is sent to the correction limit reminder, which is a module for real-time monitoring and analysis of user posture. The function of the correction limit reminder is to compare the user's current posture with the preset dynamic reciprocating posture to determine whether the user adjusts the posture according to the predetermined correction trajectory. Through this comparison, it is possible to identify whether the user has posture deviation or abnormal behavior during the correction process.

[0041] Then, the first real-time posture timing is input into the correction limit reminder, which can identify any abnormalities or deviations of the user during the posture adjustment process. For example, suppose the user is doing shoulder correction training, and the dynamic reciprocating posture is from "shoulder leaning forward" to "shoulder drooping naturally." If the user does not adjust according to the preset dynamic posture during the training process, but always maintains the wrong forward posture, this abnormality will be identified by comparing the first real-time posture timing and the dynamic reciprocating posture, and a posture abnormality pattern will be generated. At this time, a correction reminder can be issued to the target user to remind the user that he needs to readjust his posture to avoid poor training results or physical damage caused by incorrect posture.

[0042] Furthermore, the present application also includes the following steps:

[0043] The first real-time posture timing is input into the correction limit reminder, and the deviation is compared with the dynamic reciprocating posture to generate a posture deviation vector; based on the dynamic reciprocating posture, the posture deviation vector is judged for the same-direction deviation. If the deviation is in the same direction, the posture abnormal mode is the first abnormal mode; if the deviation is in the opposite direction, the posture abnormal mode is the second abnormal mode.

[0044] Specifically, the first real-time posture sequence is input into the correction limit reminder, compared with the dynamic reciprocating posture, the difference between the two is calculated, and a posture deviation vector is generated. The posture deviation vector represents the degree and direction of the difference between the user's current posture and the ideal posture. For example, if the user's shoulder fails to adjust according to the target posture during the correction training, the actual height of the shoulder may be higher than the preset height, and then a deviation vector pointing upward is generated, indicating that the user's posture is deviated. Based on the dynamic reciprocating posture, the generated posture deviation vector is analyzed to determine whether the direction of the deviation is consistent with the preset posture adjustment direction. This process is called the same-direction deviation judgment. If the deviation is detected to be consistent with the adjustment direction of the dynamic reciprocating posture, that is, although the user has a posture deviation, the overall adjustment direction is correct, then the deviation belongs to the same-direction deviation. In this case, the first abnormal mode is generated to prompt the user that although there is a slight deviation, the overall posture adjustment process is in the right direction. For example, in the user's shoulder correction training, although the shoulder is not completely lowered, it has begun to adjust to the correct position, which is identified as the first abnormal mode.

[0045] However, if the deviation direction is detected to be inconsistent with the dynamic reciprocating posture, for example, the user's posture adjustment direction is opposite to the target direction, this deviation is an anisotropic deviation. At this time, a second abnormal mode will be generated to prompt the user that there is an error in the direction of posture adjustment. For example, in waist correction training, if the user's waist bends in the opposite direction during the correction process, this anisotropic deviation is identified as the second abnormal mode, and the user is reminded to make adjustments to avoid further posture problems or injuries caused by improper posture correction. By analyzing the posture deviation vector, the user's problems in the posture correction process can be identified more accurately, ensuring that the posture adjustment is in the right direction, and providing targeted feedback and guidance based on different abnormal modes.

[0046] Further, step five of this application includes:

[0047] The electromyographic sensing data is input into a pre-trained tension recognition network to generate the muscle tension recognition result; a correction sustainability state test is performed based on the muscle tension recognition result; if the correction sustainability state test passes, a correction decision is made based on the target posture correction scheme and the first real-time posture timing to generate a first correction reminder decision; and the first correction reminder decision is sent to the terminal display device of the target user.

[0048] Specifically, during the posture correction process, the electromyographic sensor data is input into the pre-trained tension recognition network. The tension recognition network is a model based on deep learning training. It uses a large amount of training data to identify the muscle tension under different postures and finally generates a muscle tension recognition result, which is the specific value or evaluation result of the tension or relaxation state of the user's current muscle group.

