KR102584654B1 - methods for recommending and managing of exercise lectures - Google Patents

Abstract

Disclosed is a method for recommending and managing exercise lectures. The method comprises the steps of: converting normal exercise images previously stored in a server into input sequences for learning by using a preprocessing module; training a posture analysis module based on the input sequences for learning and learning posture analysis information corresponding thereto; receiving exercise images from a student terminal and an instructor terminal, and preprocessing the exercise images by using the preprocessing module to generate input sequences for a plurality of joint position changes, wherein the exercise images include the student's first exercise image and the instructor's second exercise image; applying the input sequences to the posture analysis module to generate posture analysis information for the exercise images, wherein the posture analysis information includes first posture analysis information associated with the first exercise image and second posture analysis information associated with the second exercise image; comparing the first posture analysis information and the second posture analysis information to calculate posture accuracy for each part, and deriving recommended lecture lists by using the calculated posture accuracy; and providing the recommended lecture lists to the student terminal. Therefore, the method can recommend experts that suit the student's needs.

Description

Methods for recommending and managing exercise lectures}

The present invention relates to a method for recommending and managing exercise classes. Specifically, the present invention relates to a method of providing appropriate exercise classes to students and managing the students.

The content described in this section simply provides background information for this embodiment and does not constitute prior art.

As national income increases and we enter an aging age, life expectancy in our society is gradually increasing, and society as a whole is focusing on health promotion to maintain or improve the quality of life.

Health promotion refers to taking healthy measures to eat, excrete, breathe, sleep, work, rest, and exercise in the process of physically living in daily life, and to lead a mentally healthy life.

To improve health, people choose the exercises they need to improve things like flexibility, strength, and endurance.

However, if exercising with incorrect or incorrect posture, the exerciser may get injured and inevitably require treatment due to damage to muscles/bones/nerves, and even if no injury occurs, exercise effectiveness may be reduced or non-existent.

Therefore, it is important to exercise with correct posture. Accordingly, people want to exercise while receiving guidance from experts.

However, even in order to receive guidance from an expert, when selecting an expert, the means and methods to verify the expert are limited to the expert's history, and it is not easy to find an expert who is suitable for the person who wants to exercise.

Additionally, as people are unable to exercise in offline spaces due to forced environments such as the coronavirus, the home training industry and non-face-to-face exercise industry where people exercise at home are growing.

Home training and non-face-to-face exercise cannot be directly guided by an expert who teaches correct exercise posture.

An expert (instructor) can check the student's video provided in real time and suggest direction, but since he or she can only provide guidance based on rough judgment with the human eye, it is difficult to provide precise guidance on posture, and accuracy inevitably decreases.

In addition, there are many cases where people exercise at home through home training and non-face-to-face exercise, but due to noise between floors and narrow space, it may not be possible to exercise smoothly by enduring uncomfortable factors and exercising.

Korean Patent Publication No. 10-2306055

The object of the present invention is to provide a method for analyzing the user's exercise posture based on the user's exercise image and managing the user's exercise based on the analyzed exercise posture.

In addition, another object of the present invention is to provide a method for verifying the expertise of an exercise instructor by analyzing the exercise posture based on the exercise instructor's exercise video.

In addition, another object of the present invention is to provide a method for extracting user-tailored exercise lectures and providing them to the user, and allowing the user to take appropriate exercise lectures.

In addition, another task of the present invention is to provide an environment in which users can exercise by providing exercise facilities suited to the user's environment.

The objects of the present invention are not limited to the objects mentioned above, and other objects and advantages of the present invention that are not mentioned can be understood by the following description and will be more clearly understood by the examples of the present invention. Additionally, it will be readily apparent that the objects and advantages of the present invention can be realized by the means and combinations thereof indicated in the patent claims.

A method for recommending and managing exercise classes according to some embodiments of the present invention to solve the above problems includes converting a normal exercise image pre-stored in a server into a learning input sequence using a preprocessing module, the learning input sequence and its corresponding Learning a posture analysis module based on learning posture analysis information, receiving exercise images from the student terminal and the instructor terminal, preprocessing the exercise images using the preprocessing module, and inputting a plurality of joint position changes Generating a sequence, wherein the exercise image includes a first exercise image of the student and a second exercise image of the instructor, applying the input sequence to the posture analysis module to generate posture analysis information for the exercise image. However, the posture analysis information includes first posture analysis information associated with the first exercise image and second posture analysis information associated with the second exercise image, the first posture analysis information and the second posture analysis. Comparing information to calculate posture accuracy for each part, deriving a list of recommended lectures using the calculated posture accuracy, and providing the list of recommended lectures to the student terminal.

