TWI824921B - Training system and method for body movement - Google Patents

Abstract

A training system and method for body movement. The training system includes an open source data platform, a display device, an image capture device, and an image comparison unit. When the body movement images uploaded by multiple coaches on the open source data platform, students can visually learn to imitate the coach's body movement. Then the training system can capture the body movement images of the students and compare the body movement images of the students and the coach to generate image similarity or image difference for subsequent students to correct body movements. Among them, the open source data platform has a database and AI module, which allows the AI ​​module to conduct in-depth learning and continuously optimize the body movements of the film to further enhance the teaching effect.

Description

Limb movement training system and its training methods

The present invention relates to a training system and training method for various sports such as martial arts, dance, gymnastics, sports competitions, rehabilitation, etc., especially a training system and training method for limb movements.

Driven by the epidemic, the public is relying more on online platforms to learn various skills, including various body movements. Students can use the platform to learn based on text or videos, similar to the patent number TWI760044B proposed for dementia: " An intelligent cognitive motion training method, suitable for a cognitive motion training system including at least a display device, a camera and a host."

The above-mentioned patent only has a teaching video provided by a single coach for students to learn, and the training method has a certain fixed value. When the learning ability is higher than the first exercise, a second exercise that is more difficult than the first exercise will be provided for students to learn. Compared with It is too simple and a bit easy for the general public to achieve. And because it is not easy for multiple coaches to provide teaching videos, there is often only one learning area that a single coach can provide, so the scope of learning is slightly insufficient. For example, if modern people like fitness, yoga, dancing, martial arts, etc., there is no coach to support them.

Therefore, in order to be able to learn body movements in various fields and to be continuously modified and learned, it is necessary to develop more ideal technical methods to solve the above problems.

The purpose of the present invention is to provide a training system and training method for limb movements in various sports such as martial arts, dance, gymnastics, sports competitions, rehabilitation, etc., for training the limb movement ability of at least one student.

The invention relates to a limb movement training system. The training system includes an open source data platform, a display device, an image capture device, and an image comparison unit.

In the open source data platform, multiple coaches can upload any body movement images generated by consuming energy through the body's skeletal muscles. The display device is coupled to the open source data platform to display the coach's body movement images for students to visually imitate and learn from the coach. For body movements, the open source data platform can be similar to platforms set up in network systems such as Youtube or Nintendo, and the coach can be anyone, that is, anyone can upload images to the open source data platform.

The image comparison unit is coupled to the image capture device and the open source data platform. The image capture device can capture the student's body movement images, and then compares the student's body movement images with the coach's body movement images through the image comparison unit. To generate image similarity or image difference for subsequent students to correct body movements.

The image capture device in the training system can be a mobile phone, a camera with wireless communication function, a camera external to the screen, or a built-in photography lens. The image capture device has a transmission module, and the transmission module can be Wifi. , Bluetooth receiver, etc., the transmission module can be used to transmit the student's body movement images for subsequent comparison by the image comparison unit. The open source data platform is set up in a network system, which includes the Internet, cloud servers, multiple coach terminals, and multiple student terminals.

The cloud server can communicate with the Internet, and the open source data platform is built on the cloud server. Multiple coaching terminals are respectively connected to the Internet and can be used for multiple The coach uploads the file of his body movement image and transmits it to the open source data platform of the cloud server through the Internet. In addition, there are multiple student terminals connected to the Internet. The student terminals can be, for example, mobile phones and computers. The student terminals are used to download the coach's body movement images from the open source data platform of the cloud server for learning.

The student terminal has a display device, a receiving module, and a control application module. The receiving module can be a Wifi or Bluetooth receiver, etc., and the control application module can be a program component, such as an APP on a mobile phone or a plug-in program that can be installed on a desktop computer. The control application module has an image comparison unit. When the receiving module receives body movement images of students from the transmitting module, it can be transmitted to the image comparison unit of the control application module for comparison and analysis.

Furthermore, after the receiving module in the training system receives the student's body movement image from the transmitting module, it will be displayed on the display device and overlap with the coach's body movement image. For example, the left hand of the coach's image will overlap with the left hand of the student's image, or The head of the coach image will be superimposed on the head of the student image, and the student image can be presented with low saturation such as gray, while the coach image can be marked with a skeleton structure of obvious colors such as red, that is, the gray student image will overlap with the red coach influence. , and divide the human body into left and right sides or upper and lower sides, then the left and right halves or the upper and lower halves of the student's image will overlap with the instructor's image respectively.

