A human body balance posture correction training system and method

Abstract

The present invention proposes a human body balance posture correction training system and method, which includes a data acquisition module, a posture prediction and evaluation training module and a VR virtual module. The data acquisition module obtains the joint data and posture information of the target object and transmits them to the host computer; a reinforcement learning training model is used to predict the posture that the target object should exhibit; the actual posture made by the target object according to the moving direction and stride prompt information is recorded; the overlap between the actual posture and the predicted posture is compared; finally, judgment and evaluation are made and the feedback information is presented in a VR virtual reality environment, so that the target object can realize timely posture correction in an immersive experience, effectively improving the effect and quality of correction training.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a human body balance posture correction training system and method. Background Art

[0002] As aircraft technology continues to evolve, training to maintain the pilot's balanced posture during operation has become increasingly crucial. This encompasses pilots, pilots, and operators, all of whom must ensure proper posture during flight or operation to ensure safe and efficient operations. Therefore, research and development of training devices for posture correction to provide safer and more efficient balance training has become a core area of ​​research and development.

[0003] The continuous development of liquid level information acquisition technology, virtual reality (VR) technology, and reinforcement learning has brought new prospects to posture correction training devices. For example, using liquid level information acquisition equipment can more accurately assess the center of gravity of a target object. VR technology provides an immersive experience that makes the target object feel as if they are actually there, while reinforcement learning provides an effective tool for assessing the driver's posture state. The continuous improvement of these technologies has driven technological innovation in the field of posture correction devices, making them more reliable and practical.

[0004] Several current research projects are exploring how to integrate virtual reality technology with real-world scenarios to create a more immersive pilot posture training environment. This innovative approach aims to provide a more realistic and engaging training experience, capable of simulating a variety of flight or driving scenarios. In this approach, virtual reality technology is used to simulate real-world flight or driving scenarios during training. By donning a virtual reality headset, the pilot can virtually place themselves at the controls of an aircraft or vehicle and interact with the environment. This interaction might include handling emergencies, coping with varying weather conditions, and performing complex maneuvers. Furthermore, this approach can be combined with real-world scenarios, for example, by integrating virtual reality simulations into real-world flight simulators or embedding virtual elements into actual driving training vehicles. This integration can provide a more comprehensive training experience, enabling pilots to better maintain correct posture and reactions in the face of various challenges. Overall, this technological advancement has the potential to provide a more realistic simulation environment for pilot posture correction training, helping to improve training effectiveness, increase the realism of training, and better prepare pilots for a variety of complex driving scenarios. Some research is focused on developing intelligent astronaut posture correction training devices that can automatically adjust training plans based on the pilot's actual needs. This involves using intensive techniques to personalize training and assessments to suit the unique characteristics and needs of each astronaut.

[0005] Posture correction training devices are not only used in astronaut training but also in a wide range of other fields, such as medicine, the military, and industry. In the medical field, these devices can be used for rehabilitation therapy, helping patients regain correct body posture and position after recovery. In the military, they can be used to train soldiers to ensure they can maintain proper posture and balance on the battlefield. In the industrial field, these devices can be used to train workers to improve work efficiency and reduce the risk of work-related injuries. The demand for posture control in these various fields is constantly increasing, as correct posture and position are crucial to the successful execution of missions and the health of employees. Therefore, the application range of posture correction training devices is constantly expanding to meet the needs of various different fields. This has also promoted the continuous innovation and improvement of related technologies and devices to meet the requirements of different fields. Summary of the Invention

[0006] The purpose of the present invention is to provide a training system that helps a target group to adjust their posture more skillfully and efficiently by providing a virtual real scene.

[0007] To achieve the above objectives, the present invention proposes a human body balance posture correction training system, comprising the following modules:

[0008] Data acquisition module: used to collect information about the physical scene and obtain the center of gravity, joints and posture information of the target object in real time;

[0009] Posture prediction and evaluation training module: used to predict the posture adjustment required to maintain balance in real time and compare it with the actual posture adjustment of the target object, thereby evaluating the posture adjustment quality of the target object in real time;

[0010] VR virtual module: used to build virtual reality scenes and present real-time feedback information on target object posture correction.