In addition, generating the input sequence includes recognizing the position of a main joint in each frame of a plurality of frames included in the exercise image, and applying joint recognition information about the position of the main joint to each frame. , generating a plurality of joint vectors based on changes in the joint recognition information, and arranging the plurality of joint vectors in sequential frame order to generate the input sequence.

In addition, the posture analysis module includes an autoencoder including a neural network and an abnormality determination unit that determines whether the input sequence is abnormal, and the step of generating the posture analysis information includes applying the input sequence to the autoencoder. It may further include generating an output sequence and generating posture analysis information based on the difference between the input sequence and the output sequence in the abnormality determination unit.

In addition, the step of deriving the recommended lecture list includes generating first matching instructor information that matches the student and the instructor for each region based on the posture accuracy for each region, the first matching instructor information and The method may further include deriving a recommended instructor based on a set weight of the portion and extracting a lecture corresponding to the recommended instructor from pre-stored lecture information to derive the recommended lecture list.

Additionally, the part weight may be a preset value for at least one body part according to each exercise type.

In addition, the step of deriving the recommended lecture list includes two of the posture accuracy, the student's preference received from the student terminal, the time schedule of the student and the instructor, the student's location proximity to the lecture location, and the lecture cost. A step of deriving the recommended lecture list based on the above parameters may be further included.

In addition, loading the pre-stored course schedule location, preset distance, and affiliated lecture facility information from the database unit included in the server, extracting a list of first lecture facilities available for rent located within the preset distance based on the class schedule location. A step of extracting a second lecture facility list suitable for the lecture information based on the first lecture facility list and the lecture information taken by the student from the first lecture facility list and including the second lecture facility list in the second lecture facility list. A step of generating a final lecture facility list based on the rental status information of each lecture facility may be further included.

The exercise lecture recommendation and management method of the present invention can verify the expert's expertise in detail by analyzing the exercise posture of the expert teaching the exercise lecture, and recommend an expert suitable for the needs of the student.

Additionally, the present invention can analyze the student's exercise posture using artificial intelligence and provide detailed exercise feedback to the student.

Additionally, the present invention can provide a place for students taking non-face-to-face exercise classes to take exercise classes.

In addition to the above-described content, specific effects of the present invention are described below while explaining specific details for carrying out the invention.

1 is a schematic diagram of an exercise lecture recommendation management system to illustrate a server according to some embodiments of the present invention.
FIG. 2 is a block diagram for explaining the server of FIG. 1 in detail.
FIG. 3 is a flowchart for explaining the operation of the server of FIG. 2.
Figure 4 is a block diagram for explaining in detail the learning step of the exercise posture analysis unit of Figure 2.
FIG. 5 is a block diagram for explaining the preprocessing module of FIG. 4 in detail.
FIG. 6 is a flowchart for explaining the operation of the preprocessing module of FIG. 5.
FIG. 7 is a diagram for explaining the joint recognition module of FIG. 5.
FIG. 8 is a diagram for explaining the sequence generation module of FIG. 5.
FIG. 9 is a block diagram for explaining in detail the execution steps of the exercise posture analysis unit of FIG. 2.
FIG. 10 is a block diagram showing the posture analysis module of FIG. 9.
FIG. 11 is a block diagram showing data flow to explain another embodiment of the posture analysis module of FIG. 9.
FIG. 12 is a flowchart for explaining the posture analysis module of FIG. 11 in detail.
FIG. 13 is a block diagram to explain some examples of the autoencoder of FIG. 11.
Figure 14 is a block diagram for explaining in detail the lecture recommendation unit of Figure 2.
Figure 15 is a flowchart for explaining in detail the operation of the lecture recommendation unit of Figure 14.
FIG. 16 is a diagram for explaining in detail the method of deriving the recommended instructor of FIG. 15.
FIG. 17 is a block diagram for explaining the location reservation unit of FIG. 2 in detail.
FIG. 18 is a flowchart for explaining in detail the operation of the place reservation unit of FIG. 17.
FIG. 19 is a diagram for explaining a method of generating the lecture facility list of FIG. 17.
Figure 20 is a block diagram for explaining in detail the student management unit of Figure 2.
Figure 21 is a diagram for explaining the hardware implementation of a system that performs a method for recommending and managing exercise classes according to some embodiments of the present invention.