In addition, the coach terminal in the training system further has a control setting module. The control setting module is used to set the coach's skeletal structure and predetermined comparison parts in the coach's body movement images. The predetermined comparison parts are, for example, left and right hands, Local parts such as the left and right feet, and an error threshold is set corresponding to the comparison part. The error threshold is used to provide image similarity or image difference comparison, and the error threshold can be set arbitrarily. For example, the error threshold for the left and right hands is 10%. The error threshold for the left and right feet is 7%.

When there are multiple comparison parts in the training system, the control setting module can set the weights of the multiple comparison parts, and the multiple weights are multiplied by the image similarity or image difference of the corresponding comparison parts. , then summed, and then compared with the error threshold.

Furthermore, the training system can include databases, screening modules, and AI modules. The database is used to store body movement images of multiple coaches as a coaching data set. The filtering module uses the error threshold results of image similarity or image difference comparisons (this result can be converted into scores) to filter out the results that exceed expectations. The body movement images of students with a target score (for example, 80 points) are used as a qualified student data set. That is, the screening module can form a score based on the result of the image similarity or image difference comparison error threshold, and screen out the students who exceed the expected target, such as 80 points. Body movement images with scores or above can be edited into a qualified student data set. The AI module will then receive the coach data set and the qualified student data set as the input layer for artificial intelligence learning, which can be used to optimize the student's physical movement ability.

In addition, the present invention is also a limb movement training method for training the limb movement ability of at least one student. The training method includes the following steps: Upload the limb movement images of a plurality of coaches to an open source data platform, where the open source data platform is provided with In network systems of cloud servers and the Internet.

Display the coach's body movement images for students to visually imitate and learn the coach's body movements.

Capture the body movement images of students.

Compare the student's body movement images and the coach's body movement images to generate image similarity or image difference for subsequent students to correct their body movements.

Among them, in the training method, the student's body movement images will be displayed overlapping with the coach's body movement images to facilitate the students' visual comparison of the differences.

Furthermore, the training method may further include the following steps: Set up the trainer's skeletal structure.

Set the predetermined comparison part in the coach's body movement image.

An error threshold is set corresponding to the comparison part, and the error threshold is used to provide image similarity or image difference comparison.

Among them, when there are multiple comparison parts, it is necessary to set the weights of the multiple comparison parts. The multiple weights are multiplied by the image similarity or image difference of the corresponding comparison parts, and then summed. , and then compared with the error threshold.

Furthermore, in order to further optimize the student's limb movement ability, the training method may also include the following steps: storing multiple coaches' limb movement images as a coaching data set.

According to the image similarity or image difference comparison error threshold, the limb movement images of students that exceed the expected target are screened out as a qualified student data set.

The coach data set and the qualified student data set are received as the input layer of the AI module to optimize the student's physical movement ability.

Therefore, using the limb movement training system and its training method provided by the present invention, an open source data platform can be used to allow students to improve different limb movement abilities from videos provided by multiple coaches, and further utilize the coaching materials in the database. The collection of qualified student information allows the AI module to conduct in-depth learning to continuously optimize students' body movements, thereby further optimizing teaching effectiveness.

The advantages and spirit of the present invention can be further understood through the following detailed description of the invention and the accompanying drawings.

0:Training system

1: Network system

10: Open source data platform

12:Internet

14:Cloud server

15: Coach terminal

16:Control setting module

17: Error threshold

18:Student terminal

20:Display device

22:Receive module

24:Control application module

30:Image capture device

32:Teleport module

40:Image comparison unit

50:Database

60: Screening module

70:AI module

Figure 1 is a general correlation diagram of the limb movement training system of the present invention; Figure 2 is a correlation diagram of the trainer terminal of the training system of the present invention; Figure 3 is a correlation diagram of the student terminal of the training system of the present invention; Figure 4 is a correlation diagram of the limb movement of the present invention The AI module association diagram of the training system; Figure 5 is a flow chart of the limb movement training method of the present invention; and Figure 6 is a flow chart of the further optimized training method of the present invention.

The purpose of the present invention is to provide a limb movement training system and training method that can be used in various sports such as martial arts, dance, gymnastics, sports competitions, rehabilitation, etc., for training the limb movement ability of at least one student.

Please refer to FIG. 1 , which is a general correlation diagram of the limb movement training system 0 of the present invention. The present invention relates to a limb movement training system 0. The training system 0 includes an open source data platform 10, a display device 20, an image capture device 30, and an image comparison unit 40. In addition, the open source data platform 10 is installed in the network system 1. The network system 1 further includes a cloud server 14, the Internet 12, a plurality of coach terminals 15, and a plurality of student terminals 18.

