KR102652525B1 - Artificial intelligence coaching system for motion correction - Google Patents

Abstract

The present invention relates to a motion correction coaching system, and specifically, to recognize the user's exercise motion through a terminal with a camera function at home, etc., and to perform correction and coaching through voice, etc., by collecting data through a plurality of camera terminals and This is about an artificial intelligence coaching system for motion correction that can improve motion recognition accuracy by tracking human joints through deep learning.

Description

{Artificial intelligence coaching system for motion correction}

The present invention relates to a motion correction coaching system. Specifically, the present invention relates to a motion correction coaching system. In detail, the user's exercise motion is recognized through a terminal with a camera function at home, etc., and correction and coaching are performed through voice, etc., data collection through a plurality of camera terminals, and This is about an artificial intelligence coaching system for motion correction that can improve motion recognition accuracy by tracking human joints through deep learning.

In the rehabilitation treatment of patients who have difficulty walking or moving on their own due to various causes, including brain damage and muscle damage, they receive help from their guardians or therapists to help them move their weakened muscles.

For this purpose, patients must periodically visit medical institutions such as hospitals or rehabilitation centers, and after performing rehabilitation exercises in a designated space at a set time, they have the difficulty of returning home with an uncomfortable body, especially for the elderly and those with poor walking ability or joints. These difficulties are aggravated in the case of patients requiring rehabilitation.

Accordingly, with the recent rapid development of IT technology, various methods are being researched that can help patients maintain/improve their health while improving convenience by supporting them to do rehabilitation exercises at home without having to go to a facility.

Most of these provide only a simple function for patients to follow the instructor's exemplary movements and postures, making it impossible to judge the accuracy of the movements, and may even cause side effects due to incorrect movements, which is not good in terms of user convenience or the maintenance and promotion of health. Efficiency was low.

In addition, due to the recent COVID-19 pandemic, the proportion of people doing exercise on their own at home, which they used to do under guidance at hospitals, personal training, and Pilates shops, has increased, and most people are stuck at the level of following videos.

In addition, rehabilitation exercises (physical therapy, occupational therapy) that require a lot of physical contact with the therapist trained at school are also difficult to provide close instruction, and many rehabilitation exercises were canceled during the period when strong social distancing policies were implemented to respond to the coronavirus. . As rehabilitation exercises are not carried out smoothly, public health deteriorates and this also affects the deterioration of the country's health insurance finances. In addition, rehabilitation exercise-related practice at universities etc. is also not carried out well due to the aftermath of the coronavirus, which inevitably leads to a decline in the quality of education. There is no such thing.

Republic of Korea Patent Publication No. 10-2013-0099321 (2013.09.06)

The present invention was created to solve the above problems, and the purpose of the present invention is to develop an AI coaching algorithm that can improve motion recognition accuracy by collecting data through multiple camera terminals and tracking human joints through deep learning. By applying this, we provide an artificial intelligence coaching system for movement correction that recognizes the user's exercise movements through a terminal with a camera function in various indoor spaces, such as at home, and provides accurate posture correction and coaching through voice, etc.

For the above purpose, the present invention provides a coaching system for motion correction, which includes a rail portion installed around a subject performing a set motion, and a mounting structure moving along the rail portion and having a plurality of mounting portions formed on the upper side. A frame structure consisting of a main moving body; A learning terminal mounted on the holder and having a camera for photographing a subject; An image classification unit that collects images collected through the camera in response to the movement of the main moving object and organizes them by movement, and a first attribute extraction unit that extracts first movement data based on the joint area from the classified images for each movement. A learning module including a unit and an image learning unit that learns the first motion data and generates reference information; A second attribute extraction unit that analyzes motion images captured through a user terminal equipped with a camera and extracts second motion data based on the joint area, and generates correction information by comparing the second motion data with reference information. a coaching module including a judgment unit that outputs the correction information through video and audio; It is characterized by consisting of.

At this time, the rail part is configured so that the tube body with a penetration part formed along the longitudinal direction forms a ring shape, and the main moving body includes a support roller part that is inserted into the penetration part and contacts the inner wall of the tube body, and a portion of the tube that is in contact with the bottom surface. It may be provided with a plurality of wheels that rotate through a driving means.

In addition, the holding structure may include a plurality of auxiliary supports that are formed in the form of a pillar perpendicular to the upper side of the main movable body and each side of which has wheels formed at the bottom and supported on the floor.

In addition, an auxiliary movable body that is inserted into the penetrating part and has a support roller part in contact with the inner wall of the tube body, and a plurality of wheels in contact with the bottom surface, but is disposed in a position symmetrical to the main movable body with respect to the center of the ring-shaped rail part. It further includes, and the holding structure may be composed of a support bar connecting vertically formed pillars to the main movable body and the auxiliary movable body, respectively.