Terms or words used in this specification and patent claims should not be construed as limited to their general or dictionary meaning. According to the principle that the inventor can define terms or word concepts in order to explain his or her invention in the best way, it should be interpreted with a meaning and concept consistent with the technical idea of the present invention. In addition, the embodiments described in this specification and the configurations shown in the drawings are only one embodiment of the present invention and do not completely represent the technical idea of the present invention, so they cannot be replaced at the time of filing the present application. It should be understood that there may be various equivalents, variations, and applicable examples.

Terms such as first, second, A, and B used in the present specification and claims may be used to describe various components, but the components should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another. For example, a first component may be named a second component, and similarly, the second component may also be named a first component without departing from the scope of the present invention. The term 'and/or' includes any of a plurality of related stated items or a combination of a plurality of related stated items.

The terms used in the specification and claims are merely used to describe specific embodiments and are not intended to limit the invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this application, terms such as "include" or "have" should be understood as not precluding the existence or addition possibility of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification. .

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by a person of ordinary skill in the technical field to which the present invention pertains.

Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and unless clearly defined in the present application, should not be interpreted in an ideal or excessively formal sense. No.

Additionally, each configuration, process, process, or method included in each embodiment of the present invention may be shared within the scope of not being technically contradictory to each other.

Hereinafter, with reference to FIGS. 1 to 21, we will look at the exercise lesson recommendation and management method performed in the exercise lesson recommendation management system according to some embodiments of the present invention.

1 is a schematic diagram illustrating an exercise lecture recommendation management system according to some embodiments of the present invention.

Referring to FIG. 1, an exercise lecture recommendation management system according to some embodiments of the present invention may include a server 100, a student terminal 200, and an instructor terminal 300.

The student terminal 200 and the instructor terminal 300 are user devices capable of communicating with the server 100 over a network. For example, the student terminal 200 and the instructor terminal 300 may refer to mobile electronic devices such as mobile phones, smartphones, and wearable devices (eg, watch-type terminals, etc.) that may have unique identification numbers. Here, the student terminal 200 and the instructor terminal 300 are of course not limited to a specific operating system (eg, iOS or Android operating system). In addition, the student terminal 200 and the instructor terminal 300 include a terminal input unit that receives user input, a terminal display unit that displays information received through the network 400, a terminal communication unit that transmits and receives signals to and from the outside, and It may include a terminal control unit that processes data, controls each internal unit of the student terminal 200 and the instructor terminal 300, and controls data transmission/reception between the units.

Additionally, the student terminal 200 and the instructor terminal 300 may be devices including a camera module capable of recording exercise images. The student terminal 200 may transmit the first exercise image captured by the student's exercise posture using a camera module to the server 100. The instructor terminal 300 may transmit a second exercise video captured of the instructor's exercise posture to the server 100 using a camera module. Here, the first and second exercise images may be images or videos.

Additionally, in another embodiment of the present invention, the exercise lecture recommendation management system further includes a separate photographing device, and the photographing device may be installed in an affiliated lecture facility. The imaging device may capture the first and second exercise images respectively and provide them to the server 100. Additionally, the photographing device may generate motion data consisting of continuous images and provide the motion data to the server 100. Hereinafter, for convenience of explanation, an example will be given where the student terminal 200 captures the student's first exercise video and the instructor terminal 300 captures the instructor's second exercise video.

FIG. 2 is a block diagram for explaining the server of FIG. 1 in detail, and FIG. 3 is a flow chart for explaining the operation of the server of FIG. 2.

Referring to FIG. 2, the server 100 includes a data reception unit 110, a database unit 120, an exercise posture analysis unit 130, a lecture recommendation unit 140, a venue reservation unit 150, and a student management unit 160. and a control unit 170.

Specifically, the data receiving unit 110 can receive data required by each element of the server 100. For example, the data receiving unit 110 may receive the first exercise image, the student's preference, the student's time schedule, and the scheduled class date from the student terminal 200 and transmit them to other components in the server 100. In addition, the data receiving unit 110 may receive the second exercise image and the instructor's time schedule from the instructor terminal 300, and transmit the received second exercise image and time schedule to other components in the server 100. there is. The data receiving unit 110 may use various communication modules and may exchange data between the server 100, the student terminal 200, and the instructor terminal 300 through a network. Each data received and transmitted by the data receiving unit 110 will be described in detail later while explaining each component included in the server 100.

The database unit 120 may store various data used by at least one component of the server 100. Data stored in the database unit 120 may include input data or output data for software and commands related thereto.

Additionally, the database unit 120 may store in advance a normal exercise dataset, which is a set of normal exercise images for each type of exercise, and lecture information on exercise lectures.