[0011] Furthermore, the data acquisition module includes the following collectors:

[0012] Liquid level information collector: used to obtain the center of gravity information of the target object in real time;

[0013] Visual sensor: including space camera and depth camera: Space camera: used to capture the pose information of the target object; Depth camera: used to capture the joint information of the target object;

[0014] Furthermore, the posture prediction and evaluation training module includes a reinforcement learning training module and a balance system;

[0015] The reinforcement learning training module includes the following algorithms:

[0016] Multi-data fusion algorithm: Based on the joint information captured by the depth camera, a 6D pose estimation method is used to fit the posture state of the training object, which is used to analyze the pose information of the target object for evaluation;

[0017] Posture adjustment evaluation algorithm: used to evaluate the effectiveness of the posture adjustment made by the target object to maintain balance;

[0018] Human posture tracking algorithm: used to realize the tracking and positioning tasks of human posture;

[0019] Center of gravity fitting algorithm and center of gravity analysis algorithm: relying on the liquid level information obtained by the liquid level information collector, calculate the torque within the center of gravity error range, and provide feedback to the target object on the direction and stride that should be adjusted based on the torque;

[0020] The balancing system calculates the consistency of the center of gravity based on the comparison between the liquid level information data and the reference center of gravity position, and provides feedback on the direction and stride of the posture adjustment of the target object.

[0021] Furthermore, the balancing device constituting the balancing system includes a balancing board and a chassis driving device:

[0022] The balance board is placed on the upper end of the chassis driving device through a center frame;

[0023] The mechanical structure of the chassis drive device includes six support rods with hydraulic devices, an upper platform and a lower platform. The lower platform is fixed to the infrastructure. One end of the support rod is distributed around the lower platform and is movably connected to the lower platform through a universal joint. The other end is rotationally connected to the upper platform through a spherical pair. The telescopic movement of the support rod enables the upper platform to move and rotate freely in all directions.

[0024] The chassis drive device is connected to the reinforcement learning training model signal, and cooperates with the liquid level information collector installed at the connection between the upper platform and the center frame to drive the balancing system to realize active mode and passive mode;

[0025] In the active mode, the balancing device will follow the tilting of the human body, and the host computer will provide prompt information in the virtual scene by analyzing the center of gravity information. The target object will adjust its posture in time to maintain balance according to the prompt information. In the passive mode, the balancing device is locked, and the host computer will control the chassis drive device to tilt in different directions by analyzing the center of gravity information. The target object needs to maintain balance in time.

[0026] Furthermore, the VR virtual module implements the pose fitting of the target object in the UI interface through the multi-data fusion algorithm.

[0027] The present invention also provides a method for training a human body to correct a balance posture, using the above-mentioned human body correction training system. The specific method is as follows:

[0028] Step 1: Obtain joint data of the target object through the depth camera;

[0029] Step 2: Train the model through reinforcement learning to predict the posture that the target object should exhibit;

[0030] Step 3: Recording the actual posture of the target object according to the moving direction and stride prompt information;

[0031] Step 4: Compare the degree of coincidence between the actual posture and the predicted posture;

[0032] Step 5: Determine whether the overlap is greater than a threshold, and present the feedback information in the VR virtual environment. If so, give a posture adjustment suggestion; if not, mark it as the correct posture;

[0033] Furthermore, the generation mechanism of the moving direction and stride prompt information in step 3 is:

[0034] Step 3.1: Obtaining the center of gravity data of the target object in real time through a liquid level information collector;

[0035] Step 3.2: Calculate the magnitude and direction of the moment of the target object's center of gravity relative to the center of the chassis using the balancing system;

[0036] Step 3.3: Using a perpendicular line from the center of the chassis on the chassis drive device to the direction of the torque as a reference line, calculate the direction and distance that the center of gravity of the target object deviates from the reference line;

[0037] Step 3.4: Determine whether the offset distance is greater than a threshold. If so, provide feedback information on the stride and direction of movement to the target object.