Through the present invention, various exercises, including rehabilitation training, can be performed remotely at home without the need for a visit to a specialized institution or the help of an expert, and accurate posture coaching can be achieved. Compared to existing kiosk-type equipment that requires a relatively large space and is expensive, user convenience can be greatly improved by being able to exercise while receiving AI coaching anywhere indoors with a simple preparation such as a smartphone and a stand. Even if you don't look at it, it recognizes your posture and coaches you through voice to correct it, so you can focus on exercising and it's very useful for sweaty workouts.

1 is a conceptual diagram of the present invention;
Figure 2 is a block diagram showing the configuration and connection relationship according to an embodiment of the present invention;
Figure 3 is a plan view showing the frame structure according to one embodiment of the present invention;
Figure 4 is a partial perspective view showing the coupling structure of the frame structure according to one embodiment of the present invention;
Figure 5 is a perspective view showing the structure of a frame structure according to one embodiment of the present invention;
Figure 6 is a perspective view showing the structures of the main moving body, the auxiliary moving body, and the frame structure according to the second embodiment of the present invention.

Hereinafter, the configuration of the artificial intelligence coaching system for motion correction of the present invention will be described in detail with reference to the attached drawings.

1 is a conceptual diagram of the present invention, and FIG. 2 is a block diagram showing the configuration and connection relationship according to an embodiment of the present invention. The present invention provides low motion recognition accuracy in existing remote motion coaching programs and provides guidance by viewing the screen along with exercise motions. In order to solve the problem of difficulty in receiving recognition, the data that learned the movements of the body joints performing the movement through multiple cameras is used to greatly increase recognition accuracy and improve convenience by coaching posture through voice. The main components for this purpose include a frame structure (1), a learning terminal (2), a learning module (3), and a coaching module (4).

Figure 3 is a plan view showing the frame structure according to one embodiment of the present invention. As mentioned above, 2D (x, y) for accurate motion recognition when extracting and learning exercise postures through multiple cameras is used. In order to acquire 3D (x, y, z) image data, a frame structure 1 is provided in which a plurality of cameras rotate around a subject performing an action and take pictures.

For this purpose, the frame structure 1 consists of a rail part 11 and a main moving body 12 that moves along the rail part 11 and has a plurality of cameras 21 installed.

The rail unit 11 is a structure installed around a subject performing a set operation. It is preferably installed on the floor for easy and stable installation, but is configured in an annular shape, that is, in a ring shape, so that the subject is located in the center. In addition, the cross section may be configured in various shapes, but it is preferable to be composed of a circular tubular body 111, and as will be described later, not only the main moving body 12 but also, in some cases, the auxiliary moving body 13 may be used in the rail portion 11. ), so that the penetrating portion 112 is formed along the longitudinal direction on the side, preferably on the outside, to prevent their separation, that is, it is configured to have a 'C' shaped cross section.

Exercises such as rehabilitation treatment, yoga, pilates, and home training that are usually performed at home are performed on a mat that is large enough for one person to lie down, so a rail part (11) is provided to have an inner diameter larger than the mat, and is space-efficient. For convenience of use and storage, it is desirable to configure the rail portion 11 in a form that can be folded, separated, and assembled.

Figure 4 is a partial perspective view showing the coupling structure of the frame structure according to one embodiment of the present invention, showing a rail portion that can be separated and assembled.

That is, the rail portion 11 is formed with a penetrating portion 112 along the longitudinal direction and is configured to form a ring shape by connecting a plurality of tube bodies 111 forming an arc, and the outer diameter of one end of each tube body 111 is It can be configured to be assembled in a reduced form so that it is inserted into the other side of another pipe body. At this time, to prevent the connected tube body 111 from being unintentionally separated, an elastic protrusion is formed on one tube body of the connecting part 113, and a through hole is formed on the other tube body of the connected connecting part 113 at a position corresponding to the elastic protrusion. The elastic protrusion is inserted into the through hole and a solid assembled state can be maintained.

In the embodiment of the present invention, the rail portion 11 is divided into eight parts and presented in a form that can be separated and assembled, but it is not limited to this and can be divided into a variety of numbers of tubular bodies 111, and the material of the tubular body 111 forming the rail portion is It is possible to adjust the diameter of the rail part within a certain range through the number of connected pipe bodies 111 by making it a material with flexible characteristics within a certain range.

Figure 5 is a perspective view showing the structure of a frame structure according to one embodiment of the present invention.

The main moving body 12 moves along the rail unit 11 with a plurality of cameras, essentially a learning terminal 2 with a built-in camera 21 installed, and takes pictures in all directions and produces 3D image data. In a configuration that supports collection, a mounting structure 16 having a plurality of mounting portions 161 on which learning terminals are installed is formed on the upper side.