The normal exercise dataset may include images of normal exercise in the correct posture for each type of exercise. Additionally, the normal exercise dataset includes at least one normal exercise image for each exercise, allowing the server 100 to accurately analyze the exercise posture. For example, the normal exercise dataset may include at least one normal exercise image for the squat exercise for each age. Additionally, the normal exercise dataset may include mapping of learning posture analysis information corresponding to the normal exercise image. Here, the learning posture analysis information is posture analysis information and may include a numerical value indicating the accuracy of the normal exercise image. For example, if accuracy is defined as a probability value in the range from 0 to 1, the learning attitude analysis information may be 1 or a value close to 1. However, this description is merely illustrative, and the embodiments of the present invention are not limited thereto.

Lecture information is information about lectures offered for at least one exercise, and for each lecture, it may include lecture identification, lecture name, exercise type, instructor identification, lecture schedule, tuition, lecture location, and non-face-to-face status. In addition, the lecture information includes an instructor rating that evaluates the instructor calculated based on posture analysis information (hereinafter referred to as second posture analysis information) that analyzes the instructor's exercise posture based on the second exercise video of each instructor in the present invention. It can be included.

Additionally, the database unit 120 may store and manage the input sequence preprocessed by the exercise posture analysis unit 130 and the posture analysis information generated by the exercise posture analysis unit 130.

The exercise posture analysis unit 130 is learned based on the normal exercise data set stored in the database unit 120, and the first exercise image received from the student terminal 200 and the second exercise image received from the instructor terminal 300. By analyzing each exercise posture, posture analysis information can be generated. For example, the exercise posture analysis unit 130 may analyze the first exercise image to generate first posture analysis information and analyze the second exercise image to generate second posture analysis information. A detailed description of the exercise posture analysis unit 130 will be described later with reference to FIGS. 4 to 13.

The lecture recommendation unit 140 may derive a list of recommended instructors and recommended lectures based on the posture analysis information generated by the exercise posture analysis unit 130. For example, the lecture recommendation unit 140 derives a list of recommended lectures based on at least one of the first and second posture analysis information, the student's preference, the time schedule for the student and the lecture, location proximity, and lecture cost. and can be provided to the student terminal 200. A detailed description of this will be provided later with reference to FIGS. 14 to 16.

The venue reservation unit 150 may generate a final lecture facility list of lecture facilities available for rental among lecture facilities associated with course application information and provide the final lecture facility list to the student terminal 200. The course application information is received from the student terminal 200 and is information about a course application corresponding to at least one of the recommended course list generated by the course recommendation unit 140. A detailed description of the place reservation unit 150 will be described later with reference to FIGS. 17 to 19.

The student management unit 160 may provide schedule notification information for course management to the student terminal 200 based on the course history including course application information corresponding to the student terminal 200. In addition, the student management unit 160 may provide exercise feedback information to the student terminal 200 based on the first posture analysis information and InBody information generated by the exercise posture analysis unit 130 in association with the student terminal 200. there is. A detailed description of this will be provided later with reference to FIG. 20.

The control unit 170 performs a function of controlling the operation of each component included in the server 100. The control unit 170 can perform operations such as connection between each component, data exchange, and command execution. Accordingly, the controller 170 may be the entity responsible for recommending and managing exercise classes performed by the server 100.

Below, the operation of the server 100 according to some embodiments of the present invention will be described with reference to FIGS. 3 and 4.

FIG. 3 is a flowchart for explaining the operation of the server of FIG. 2. Figure 4 is a block diagram for explaining in detail the learning step of the exercise posture analysis unit of Figure 2.

3 and 4, the exercise posture analysis unit 130 can learn the posture analysis module 133 included in the exercise posture analysis unit 130 based on the normal exercise image stored in the database unit 120. there is.

Specifically, the exercise posture analysis unit 130 may convert the normal exercise image stored in the database unit 120 into an input sequence for learning using the preprocessing module 131 (S110). A detailed description of how the exercise posture analysis unit 130 preprocesses the exercise image using the preprocessing module 131 will be described later with reference to FIGS. 5 to 8 .

The exercise posture analysis unit 130 may learn the posture analysis module 133 based on the learning input sequence and the corresponding posture analysis information (S120). The operations of the pre-processing and learning steps of the exercise posture analysis unit 130 of the present invention will be described in detail later with reference to FIGS. 4 to 8.

Subsequently, the data receiving unit 110 receives exercise images for the exercise posture from the student terminal 200 and the instructor terminal 300, and the exercise posture analysis unit 130 preprocesses each exercise image to change a plurality of joint positions. An input sequence for can be created (S130).

Next, the exercise posture analysis unit 130 may generate posture analysis information for the exercise image by applying the generated input sequence to the posture analysis module 133 (S140).