[0049] Next, based on the generated muscle tension recognition results, a correction sustainability test is performed. The purpose of the correction sustainability test is to evaluate whether the user can continue the current posture correction training within a reasonable range of muscle tension. If the user's muscle tension is within a reasonable range, it means that the user can continue posture correction without causing muscle fatigue or injury; on the contrary, if the muscle tension is too high or too low, it may mean that the user cannot continue to maintain the current correction state. For example, when the user is doing waist straightening training, if the waist muscle tension is identified to be moderate, it means that the user can effectively maintain this posture and the correction sustainability test is passed.

[0050] If the correction sustainable state test passes, the next correction decision will be made in combination with the target posture correction scheme and the first real-time posture sequence. The first real-time posture sequence refers to the data sequence that monitors the user's posture changes in real time. By comparing with the target posture correction scheme, it can be identified whether the current posture of the target user is adjusted on the right track, and the first correction reminder decision is generated. The first correction reminder decision refers to the real-time suggestions or instructions provided to the target user to help the user further optimize the posture during the training process. For example, if the user's posture is slightly deviated but overall meets the target scheme, the user may be advised to slightly adjust certain body parts or continue to maintain the current posture. Finally, the first correction reminder decision is sent to the user's terminal display device. The terminal display device can be the user's mobile phone, tablet computer or smart watch, and reminder information is sent to the user through these devices. For example, when the user's shoulder correction training is identified as slightly forward, the muscle tension is moderate and the training can continue, the user's smart watch may display "Please adjust the shoulder slightly backward and maintain the posture".

[0051] Through this complete process, accurate and personalized posture correction suggestions are provided to ensure that users can complete posture correction training efficiently under the premise of safety.

[0052] Furthermore, the present application also includes the following steps:

[0053] If the correction sustainable state test fails, a second abnormal warning signal is generated; wherein, the second abnormal warning signal is used to remind that the target posture correction plan does not match the target user; and the second abnormal warning signal is sent to the terminal display device of the target user.

[0054] Specifically, the purpose of the correction sustainable state test is to evaluate whether the user can continue the current posture correction training within a reasonable range of muscle tension. When the test fails, it means that the user's muscle tension exceeds the reasonable range, or other problems occur in the posture adjustment process, indicating that the user cannot continue the correction training safely or effectively. The generated second abnormal warning signal is used to remind the user that the current target posture correction scheme does not match the user's actual situation. In other words, the user may not be able to train according to the preset posture correction scheme due to physical strength, muscle state, posture amplitude and other reasons. For example, in back correction training, if the user's back muscle tension is too high, it may be because the intensity required for correcting the posture exceeds the user's physical tolerance. The user is reminded through an abnormal warning signal that the correction scheme may be inappropriate for the user or needs to be adjusted.

[0055] After the second abnormal warning signal is generated, the signal is sent to the user's terminal display device. For example, when the user is doing shoulder correction training and it is detected that the user's shoulder muscles are overly tense and cannot continue to train, a prompt message "Shoulder muscles are overly tense, please pause training and adjust the correction plan" will be displayed on the user's smart watch. Through this process, the user can be effectively prevented from potential physical injuries caused by the mismatch between the correction training and their own state, and the user is reminded that the current correction plan may need to be modified.

[0056] Furthermore, the present application also includes the following steps:

[0057] If the posture abnormality mode is the second abnormality mode, a second abnormal warning signal is generated; a posture correction decision is made based on the abnormal warning signal, and a second correction reminder decision is generated; the second abnormal warning signal and the second correction reminder decision are sent to the terminal display device of the target user.