[0038] Step 3.5: The feedback information is optimized through PID adjustment and presented in a VR virtual environment.

[0039] Furthermore, the evaluation mechanism for the actual moving stride of the target object is as follows:

[0040] S1: Acquire the center of gravity data of the target object in real time through a liquid level information collector;

[0041] S2: calculating, by the balance system, the stride length required for the target object to reach the target center of gravity;

[0042] S3: Determine whether the actual movement stride deviation is greater than a threshold; if so, perform feedforward adjustment through the visual sensor and return to S1; if not, optimize the actual movement stride through the PID control algorithm;

[0043] S5: When the actual moving stride is not greater than the threshold, determine whether the position deviation between the current center of gravity and the target center of gravity is less than the threshold; if not, return to step S1.

[0044] Furthermore, during the posture correction training process, the host computer records each training data of the target object for reference in the next training plan, thereby achieving personalized training for the target object.

[0045] Compared with the prior art, the advantages of the present invention are:

[0046] 1. The present invention uses a VR virtual module in conjunction with two modes of the balance system to simulate real training scenarios and provide virtual prompts in the scenarios as guidance, allowing trainers to obtain an immersive training experience and feel and operate in a virtual environment similar to actual work scenarios. This experience helps trainers to more deeply understand and master the operating skills and processes in actual work, improve their adaptability and proficiency in actual work, and better cope with various challenges and tasks.

[0047] 2. The present invention uses a data acquisition module and a posture prediction and evaluation training module to obtain the changes in the target object's human posture in real time, and through prediction and evaluation, promptly gives feedback and suggestions, effectively improving the posture correction effect and quality of the target object, without the need for long waiting times, which greatly saves time.

[0048] 3. The present invention records the training data of the target object each time and uses this data as a reference for adjusting various parameters in the next training plan, thereby formulating a specific training plan for each target object and improving training efficiency and quality.

[0049] 4. The present invention uses various algorithms in the posture prediction and evaluation training module to achieve higher posture detection accuracy and more accurate data, thereby more accurately identifying and correcting the target object's bad posture.

[0050] 5. The mechanical structure of the chassis drive device in the present invention has a parallel structure, that is, six drives act together on one platform, thereby optimizing the system stiffness, increasing the load-bearing capacity, and preventing the accumulation of position errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic diagram illustrating the design of the human body balance posture correction training system of this embodiment.

[0052] Figure 2 Schematic diagram of the human body balance posture correction training system architecture of this embodiment.

[0053] Figure 3Schematic diagram of the placement of various devices in the human body balance posture correction training system of this embodiment.

[0054] Figure 4 Schematic diagram of the planar structure of the chassis drive device in this embodiment.

[0055] Figure 5 This is a schematic diagram of the three-dimensional structure of the chassis drive device.

[0056] Figure 6 This is a schematic diagram of the rotation direction of the chassis drive device in this embodiment.

[0057] Figure 7 This is a flow chart of the human body balance posture correction training method in this embodiment.

[0058] Figure 8 Flowchart of the generation mechanism of the moving direction and stride prompt information in this embodiment.

[0059] Figure 9 Flowchart of the evaluation mechanism of the actual moving stride of the target object in this embodiment. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be further described below.

[0061] This embodiment proposes a human body balance posture correction training system, the design of which mainly focuses on five parts: mechanical structure design, data acquisition, VR virtual scene display, posture prediction and evaluation, and operating environment. Figure 1 As shown, the human body balance posture correction training system of this embodiment will be specifically described below.