Next, the lecture recommendation unit 140 may calculate the posture accuracy for each part based on the posture analysis information generated by the exercise posture analysis unit 130 (S150).

Subsequently, the lecture recommendation unit 140 may derive a recommended instructor based on the calculated posture accuracy and the preset weight of the part (S160). Additionally, the lecture recommendation unit 140 may extract a list of recommended lectures corresponding to the recommended instructor (S170).

Next, the data receiver 110 may provide the list of recommended lectures generated by the lecture recommendation unit 140 to the student terminal 200 and receive course registration information (S180).

Hereinafter, the preprocessing and learning operations of the exercise posture analysis unit 130 according to some embodiments of the present invention will be described in detail.

FIG. 5 is a block diagram for explaining the preprocessing module of FIG. 4 in detail. FIG. 6 is a flowchart for explaining the operation of the preprocessing module of FIG. 5. FIG. 7 is a diagram for explaining the joint recognition module of FIG. 5. FIG. 8 is a diagram for explaining the sequence generation module of FIG. 5.

Referring to FIGS. 4 and 5 , the exercise posture analysis unit 130 may include a preprocessing module 131 and a posture analysis module 133.

At this time, the preprocessing module 131 may include a joint recognition module 131a, a joint vector conversion module 131b, and a sequence generation module 131c.

The exercise posture analysis unit 130 may load the normal exercise image stored in the database unit 120 and apply it to the preprocessing module 131.

Specifically, referring to FIGS. 5 to 7, the joint recognition module 131a of the preprocessing module 131 recognizes the position (b) of the main joint in each frame of the plurality of frames (FS) included in the normal motion image. You can do it (S111).

Additionally, the joint recognition module 131a can recognize the positions (b) of the recognized joints by grouping them. For example, as shown in e1 of Figure 7, the position of the recognized joint (B) is the upper body (UB), the lower body (LB), the arm (A) included in the upper body (UB), the lower body (LB), etc. Can be grouped into .

Subsequently, the joint recognition module 131a may apply joint recognition information, including the type of recognized joint and the connection relationship between the positions of each recognized joint, to each frame (S112).

Here, with reference to FIG. 7 , the joint recognition information may include information about the connection relationship between the type of recognized joint and the position (b) of each joint. For example, the joint recognition information may include each joint type of the recognized first joint (b1) and the second joint (b2) and relationship information between the first joint (b1) and the second joint (b2). Specifically, when the type of the first joint (b1) is the ankle and the type of the second joint (b2) is the knee, the first joint (b1) and the second joint (b2) are connected by the tibia and fibula. Relationship information about the first joint and the second joint can be generated by grouping them into groups that affect each other's movements.

Additionally, when the joint recognition module 131a is grouped for at least one recognized joint, each recognized joint may be grouped as an overlap. For example, the joint recognition module 131a may generate a first grouping by grouping the first joint (b1), the second joint (b2), and the third joint (b3) into one, and the third joint (b3) may be included in a second grouping that is different from the first grouping as a hip joint.

Additionally, grouping may be done according to criteria set according to the type of exercise. However, this description is merely illustrative, and the embodiments are not limited to this description.

In still other embodiments of the present invention, the joint recognition module 131a may receive a motion image including a plurality of frames FS and a corresponding exercise type. The joint recognition module 131a generates a recognition joint list corresponding to the received exercise type, recognizes the position (b) of the joint in each frame based on the generated recognition joint list, and generates a plurality of joint frames (JFS). can do. The type and number of joints to be recognized for each type of exercise may be the same or different. The embodiments of the present invention are not limited thereto. In other embodiments of the present invention, the joint recognition module may be implemented in various modifications.

Subsequently, the joint recognition module 131a may transmit a plurality of joint frames (JFS) to which joint recognition information is applied to each frame to the joint vector conversion module 131b.

A plurality of joint frames (JFS) may be a frame in which each frame and joint recognition information are integrated. Additionally, a plurality of joint frames (JFS) may be generated for each grouping according to the grouping criteria of the joint recognition information. However, this description is merely illustrative, and the embodiments are not limited to this description.

Next, the joint vector conversion module 131b may generate a plurality of joint vectors based on changes in joint recognition information included in the frame (S113).

For example, the joint vector conversion module 131b may generate a plurality of joint vectors (JFS') by converting the dimensions of a plurality of joint frames (JFS) through a plurality of layer sets. Each layer set may have a different dimension, and the last layer set may have a preset dimension size.