In open source data platforms 10 similar to Youtube, Nintendo, and Will, which have cloud servers 14 and the Internet 12 and can bring together people's creations, multiple coaches can upload any body movement images generated by consuming energy through the body's skeletal muscles. to the open source data platform 10, and the open source data platform 10 is coupled to the display device 20 of the student terminal 18 through the Internet 12. The display device 20 is used to display the coach's body movement image or skeletal structure for students to visually imitate and learn. Coaching body movement.

The student's body movement images can be captured through an image capture device 30 such as a network camera, a mobile phone, a laptop, etc., and the student's body movement images are transmitted to the student terminal through the transmission module 32 in the image capture device 30 The receiving module 22 of the student terminal 18 then compares the student's body movement image and the coach's body movement image through the image comparison unit 40 in the student terminal 18 to generate image similarity or image difference for the student to subsequently correct himself. body movements, that is, when the image similarity or image difference between the student and the coach is too different, the student needs to improve the movements to make the movements of the two more consistent. The image comparison unit 40 is also coupled to the computer through the Internet 12 The open source data platform 10 is provided, and the image comparison unit 40 can be coupled to the image capture device 30 through short-range wireless communication.

The image capture device 30 in the training system 0 can be used in practice such as a mobile phone, a camera with wireless communication function, a camera external to the screen or a built-in camera lens, etc. The image capture device 30 has a transmission module 32, and the trainee The terminal 18 has a receiving module 22. The transmitting module 32 and the receiving module 22 can be short-distance wireless communications such as Wifi, Bluetooth, etc. The transmitting module 32 can be used to transmit the student's body movement images to the receiving module 22. , for subsequent comparison by the image comparison unit 40.

The open source data platform 10 is installed on a cloud server 14, and the cloud server 14 can be communicatively connected to the Internet 12. A plurality of coaching terminals 15 are respectively connected to the Internet 12 for communication, allowing multiple coaches to upload files of their body movement images and transmit them to the open source data platform 10 of the cloud server 14 through the Internet 12. In addition, there are Multiple student terminals 18 are communicatively connected to the Internet 12 . The display device 20 of the student terminal 18 may be, for example, a mobile phone or a computer, and is used to download the coach's body movement images from the open source data platform 10 of the cloud server 14 for students to visually learn.

When the receiving module 22 receives the student's body movement image from the transmitting module 32, it will also be displayed on the display device 20 and overlapped with the coach's body movement image, for example, on the display. In the display device 20, the torso of the coach image will overlap with the torso of the student image, or the head of the coach image will overlap with the head of the student image, and the student image can be presented with low saturation such as gray, and the coach image can be presented with obvious colors such as red. The skeletal structure is marked, that is, the gray student image will overlap with the red coach's influence, allowing students to easily visually discover the difference in posture between themselves and the coach's body movements. Alternatively, the human body can also be divided into left and right sides or upper and lower sides. Then the left and right halves or the upper and lower halves of the student's image will overlap with the instructor's image respectively, allowing for local comparison for more detailed analysis.

Please refer to Figure 2. Figure 2 is a correlation diagram of the coach terminal 15 of the training system 0 of the present invention. The coach terminal 15 in the training system 0 further has a control setting module 16. The control setting module 16 can be used to set the coach's skeletal structure, such as the head, torso, limbs and other stickman-like backbones. The control setting module 16 can also be used to set local parts of the coach's body movement images that are scheduled to be compared, such as left and right hands, left and right feet, and other local parts, and set an error threshold 17 corresponding to the compared local parts. The error threshold 17 is used to provide a reference for comparison of image similarity or image difference. The error threshold 17 can be set arbitrarily. For example, the error threshold 17 of the male coach's left and right hands is 10%, while the female coach's left and right feet The error threshold 17 is 7%, that is, when the student is learning, the comparison position of the left and right hands with the male instructor is less than 10%, which is a qualified standard movement. When the student is learning, the comparison position with the female instructor’s left and right feet is When the error threshold 17 is less than 7%, it is also a qualified standard action.

When there are multiple comparison parts in the training system 0, the control setting module 16 can set the weights of the multiple comparison parts respectively, and the multiple weights are multiplied by the image similarity of the corresponding comparison parts. Or the image difference, and then add it up, and then compare it with the error threshold 17. When it is greater than the error threshold 17, the movement is not standard, and the student needs to continue practicing. When it is less than the error threshold 17, it is a qualified standard movement. Finally, the above settings can be transmitted to the open source data platform 10 of the cloud server 14 through the Internet 12 .