[0062] This embodiment proposes a human body balance posture correction training system, such as Figure 2 As shown, it includes a data acquisition module, a posture prediction and evaluation training module and a VR virtual module, among which the data acquisition module is used to collect information about the physical scene and obtain the center of gravity, joints and posture information of the target object in real time; the posture prediction and evaluation training module is used to predict the posture adjustment required to maintain balance in real time, and compare it with the actual posture adjustment of the target object, so as to evaluate the posture adjustment quality of the target object in real time; VR virtual module: used to construct a virtual reality scene and present the target object posture correction feedback information in real time.

[0063] In the data acquisition module, Figure 4 The liquid level information collector 6 shown in the figure obtains the center of gravity information of the target object in real time; Figure 3 The space camera 2 shown captures the position information of the target object; Figure 3The depth camera 1 shown captures joint information of the target object.

[0064] In this embodiment, the posture prediction and evaluation training module includes a reinforcement learning training module and a balance system. In the reinforcement learning training module, based on the joint information captured by the depth camera 1, a multi-data fusion algorithm is used to fit the posture state of the training object using a 6D posture estimation method, thereby analyzing the posture information of the target object for evaluation; a human posture tracking algorithm is used to achieve the tracking and positioning tasks of the human posture; a posture adjustment evaluation algorithm is used to evaluate the effectiveness of the posture adjustment made by the target object to maintain balance; a center of gravity fitting algorithm and a center of gravity analysis algorithm are used to calculate the torque within the center of gravity error range based on the liquid level information obtained by the liquid level information collector 6, and based on this torque, feedback is provided to the target object regarding the direction and stride that should be adjusted. The balance system calculates the consistency of the center of gravity based on the comparison of the liquid level information data and the reference center of gravity position, and provides feedback on the direction and stride of the target object's posture adjustment.

[0065] The balancing device constituting the balancing system, such as Figure 3 As shown, it includes a balance board 3 and a chassis driving device 4. The balance board 3 is placed on the upper end of the chassis driving device 4. The chassis driving device 4 has a structure as shown in FIG. Figure 4 and Figure 5 As shown, it includes an upper platform 11, six support rods 8 with hydraulic devices and a lower platform 9: the lower platform 9 is fixed on the infrastructure, one end of the support rod 8 is distributed on the circumference of the lower platform 9, and is movably connected to the lower platform 9 through a universal ball, and the other end is rotationally connected to the upper platform 11 through a spherical pair 7. A center frame 5 is provided on the upper end of the upper platform 11, and the balance plate 3 is placed on the upper end of the chassis drive device 4 through the center frame 5. The above-mentioned liquid level information collector 6 is provided at the connection between the upper platform 11 and the center frame 5. The balancing system drives the control of six degrees of freedom of the upper platform 11 in three-dimensional space through the telescopic movement of the six support rods 8. These six degrees of freedom include translation in the XYZ axis direction and rotation around the XYZ axis, as shown in FIG. Figure 6As shown, the upper platform 11 is free to move and rotate in all directions, providing diversified motion control. The mechanical structure of the chassis drive device 4 has a parallel structure, that is, six drivers act together on one platform, thereby optimizing the system stiffness, strong load-bearing capacity, and no cumulative position error. In addition, the chassis drive device 4 is connected to the above-mentioned reinforcement learning training model signal to drive the balance system structure to realize active mode and passive mode. In the active mode, the balance device will follow the human body's tipping, and the host computer will analyze the center of gravity information based on the liquid level information collector 6, and provide prompt information in the virtual scene. The target object will adjust its posture in time to maintain balance according to the prompt information. In the passive mode, the balance device will be locked, and the host computer will control the chassis system's device to tip in different directions by analyzing the center of gravity information. The target object needs to maintain balance in time. In addition, the chassis drive device 4 is also configured with a corresponding housing 10.

[0066] In this embodiment, the various devices constituting the human body balance posture correction training system are arranged as shown in FIG3 , and the depth camera 1 and the space camera 2 are symmetrically distributed around the chassis drive device 4 .

[0067] In this embodiment, the VR virtual module uses a multi-data fusion algorithm to achieve the posture fitting of the target object in the UI interface, and at the same time uses the host computer to present the feedback information of the posture evaluation and correction in the constructed virtual reality (VR) scene.