Specifically, a support roller part 14 is formed on the side of the main moving body 12, which is inserted into the penetrating part 112 and contacts the inner wall of the tube body 111 to prevent it from being separated while moving along the rail part 11. It is configured and is provided with a plurality of wheels (W) that are in contact with the floor surface and some of which rotate through the driving means (15). As the driving means 15, an electric driving means such as a motor may be applied. However, in the present invention, shooting and movement of the main moving object 12 are performed in a relatively short period of time while maintaining a specific motion, rather than being driven for a long time. As this is achieved, movement can be achieved without a separate power source or control means by applying a driving means that operates through a mainspring.

The mounting portion 161 is made up of various types of grips that can temporarily hold the learning terminal 2, and in one embodiment of the present invention, the mounting structure 16 is formed vertically upward to the main moving body 12. ) is provided with a mounting portion 161 at a set interval in the vertical direction. In the attached drawing, as three mounting units 161 are provided, the actions taken by the central subject can be filmed by moving the main moving object 12 around the area so that the entire subject can be filmed through the three cameras 21. is shown, and the number of mounting parts provided can be appropriately added or subtracted depending on the size of the rail part.

At this time, the main moving body (12) is configured to have an appropriate area and wheels (W) to stably support and move the weight of the installation of a plurality of learning terminals (2), and the holding structure (16) has lower ends on both sides. A plurality of auxiliary supports 162 formed with wheels W supported on the floor are formed to prevent shaking or falling due to centrifugal force during movement.

In this way, practically necessary 3D image data can be obtained by rotating the main moving object 12 360 degrees around the subject, but image data can also be obtained by simultaneously shooting from the side and top of the subject.

Figure 6 is a perspective view showing the structure of the main moving body, the auxiliary moving body, and the frame structure according to the second embodiment of the present invention, which is disposed at a position symmetrical to the main moving body 12 with respect to the center of the ring-shaped rail portion 11. An auxiliary movable body 13 is provided, and the holding structure 16 presents a structure consisting of a support bar 164 connecting vertically formed pillars 163 to the main movable body 12 and the auxiliary movable body 13, respectively. I'm doing it.

The auxiliary moving body 13 has substantially the same structure as the main moving body 12 and is configured to move along the rail unit 11, but is different in that it is not equipped with a separate driving means and moves freely by external force. .

At this time, the main moving body 12 and the auxiliary moving body 13 are connected through an interlocking means 165 such as a wire installed in a ring shape along the inside of the rail portion 11, and the interlocking means ( 165) can be configured to allow connected auxiliary mobile devices to move together.

The holding structure 16 is a support bar 164 that connects the pillars 163 formed perpendicularly to the main mobile body 12 and the auxiliary mobile body 13, and crosses the upper side of the subject and supports the main mobile body 12. It is configured to connect the auxiliary mobile body 13, and a mounting portion 161 is formed at a set interval along the pillar 163 and the support bar 164, so that the learning terminal 2 installed on the mounting portion 161 is Take a picture of the subject in the center. In the attached drawing, two mounting parts 161 are installed on each pillar 163 and five mounting parts 161 are installed on the support bar 164 at intervals, so that the subject's whole body can be used through a total of nine learning terminals (2). A structure in which photography can be performed is shown, but the present invention is not limited to this and the number of mounting portions 161 can be adjusted appropriately.

Likewise, a plurality of auxiliary supports 162 formed with wheels W supported on the floor are formed on each pillar to prevent shaking or falling due to centrifugal force during movement.

The learning terminal 2 is mounted on the holder 161 and has a camera 21 for photographing the subject to acquire images for learning purposes. It includes a communication unit 22 that transmits the captured images and an acceleration sensor. Including a sensor unit 23 capable of measuring movement according to the movement of the main mobile body based on a sensor, in the same structure as the main mobile body 12 and the second embodiment, movement location information according to the movement of the auxiliary mobile body 13 After recording the video, it is transmitted to the learning module (3).

The learning module (3) is configured to collect and learn images transmitted through the learning terminal (2), and is comprised in detail of an image classification unit (31), a first attribute extraction unit (32), and an image learning unit. (33) is provided.

The image classification unit 31 collects images collected through the camera 21 in response to the movement of the main moving object 12 and organizes them by operation. Basically, videos of the same action are grouped by the shooting angle, so that learning can be done by classifying the same action and the same shooting angle.

The first attribute extraction unit 32 extracts first motion data based on the joint area from the images for each motion classified through the image classification unit 31. Basically, the human body moves around joints, so major joints such as neck, shoulder, arm, wrist, waist, hip joint, knee, and ankle are extracted and 3D movement on the xyz axis centered on these as first motion data. It is extracted.