Please refer to Figure 3, which is a correlation diagram of the student terminal 18 of the training system 0 of the present invention. The student terminal 18 further has a display device 20, a receiving module 22, and a control application module 24. The receiving module 22, such as a Wifi or Bluetooth receiver, is used to receive images from the transmitting module 32, and the control application module 24 has an image comparison unit 40. The control application module 24 can be a program component such as a mobile phone. When the receiving module 22 receives the student's body movement images from the transmitting module 32, it can be transmitted to the image comparison unit 40 of the control application module 24 for comparison. pair analysis.

Please refer to FIG. 4 , which is a correlation diagram of the AI module 70 of the limb movement training system 0 of the present invention. The database 50 can be set up in the cloud server 14 to store multiple coaches' body movement images as a coaching data set, and the filtering module 60 can be set up in the student terminal 18 to compare based on image similarity or image difference. Based on the error threshold value of 17 (this result can be converted into a fraction according to the preset), the body movement images of students who exceed the expected target (for example, 80 points) are selected as qualified student data sets, that is, the screening module 60 can be based on image similarity The result of comparing the degree or image difference with an error threshold of 17 forms a score, and the limb movement images that exceed the expected target, such as 80 points or more, are screened out and can be edited into a qualified student data set. In addition, the AI module 70 can be configured in the cloud server 14 to receive the coach data set and the qualified student data set as the input layer data for artificial intelligence learning, whether machine learning, deep learning, or other artificial intelligence learning modes that are combined with each other are used. In this way, the simultaneous optimization of learning from multiple data sources can be used to further optimize students' physical movement abilities without being limited to the learning comparison results of a single coach.

Please refer to FIG. 5 , which is a flow chart of the limb movement training method of the present invention. In addition, the present invention is also a limb movement training method, which is a limb movement training method for various types of sports such as martial arts, dance, gymnastics, sports competitions, rehabilitation... and is used to train the limb movement ability of at least one student. The method includes the following steps:

Step S01: Multiple coaches can shoot body videos to be shared in this teaching.

Step S02: Multiple coaches can upload their own body movement images to an open source data platform 10 that can bring together everyone's creations. The open source data platform 10 is similar to Youtube, Nintendo, and will, and is set up with a cloud server 14 and the Internet. Road 12 in network system 1.

Step S03: Set the coaching parameters, including setting the coaching skeleton structure, predetermined comparison parts, and the error threshold 17 of the comparison parts. Set the predetermined comparison parts in the coach's body movement images to confirm the parts that students need to learn and pay attention to this time, and the error threshold 17 is used to provide image similarity or image difference comparison. Among them, when there are multiple comparison parts, it is necessary to set the weights of the multiple comparison parts. The multiple weights are multiplied by the image similarity or image difference of the corresponding comparison parts, and then summed. , and then compared with the error threshold 17.

Step S04: The display device 20 can display the coach's body movement images so that students can visually imitate and learn the coach's body movements.

Step S05: The student's image can be uploaded to the student's terminal 18, and when the student learns the movements, the student's terminal 18 can capture the student's body movement images and display them in the display device 20, where the student's body movement images will be communicated with the coach's. The body movement images of the students are overlapped and displayed. For example, the student's body movement images are in low-saturation gray, while the coach's body movement images are in high-saturation red, so as to visually confirm whether the two sides overlap.

Step S06: Use the image comparison unit 40 to compare the student's body movement images and the coach's body movement images to generate image similarity or image difference for subsequent students to correct their own body movements.

Please refer to Figure 6, which is a flow chart of further optimizing the training method of the present invention. In order to further optimize students' physical movement abilities, the training method further includes the following steps:

Step S11: In the cloud server 14, the body movement images of multiple coaches can be stored as a coaching data set.

Step S12: Based on the image similarity or image difference comparison error threshold 17 described in step S06, select the body movement images of students that exceed the expected target as a qualified student data set.

Step S13: Receive the coach data set and the qualified student data set as the input layer of the AI module 70, allowing the AI module 70 to continue deep learning using machine learning or deep learning or an artificial intelligence learning mode that is a combination of them, thereby using a variety of data The source's collaborative optimization learning can be used to further optimize students' limb movement abilities to provide a better data set to optimize students' limb movement abilities, that is, to provide more advanced limb movement images for students to learn.