[0058] Specifically, when the posture abnormality mode is detected as the second abnormality mode, a second abnormal warning signal is immediately generated. The second abnormal mode means that the posture adjustment direction of the target user is opposite to the preset target direction, or the posture deviation is too large, which may have an adverse effect on the user's body. The role of the second abnormal warning signal is to promptly remind the user that there is a major deviation in the current posture correction and corrective measures need to be taken immediately. For example, when the user is training to straighten the waist, if it is detected that the user's waist is not only not straightened, but bent in the opposite direction, a second abnormal warning signal will be generated.

[0059] Subsequently, a posture correction decision analysis is performed based on the generated second abnormal warning signal to generate a second correction reminder decision. The second correction reminder decision is a specific correction suggestion or instruction given based on the current posture state and deviation of the target user to help the target user correct the wrong posture. For the second correction reminder decision, depending on the severity of the abnormality, the user may be advised to stop the current training immediately and readjust the posture, or provide specific correction steps to help the user adjust in the right direction. For example, if the user leans forward excessively during shoulder correction training, the second correction reminder decision may suggest that the user pull his shoulders back and remind the user to pay more attention to the balance of the body. Specifically, a correction decision library can be established by professional and technical personnel in this field for possible abnormalities, and the second correction reminder decision can be matched by calling the correction decision library.

[0060] Finally, the second abnormal warning signal and the second correction reminder decision are sent to the terminal display device of the target user. The terminal display device can be a mobile phone, tablet, wearable device, etc., which can present the reminder information to the user in the form of vision or sound. For example, when the user's waist posture correction has a serious deviation, the user's smart watch may vibrate and display a prompt on the screen saying "Please adjust your waist posture immediately to avoid bending backward too much."

[0061] Through this process, major deviations in the user's posture can be monitored in a timely manner, and specific correction plans can be provided to help users quickly correct incorrect postures and ensure the safety and effectiveness of posture correction training.

[0062] In summary, the human body posture auxiliary correction method based on deep learning provided by this application has the following technical effects:

[0063] Receive the target posture correction scheme of the target user, wherein the target posture correction scheme includes a preset correction starting posture and a preset correction end posture; when the target user performs posture correction training, connect the posture capture network to monitor the posture of the target user and generate a first real-time posture sequence; based on the preset correction starting posture and the preset correction end posture, call the correction limit reminder to identify the correction abnormality mode of the first real-time posture sequence and generate a posture abnormality mode; if the posture abnormality mode is the first abnormal mode, connect the surface electromyography sensor to receive electromyography sensing data; perform muscle tension recognition based on the electromyography sensing data, and perform posture correction abnormality warning according to the muscle tension recognition result. Through the posture capture network and electromyography sensing technology, the user's posture changes and muscle tension state are monitored in real time, and abnormality recognition and warning are performed based on the deep learning algorithm, so as to provide the user with more accurate posture correction feedback, and achieve the technical effect of ensuring the safety and effectiveness of posture correction training.

[0064] Embodiment 2, based on the human body posture auxiliary correction method based on deep learning in the aforementioned embodiment 1, with the same inventive concept, this application also provides a human body posture auxiliary correction system based on deep learning, please refer to the attached Figure 2 , the human posture auxiliary correction system based on deep learning includes:

[0065] The correction scheme receiving module 11 is used to receive a target posture correction scheme of a target user, wherein the target posture correction scheme includes a preset correction starting point posture and a preset correction end point posture.

[0066] The posture monitoring module 12 is used to connect to the posture capture network to perform posture monitoring on the target user when the target user performs posture correction training, and generate a first real-time posture time series.

[0067] The correction abnormality pattern recognition module 13 is used to call the correction limit reminder to perform correction abnormality pattern recognition on the first real-time posture sequence based on the preset correction starting point posture and the preset correction end point posture, and generate a posture abnormality pattern.

[0068] The surface electromyography sensing module 14 is used to connect to the surface electromyography sensor and receive electromyography sensing data if the abnormal posture mode is the first abnormal mode.