[0068] By using the human body balance posture correction training system of the above embodiment, this embodiment also proposes a human body balance posture correction training method, such as Figure 7 As shown, the specific steps are:

[0069] Step 1: Obtain joint data of the target object through depth camera 1;

[0070] Step 2: Train the model through reinforcement learning to predict the posture that the target object should exhibit;

[0071] Step 3: Record the actual posture of the target object according to the moving direction and stride prompt information;

[0072] Step 4: Compare the degree of coincidence between the actual pose and the predicted pose;

[0073] Step 5: Determine whether the overlap in step 4 is greater than the threshold, and present the feedback information in the VR virtual environment. If so, give posture adjustment suggestions; if not, mark it as the correct posture.

[0074] Among them, the generation mechanism of the moving direction and stride prompt information in the above step 3 is as follows: Figure 8 As shown, specifically:

[0075] Step 3.1: Obtain the target object's center of gravity data in real time through the liquid level information collector 6;

[0076] Step 3.2: Calculate the magnitude and direction of the moment about the center of gravity of the target object relative to the center of the chassis using the balancing system.

[0077] Step 3.3: Using the perpendicular line from the center of the chassis on the chassis drive device 4 to the direction of the torque as the reference line, calculate the direction and distance that the center of gravity of the target object deviates from the reference line;

[0078] Step 3.4: Determine whether the offset distance is greater than a threshold. If so, provide the target object with feedback on the stride and direction of movement, and present it in the VR scene. If not, assume that the target object does not need to move.

[0079] In this embodiment, the evaluation mechanism for the actual moving stride of the target object is as follows: Figure 9 As shown, specifically:

[0080] S1: Obtaining the center of gravity data of the target object in real time through the liquid level information collector 6;

[0081] S2: Calculate the stride length required for the target object to reach the target center of gravity through the balance system;

[0082] S3: Determine whether the actual movement stride deviation is greater than the threshold; if so, perform feedforward adjustment through the visual sensor and return to S1; if not, optimize the actual movement stride through the PID control algorithm;

[0083] S5: When the actual moving stride is not greater than the threshold, determine whether the position deviation between the current center of gravity and the target center of gravity is less than the threshold; if not, return to step S1.

[0084] In this embodiment, during the posture correction training process, the host computer records and saves all training data of the target object in real time for reference in the next training plan, thereby meeting the personalized training needs of the target object.

[0085] The above description is merely a preferred embodiment of the present invention and does not limit the present invention in any way. Any person skilled in the art who, without departing from the scope of the present invention, makes any equivalent substitution, modification, or other changes to the technical solution and technical content disclosed in the present invention shall be deemed to be within the scope of the present invention and still fall within the scope of protection of the present invention.