For example, when the plurality of jointed frames (JFS) includes a first frame (F1), a second frame (F2), a third frame (F3), and a fourth frame (F4), the The first frame (F1) having dimensions may be converted to dimensions of 100 The final converted dimension can be converted to a 1 x 1 x 1000 dimension. The dimension that is finally determined may be a preset dimension size. However, this description is merely illustrative, and the embodiments are not limited to this description.

In addition, in other embodiments of the present invention, the plurality of joint vectors (JFS') are grouped based on joint recognition information included in one frame of the plurality of joint frames (JFS), and are grouped into a plurality of joint vectors per frame. A joint vector (JFS') may be created. However, this is only an example, and the embodiments are not limited to this description.

Next, referring to FIGS. 5, 6, and 8, the sequence generation module 131c arranges the plurality of joint vectors (JFS') generated by the joint vector conversion module 131b in the order of consecutive frames to create an exercise image. An input sequence for can be generated (S114). Additionally, in another embodiment, the sequence generation module 131c may generate an input sequence by arranging a plurality of joint vectors (JFS') in the order of consecutive frames and performing a mutual inner product between the plurality of joint vectors (JFS'). there is.

Additionally, the sequence generation module 131c may generate an input sequence by applying a window (W) of a predetermined unit to a plurality of joint vectors (JFS'). However, this description is merely illustrative, and the embodiments are not limited to this description.

Next, referring to FIG. 4, the exercise posture analysis unit 130 applies the learning input sequence and learning posture analysis information (PI) generated by applying the normal exercise image to the preprocessing module 131 to the posture analysis module 133. Thus, the posture analysis module 133 can be learned.

For example, the exercise posture analysis unit 130 may perform supervised learning for the posture analysis module 133 based on learning analysis information (PI) with the correct answer for a normal exercise image and a learning input sequence.

Here, learning posture analysis information (PI) is posture analysis information for a normal exercise image and can be generated by dividing the accuracy for each body part. For example, learning posture analysis information (PI) may include upper body motion accuracy, arm motion accuracy, lower body motion accuracy, leg motion accuracy, etc. For example, the posture analysis information is a probability value, and each motion accuracy included in the learning posture analysis information (PI) may be close to 1 or may be 1. Additionally, the type of motion accuracy included in the learning posture analysis information (PI) may be configured differently depending on the type of exercise. However, this description is merely illustrative, and the embodiments are not limited to this description.

Hereinafter, operations performed by the exercise posture analysis unit 130 according to some embodiments of the present invention will be described in detail with reference to FIGS. 9 and 10.

FIG. 9 is a block diagram for explaining in detail the execution steps of the exercise posture analysis unit of FIG. 2. FIG. 10 is a block diagram showing the posture analysis module of FIG. 9.

Referring to FIG. 9, the pre-processing module 131 may receive the user exercise image, which is the student's first exercise image, and generate a first input sequence.

Additionally, the pre-processing module 131 may receive approval of the instructor's second exercise image, which is the instructor's exercise image, and generate a second input sequence.

The posture analysis module 133 may receive the first and second input sequences and generate posture analysis information. Specifically, the posture analysis module 133 may receive a first input sequence and generate first posture analysis information, and may receive a second input sequence and generate second posture analysis information.

At this time, referring to FIG. 10, the posture analysis module 133 may use already known deep learning technology. For example, the posture analysis module 133 may use technologies such as Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), and Graph Neural Network (GNN).

Specifically, CNN is a deep learning model used to extract features based on image or video data and recognize objects included in the image or video data, and has the characteristic of considering the correlation between each pixel of the image.

RNN is a deep learning model generally used to process time series data and is widely used in natural language processing.

GNN is a deep learning model used to derive relationships and interactions between data, and represents an artificial neural network structure implemented by deriving similarities and feature points between modeling data.

Accordingly, the posture analysis module 133 learned from the normal motion image may analyze the posture of the input motion image based on the normal motion image.

Additionally, the lecture recommendation unit 140 may generate a list of recommended lectures based on the first and second posture analysis information generated by the posture analysis module 133. A detailed description of this will be provided later with reference to FIGS. 14 to 16.

Hereinafter, the operations performed by the exercise posture analysis unit 130 according to another embodiment of the present invention will be described in detail with reference to FIGS. 11 and 13.

FIG. 11 is a block diagram showing data flow to explain another embodiment of the posture analysis module of FIG. 9. FIG. 12 is a flowchart for explaining the posture analysis module of FIG. 11 in detail. FIG. 13 is a block diagram to explain some examples of the autoencoder of FIG. 11.

Referring to FIG. 11, the posture analysis module 133 may include an autoencoder 133a including a neural network and an abnormality determination unit 133b that determines whether the input sequence is abnormal.