[0069] The muscle tension recognition module 15 is used to recognize muscle tension based on the electromyographic sensing data, and to issue an abnormal posture correction warning according to the muscle tension recognition result.

[0070] Furthermore, the correction abnormal pattern recognition module 13 in the human posture auxiliary correction system based on deep learning is also used for:

[0071] Generate a dynamic reciprocating posture with the preset correction starting point posture and the preset correction end point posture; send the dynamic reciprocating posture to the correction limit reminder; input the first real-time posture time sequence into the correction limit reminder, perform abnormal pattern recognition, and generate the posture abnormal pattern.

[0072] Furthermore, the correction abnormal pattern recognition module 13 in the human posture auxiliary correction system based on deep learning is also used for:

[0073] The first real-time posture timing is input into the correction limit reminder, and the deviation is compared with the dynamic reciprocating posture to generate a posture deviation vector; based on the dynamic reciprocating posture, the posture deviation vector is judged for the same-direction deviation. If the deviation is in the same direction, the posture abnormal mode is the first abnormal mode; if the deviation is in the opposite direction, the posture abnormal mode is the second abnormal mode.

[0074] Furthermore, the muscle tension recognition module 15 in the human posture auxiliary correction system based on deep learning is also used for:

[0075] The electromyographic sensing data is input into a pre-trained tension recognition network to generate the muscle tension recognition result; a correction sustainability state test is performed based on the muscle tension recognition result; if the correction sustainability state test passes, a correction decision is made based on the target posture correction scheme and the first real-time posture timing to generate a first correction reminder decision; and the first correction reminder decision is sent to the terminal display device of the target user.

[0076] Furthermore, the muscle tension recognition module 15 in the human posture auxiliary correction system based on deep learning is also used for:

[0077] If the correction sustainable state test fails, a second abnormal warning signal is generated; wherein, the second abnormal warning signal is used to remind that the target posture correction plan does not match the target user; and the second abnormal warning signal is sent to the terminal display device of the target user.

[0078] Further, the posture monitoring module 12 in the human posture auxiliary correction system based on deep learning is also used for:

[0079] The posture capture network includes multiple visual sensors, which are arranged at local training joints of the target user; the data collected by the multiple visual sensors are processed by the data processing layer in the posture capture network to generate the first real-time posture time series; wherein the first real-time posture time series includes the continuously changing posture of the target user in the time dimension.

[0080] Furthermore, the human posture auxiliary correction system based on deep learning also includes an abnormal mode judgment module, and the abnormal mode judgment module is used to:

[0081] If the posture abnormality mode is the second abnormality mode, a second abnormal warning signal is generated; a posture correction decision is made based on the abnormal warning signal, and a second correction reminder decision is generated; the second abnormal warning signal and the second correction reminder decision are sent to the terminal display device of the target user.

[0082] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. Figure 1The human posture auxiliary correction method based on deep learning and the specific examples in Example 1 are also applicable to the human posture auxiliary correction system based on deep learning in this embodiment. Through the above detailed description of the human posture auxiliary correction method based on deep learning, those skilled in the art can clearly know the human posture auxiliary correction system based on deep learning in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0083] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0084] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A human posture auxiliary correction method based on deep learning, characterized in that: include: Receiving a target posture correction scheme of a target user, wherein the target posture correction scheme includes a preset correction starting posture and a preset correction end posture; When the target user performs posture correction training, a posture capture network is connected to monitor the posture of the target user to generate a first real-time posture time series; Based on the preset correction starting point posture and the preset correction end point posture, a correction limit reminder is called to perform correction abnormality pattern recognition on the first real-time posture time sequence to generate a posture abnormality pattern; If the abnormal posture mode is the first abnormal mode, connecting a surface electromyography sensor to receive electromyography sensing data; Perform muscle tension recognition based on the electromyographic sensor data, and issue a posture correction abnormality warning according to the muscle tension recognition result; Performing muscle tension recognition based on the electromyographic sensor data and providing posture correction abnormality warning according to the muscle tension recognition result includes: Inputting the electromyographic sensing data into a pre-trained tension recognition network to generate the muscle tension recognition result; Performing correction sustainable state test according to the muscle tension identification result; If the correction sustainable state test passes, a correction decision is made based on the target posture correction scheme and the first real-time posture time sequence to generate a first correction reminder decision; The first correction reminder decision is sent to a terminal display device of the target user.