The image learning unit 33 is configured to learn the first motion data extracted through the first attribute extraction unit 32, and generates reference information that becomes the standard for motion recognition in the future. In this process, frames containing unnecessary movements and postures can be removed to suit the characteristics of each movement, and frame images are analyzed to learn common parts of movement centered on joints that are important in the movement to generate reference information.

The coaching module (4) is a configuration for coaching users to take accurate movements by filming movements such as rehabilitation treatment, yoga, pilates, etc. at home using a user terminal using previously learned data. It is structured in detail. It is provided with a second attribute extraction unit 41, a judgment unit 42, and a guidance unit 43.

The user terminal is a terminal owned by a user who wishes to receive the coaching service according to the present invention, and is equivalent to a smartphone equipped with a camera for motion capture and an interface for voice guidance.

The second attribute extraction unit 41 is configured to extract second motion data based on joint parts by analyzing motion images captured through the camera of the user terminal, and is the first attribute extraction unit 32 described above. In the same way, major joints such as neck, shoulder, arm, wrist, waist, hip joint, knee, and ankle are extracted, and 3D movement on the xyz axis centered on these is extracted as second motion data.

The determination unit 42 is configured to generate correction information by comparing the second motion data extracted through the second attribute extraction unit 41 with reference information generated by the learning module 3. Basically, reference information for the same motion as the user's motion is retrieved, the movement centered on each joint is compared, and correction information related to the correct angle or movement is generated based on this.

The information unit 43 is configured to output the correction information generated through the determination unit 42 through video and audio. As mentioned earlier, the correction information is output as audio along with the video by instructing to raise or lower the height of each joint of the human body or to turn it left or right. Even if you do not see the video, it is clearly displayed through the audio. We provide detailed explanations so that you can understand clearly, and we ensure that movement coaching can be easily done even in situations where you cannot see the video due to the nature of the movement.

The rights of the present invention are not limited to the embodiments described above but are defined by the claims, and those skilled in the art can make various changes and modifications within the scope of the claims. This is self-evident.

1: Frame structure 11: Rail part 111: Pipe body
112: Penetrating part 113: Connection part
12: Main moving body 13: Secondary moving body
14: support roller part 15: driving means
16: Holding structure 161: Holding part
162: auxiliary support 163: pillar
164: support bar 165: interlocking means
2: Learning terminal 21: Camera 22: Communication department
23: Sensor unit
3: Learning module 31: Image classification unit 32: First attribute extraction unit
33: Video Learning Department
4: Coaching module 41: Second attribute extraction unit 42: Judgment unit
43: Information department
W: wheel

Claims (4)

In the coaching system for movement correction,
The main moving body (12) is formed on the upper side with a rail portion (11) installed surrounding the subject performing the set motion, and a mounting structure (16) moving along the rail portion (11) and having a plurality of mounting portions (161). ) frame structure (1) consisting of;
A learning terminal (2) mounted on the mounting unit (161) and equipped with a camera (21) for photographing the subject;
An image classification unit 31 that collects images collected through the camera 21 in response to the movement of the main moving object 12 and organizes them by movement, and a first classifier based on the joint area in the classified images for each movement. A learning module (3) including a first attribute extraction unit (32) that extracts motion data, and an image learning unit (33) that learns the first motion data and generates reference information;
A second attribute extraction unit 41 that analyzes motion images captured through a user terminal equipped with a camera and extracts second motion data based on the joint region, and corrects the second motion data by comparing it with reference information. A coaching module (4) including a determination unit (42) that generates information and a guidance unit (43) that outputs the correction information through video and audio; An artificial intelligence coaching system for motion correction, characterized by consisting of.
According to paragraph 1,
The rail portion 11 is,
The tube body with the penetrating portion 112 formed along the longitudinal direction is configured to form a ring shape,
The main moving body 12 is,
It is characterized by having a support roller part (14) inserted into the penetration part (112) and in contact with the inner wall of the tube body, and a plurality of wheels (W) that are in contact with the bottom surface and some of which rotate through the driving means (15). An artificial intelligence coaching system for movement correction.
According to paragraph 2,
The holding structure 16 is formed in the form of a pillar vertically formed on the upper side of the main moving body 12, and includes a plurality of auxiliary supports 162 with wheels (W) formed at the bottom on both sides and supported on the floor. An artificial intelligence coaching system for motion correction, characterized by:
According to paragraph 2,
It is provided with a support roller part (14) inserted into the penetration part (112) and in contact with the inner wall of the tube body, and a plurality of wheels (W) in contact with the bottom surface, and the main It further includes an auxiliary mobile body (13) disposed in a position symmetrical to the mobile body (12),
The holding structure 16 is composed of a support bar 164 connecting vertically formed pillars 163 to the main mobile body 12 and the auxiliary mobile body 13, respectively. Artificial intelligence coaching for motion correction system.

Images