Specifically, referring to FIGS. 11 and 12 , the exercise posture analysis unit 130 may learn based on a learning input sequence generated by applying a normal exercise image (SE) to the preprocessing module 131.

Specifically, the exercise posture analysis unit 130 may convert a pre-stored normal exercise image into a learning input sequence (SIS) using the preprocessing module 131 (S210).

Next, the exercise posture analysis unit 130 may generate a learning output sequence (SIS) by applying the learning input sequence to the LSTM (Long Short Term Memory) autoencoder 133a (S220).

Next, the abnormality determination unit 133b calculates an error value (E), which is the difference value between the learning input sequence (SIS) and the learning output sequence (SIS') (S230), and the exercise posture analysis unit 130 calculates the error value ( Repeated learning can be determined based on E) (S240). For example, if the error value E is smaller than a preset reference value, the exercise posture analysis unit 130 may end learning. Conversely, when the error value E is greater than a preset reference value, the exercise posture analysis unit 130 may repeatedly perform exercise posture pattern learning to reduce the error value E. At this time, the exercise posture analysis unit 130 can repeatedly learn the same loaded normal exercise image and perform learning by newly loading another normal exercise image included in the normal exercise dataset. Accordingly, the exercise posture analysis unit 130 can increase posture analysis accuracy through repetitive learning.

Subsequently, the data receiver 110 receives the first exercise image (UE) for the exercise posture from the student terminal 200, and the preprocessing module 131 of the exercise posture analysis unit 130 receives the first exercise image (UE) can be preprocessed to generate a first input sequence (1IS) for multiple joint position changes (S250).

Next, the exercise posture analysis unit 130 may apply the first input sequence (1IS) to the LSTM autoencoder (133a) to generate the first output sequence (1IS') (S260).

Next, the abnormality determination unit 133b may calculate the difference between the first input sequence 1IS and the first output sequence 1IS' to generate first posture analysis information for the first motion image (S270). For example, the abnormality determination unit 133b indexes and extracts the difference between the first input sequence (1IS) and the first output sequence (1IS'), and generates a student analysis image in which the extracted portion is displayed on the student's exercise image. can do.

In an additional embodiment, the abnormality determination unit 133b extracts the difference between the first input sequence 1IS and the first output sequence 1IS' and generates joint feedback information, which is joint recognition information, corresponding to the extracted part. You can.

The first posture analysis information may include exercise achievement level, achievement probability, student analysis image, and joint feedback information.

Additionally, the exercise posture analysis unit 130 may generate second posture analysis information by performing the same operation on the second exercise image, which is the instructor's exercise image, as the first exercise image. For example, the exercise posture analysis unit 130 may generate second posture analysis information based on a second input sequence (2IS) and a second output sequence (2IS') associated with the instructor.

The exercise posture analysis unit 130 determines whether the instructor is suitable as an instructor by comparing the error value (E), which is the difference between the second input sequence (2IS) and the second output sequence (2IS'), with a preset instructor standard value. Instructor grade can be calculated.

In addition, instructor suitability and instructor rating can be calculated separately according to exercise type and body part. However, this explanation is merely illustrative and may be configured with some modifications.

Additionally, the LSTM autoencoder 133a can be divided into a student autoencoder that receives a first input sequence associated with the student and an instructor autoencoder that receives a second input sequence associated with the instructor, and a student autoencoder and an instructor autoencoder. may be an autoencoder. However, this explanation is illustrative and may be configured with some modifications.

Here, referring to FIG. 13, the LSTM autoencoder 133a includes an autoencoder including an encoder (EN) and a decoder (DC), and can make output data identical to input data.

Specifically, the LSTM autoencoder compresses the data input through the encoder (EN) to reduce the data, and uses a decoder (DC) to convert the reduced data to the same size as the data input from the encoder (EN) and output it. I do it.

For example, the encoder (EN) reduces the dimension of a plurality of input sequences and outputs an encoded feature value (EF), and the encoded feature value (EF) output from the encoder (EN) is transmitted to the decoder (DC). It is passed repeatedly. At this time, the encoded feature value (EF) may be composed of a vector value of a certain size.

Additionally, when a plurality of input sequences are sequentially applied to the input terminal of the encoder EN, the cell status of the LSTM autoencoder constituting the encoder EN can be updated. The updated cell state may be passed to the decoder (DC).

Subsequently, the decoder (DC) generates an output sequence based on the feature value (EF) encoded using the cell state (ES) received from the encoder (EN). The output sequence may be formed to have the same size as the input sequence (IS). The output sequence output from the decoder (DC) is transmitted to the abnormality determination unit 133b to analyze the exercise posture for the exercise image.