2. The method for assisting human posture correction based on deep learning according to claim 1, characterized in that: Based on the preset correction starting point posture and the preset correction end point posture, calling the correction limit reminder to perform correction abnormality pattern recognition on the first real-time posture time sequence to generate a posture abnormality pattern, including: Generating a dynamic reciprocating posture with the preset correction starting point posture and the preset correction end point posture; sending the dynamic reciprocating posture to the correction limit reminder; The first real-time posture time sequence is input into the correction limit reminder to perform correction abnormality pattern recognition to generate the posture abnormality pattern.

3. The method for assisting human posture correction based on deep learning as claimed in claim 2, characterized in that: Inputting the first real-time posture time sequence into the correction limit reminder, performing correction abnormality pattern recognition, and generating the posture abnormality pattern, comprises: Inputting the first real-time posture time sequence into the correction limit reminder, performing deviation comparison with the dynamic reciprocating posture, and generating a posture deviation vector; Taking the dynamic reciprocating posture as a reference, the posture deviation vector is judged to have a same-direction deviation. If the deviation is in the same direction, the posture abnormal mode is the first abnormal mode. If the deviation is in an anisotropic direction, the posture abnormal mode is the second abnormal mode.

4. The method for assisting human posture correction based on deep learning according to claim 1, characterized in that: Also includes: If the correction sustainable state test fails, generating a second abnormal warning signal; Wherein, the second abnormal warning signal is used to remind that the target posture correction scheme does not match the target user; The second abnormal warning signal is sent to the terminal display device of the target user.

5. The method for assisting human posture correction based on deep learning according to claim 1, characterized in that: When the target user performs posture correction training, a posture capture network is connected to monitor the posture of the target user to generate a first real-time posture time series, including: The posture capture network includes a plurality of visual sensors, and the plurality of visual sensors are arranged at local training joints of the target user; Processing the data collected by the multiple visual sensors through a data processing layer in the gesture capture network to generate the first real-time gesture time series; The first real-time posture sequence includes the continuously changing posture of the target user in the time dimension.

6. The method for assisting human posture correction based on deep learning as claimed in claim 3, characterized in that: Also includes: If the posture abnormality mode is the second abnormality mode, generating a second abnormality warning signal; Making a posture correction decision based on the second abnormal warning signal and generating a second correction reminder decision; The second abnormal warning signal and the second correction reminder decision are sent to the terminal display device of the target user.

7. A human posture auxiliary correction system based on deep learning, characterized in that: The steps for implementing the deep learning-based human posture auxiliary correction method according to any one of claims 1 to 6 include: A correction scheme receiving module, used to receive a target posture correction scheme of a target user, wherein the target posture correction scheme includes a preset correction starting point posture and a preset correction end point posture; A posture monitoring module, used for connecting to a posture capture network to monitor the posture of the target user when the target user is performing posture correction training, and generating a first real-time posture time series; A correction abnormality pattern recognition module, used to call a correction limit reminder to perform correction abnormality pattern recognition on the first real-time posture sequence based on the preset correction starting point posture and the preset correction end point posture, and generate a posture abnormality pattern; A surface electromyography sensing module, used for connecting to a surface electromyography sensor and receiving electromyography sensing data if the abnormal posture mode is the first abnormal mode; The muscle tension recognition module is used to recognize muscle tension based on the electromyographic sensing data and to provide posture correction abnormality warning according to the muscle tension recognition result.

Images