Therefore, using the limb movement training system 0 and its training method provided by the present invention, the open source data platform 10 can be used to allow students to improve different limb movement abilities from videos provided by multiple coaches, and further utilize the database 50 The coach data set and the qualified student data set allow the AI module 70 to conduct in-depth learning to continuously optimize the students' body movements, thereby further optimizing the teaching effectiveness of body movement abilities.

Through the above detailed description of the preferred embodiments, it is hoped that the characteristics and spirit of the present invention can be more clearly described, but the scope of the present invention is not limited by the above disclosed preferred embodiments. On the contrary, the intention is to cover various modifications and equivalent arrangements within the scope of the patent for which the present invention is intended.

0:Training system

1: Network system

10: Open source data platform

12:Internet

14:Cloud server

15: Coach terminal

18:Student terminal

20:Display device

30:Image capture device

32:Teleport module

40:Image comparison unit

50:Database

60: Screening module

70:AI module

Claims (8)

A limb movement training system used to train the limb movement ability of at least one student. The training system includes: an open source data platform for a plurality of coaches to upload their body movement images; a display device coupled to the open source data platform, The display device is used to display the coach's body movement images for the student to visually imitate and learn the coach's body movements; an image capture device is used to capture the student's body movement images; an image comparison unit , coupled to the image capture device and the open source data platform, used to compare the student's body movement image and the coach's body movement image to generate image similarity or image difference for subsequent correction of the body movement by the student. ; A database that stores multiple coaches' body movement images as a coaching data set; a screening module that selects students who exceed the expected target based on the result of comparing the image similarity or image difference with an error threshold. The body movement images serve as a qualified student data set; and an AI module receives the coach data set and the qualified student data set as input layers to optimize the student's body movement ability. The training system as described in item 1 of the patent application, wherein the image capture device has a transmission module, the transmission module is used to transmit body movement images of the student, and the open source data platform is installed in a network system , the network system includes: an Internet; A cloud server, communication connected to the Internet, the open source data platform is set up on the cloud server; a plurality of coaching terminals, communication connected to the Internet respectively, for multiple coaches to upload their body movement images files are transmitted to the open source data platform of the cloud server through the Internet; and multiple student terminals are communicatively connected to the Internet for downloading the coaches from the open source data platform of the cloud server. The student terminal has the display device, a receiving module, and a control application module. The control application module has the image comparison unit. The receiving module receives the image from the transmitting module. The student's body movement images are transmitted to the image comparison unit of the control application module. For example, in the training system described in Item 2 of the patent application, after the receiving module receives the body movement image of the student from the transmission module, it will be displayed on the display device and overlapped with the coach's body movement image. The training system as described in item 2 of the patent application, wherein the coach terminal has a control setting module, and the control setting module is used to set a predetermined ratio between the coach's skeletal structure and the coach's body movement image. The error threshold is set corresponding to the comparison part, and the error threshold is used to provide image similarity or image difference comparison. As in the training system described in item 4 of the patent application, when there are multiple comparison parts, the control setting module can set the weights of the multiple comparison parts, and the multiple weights are respectively Multiply the image similarity or image difference of the corresponding comparison parts, then add them up, and compare them with the error threshold. A limb movement training method for training the limb movement ability of at least one student. The training method includes the following steps: uploading a plurality of coaches' body movement images to an open source data platform, wherein the open source data platform is set up with a cloud server In the network system of the computer and the Internet; display the coach's body movement images for the student to visually imitate and learn the coach's body movements; capture the student's body movement images; compare the student's body movement images And the coach's body movement images to generate image similarity or image difference for the student to subsequently correct the body movement; store multiple coaches' body movement images as a coaching data set; based on the image similarity or image difference ratio Based on an error threshold, select the body movement images of students who exceed the expected target as a qualified student data set; and receive the coach data set and the qualified student data set as the input layer of an AI module to optimize the student's limbs Athletic ability. For example, in the training method described in Item 6 of the patent application, the student's body movement images will be displayed overlapping with the coach's body movement images. The training method as described in item 6 of the patent application, wherein the training method further includes the following steps: setting the coach's skeletal structure; setting predetermined comparison parts in the coach's body movement images; and The error threshold is set corresponding to the comparison part, and the error threshold is used to provide image similarity or image difference comparison; wherein, when there are multiple comparison parts, the weights of the multiple comparison parts are set. , the multiple weights are multiplied respectively by the image similarity or image difference of the corresponding comparison parts, and then added together, and then compared with the error threshold.