Claims

1. A human body balance posture correction training system, characterized in that: Includes the following modules: Data acquisition module: used to collect information about the physical scene and obtain the center of gravity, joints and posture information of the target object in real time; Posture prediction and evaluation training module: used to predict the posture adjustment required to maintain balance in real time and compare it with the actual posture adjustment of the target object, thereby evaluating the posture adjustment quality of the target object in real time; The posture prediction and evaluation training module includes a reinforcement learning training module and a balance system; The reinforcement learning training module includes the following algorithms: Multi-data fusion algorithm: Based on the joint information captured by the depth camera, a 6D pose estimation method is used to fit the posture state of the training object, which is used to analyze the pose information of the target object for evaluation; Posture adjustment evaluation algorithm: used to evaluate the effectiveness of the posture adjustment made by the target object to maintain balance; Human posture tracking algorithm: used to realize the tracking and positioning tasks of human posture; Center of gravity fitting algorithm and center of gravity analysis algorithm: rely on the liquid level information obtained by the liquid level information collector to calculate the torque within the center of gravity error range. Based on the torque, feedback is provided to the target object on the direction and stride that should be adjusted; The balancing system calculates the consistency of the center of gravity based on the comparison of the liquid level information data and the reference center of gravity position, and provides feedback on the direction and stride of the target object posture adjustment; The balancing device constituting the balancing system includes a balancing board and a chassis driving device: the balancing board is placed on the upper end of the chassis driving device; The mechanical structure of the chassis drive device includes six support rods with hydraulic devices, an upper platform and a lower platform: the lower platform is fixed to the infrastructure, one end of the support rods is distributed around the lower platform and is movably connected to the lower platform, and the other end is rotatably connected to the upper platform. Through the telescopic movement of the support rods, the upper platform can be freely moved and rotated in all directions; The chassis drive device is connected to the reinforcement learning training model signal, and cooperates with the liquid level information collector to drive the balancing system to realize active mode and passive mode; In the active mode, the balancing device structure will tilt with the human body, and the host computer will provide prompt information in the virtual scene by analyzing the center of gravity information. The target object will adjust its posture in time according to the prompt information to maintain balance. In the passive mode, the balancing device structure is locked, and the host computer will control the chassis drive device to tilt in different directions by analyzing the center of gravity information. The target object needs to maintain balance in time. VR virtual module: used to build virtual reality scenes and present real-time feedback information on target object posture correction.

2. The human body balance posture correction training system according to claim 1, characterized in that: The data acquisition module includes the following collectors: Liquid level information collector: used to obtain the center of gravity information of the target object in real time; The visual sensor includes a spatial camera and a depth camera. The spatial camera is used to capture the position information of the target object; the depth camera is used to capture the joint information of the target object.

3. The human body balance posture correction training system according to claim 1, characterized in that: The VR virtual module realizes the posture fitting display of the target object in the UI interface through the multi-data fusion algorithm.

4. A method for training a human body balance posture correction, using a human body balance posture correction training system according to any one of claims 1 to 3, characterized in that: The posture correction training method is as follows: Step 1: Obtain joint data of the target object through the depth camera; Step 2: Train the model through reinforcement learning to predict the posture that the target object should exhibit; Step 3: Recording the actual posture of the target object according to the moving direction and stride prompt information; The generation mechanism of the moving direction and stride prompt information is: Step 3.1: Obtaining the center of gravity data of the target object in real time through a liquid level information collector; Step 3.2: Calculate the magnitude and direction of the moment of the target object's center of gravity relative to the center of the chassis using the balancing system; Step 3.3: Using a perpendicular line from the center of the chassis on the chassis drive device to the direction of the torque as a reference line, calculate the direction and distance that the center of gravity of the target object deviates from the reference line; Step 3.4: Determine whether the offset distance is greater than a threshold. If so, provide feedback information on the stride and direction of movement to the target object. Step 3.5: Optimizing the feedback information through PID adjustment and presenting it in a VR virtual environment; Step 4: Compare the degree of coincidence between the actual posture and the predicted posture; Step 5: Determine whether the overlap is greater than a threshold, and present the feedback information in the VR virtual environment. If so, give a posture adjustment suggestion; if not, mark it as the correct posture.

5. The method for correcting the balance posture of the human body according to claim 4, characterized in that: The evaluation mechanism for the actual moving stride of the target object is as follows: S1: Acquire the center of gravity data of the target object in real time through a liquid level information collector; S2: calculating, by the balance system, the stride length required for the target object to reach the target center of gravity; S3: Determine whether the actual movement stride deviation is greater than a threshold; if so, perform feedforward adjustment through the visual sensor and return to S1; if not, optimize the actual movement stride through the PID control algorithm; S5: When the actual moving stride is not greater than the threshold, determine whether the position deviation between the current center of gravity and the target center of gravity is less than the threshold; if not, return to step S1.

6. The method for correcting the balance posture of the human body according to claim 4, characterized in that: During the posture correction training process, the host computer records each training data of the target object for reference in the next training plan, thereby realizing personalized training for the target object.

Images