Additionally, the LSTM autoencoder can be composed of RNN, and can be composed by adding artificial neural networks such as CNN, DNN, and GNN.

However, the autoencoder of the present invention is not limited to the structure of the autoencoder described with reference to FIG. 13, and it is obvious to those skilled in the art that the autoencoder can be implemented in various modifications.

In an additional embodiment, the posture analysis module 133 may perform machine learning using a neural network on the input sequence and output sequence. The posture analysis module 133 may generate posture analysis information by performing machine learning on the received input sequence and output sequence. For example, the posture analysis module 133 performs machine learning on the first input sequence and the first output sequence associated with the student to generate one of a probability value related to posture analysis and a Gradient-weighted CAM (Grad-CAM) image. can be created. The first posture analysis information generated based on the first input sequence and the first output sequence may include a probability value generated by machine learning and Grad-CAM. The first posture analysis information including Grad-CAM may be provided to the student terminal 200 corresponding to the first posture analysis information by displaying an area where an intuitively incorrect exercise posture is taken. Therefore, even students who are not experts can easily recognize the exercise posture and body part to work on intuitively based on the first posture analysis information.

Specifically, the posture analysis module 133 may calculate a CAM (Class Activation Map) in the last layer of the neural network. In other words, weights for analyzing exercise posture can be calculated and multiple heat maps (HeatMap) can be generated for each node. The method for deriving the Grad CAM image is already known, so detailed explanation will be omitted. Additionally, it is obvious to those skilled in the art that in the method of generating a Grad CAM image, various known methods for displaying areas different from the correct exercise posture can be applied.

Hereinafter, the lecture recommendation unit 140 according to some embodiments of the present invention will be described with reference to FIGS. 14 to 16.

Figure 14 is a block diagram for explaining in detail the operation of the lecture recommendation unit of Figure 2. Figure 15 is a flowchart for explaining in detail the operation of the lecture recommendation unit of Figure 14. Figure 16 is a diagram for explaining in detail the method of deriving the recommended instructor of Figure 15.

Referring to FIG. 14, the lecture recommendation unit 140 may derive a list of recommended lectures based on two or more parameters among posture analysis information, student preference, time schedule, location proximity, and lecture cost. The time schedule includes the time schedule of students, lectures, and instructors. Location proximity refers to the student's location proximity to the lecture location.

Specifically, referring to FIGS. 15 and 16, the lecture recommendation unit 140 provides posture accuracy for each part ( cp) can be calculated (S151).

For example, posture accuracy (cp) may be a value calculated as a percentage of posture analysis information, which is a probability value. Additionally, posture accuracy (cp) can be calculated for each body part, and the considered body parts can be configured differently depending on the type of exercise.

Next, the lecture recommendation unit 140 may generate first matching instructor information by deriving an instructor with the maximum specific posture accuracy based on the calculated posture accuracy (cp) (S161). Here, the first matching instructor information is information about the recommended instructor recommended for each part.

Additionally, posture accuracy can be calculated for each body part or for each type of exercise. For example, posture accuracy for the squat exercise may be calculated for the first part, the second part, the third part, and the fourth part, respectively.

At this time, the lecture recommendation unit 140 may derive an instructor with the maximum posture accuracy for each body part. Referring to FIG. 16, the user's posture accuracy in the first region is 20, and the maximum posture accuracy within the first region is 90, corresponding to instructor B. Therefore, the instructor matching the first part can be derived as instructor B.

Additionally, in another embodiment, the lecture recommendation unit 140 may exclude from the matching instructors an instructor whose posture accuracy is lower than that of the student. For example, the student's posture accuracy for part 4 is 35, and the matching instructor is instructor C. At this time, instructor D's posture accuracy for the fourth part is 20, which is lower than the student's posture accuracy. Therefore, even if instructor D has the maximum posture accuracy for parts other than the fourth part, he is judged to have failed for the fourth part and cannot be selected as a matching instructor. However, this description is merely illustrative, and the embodiments are not limited to this description. It may be configured with some modifications.

Subsequently, the lecture recommendation unit 140 may derive a recommended instructor suitable for the user based on the first matching instructor information and the preset weight of the portion (S163). At this time, the lecture recommendation unit 140 may calculate the score for each matched instructor by calculating the posture accuracy and part weight of the matched instructor for each body part. The lecture recommendation unit 140 may select an instructor corresponding to the maximum value among the calculated scores as a recommended instructor, and may select one or two more recommended instructors as spares.

Next, the course recommendation unit 140 may extract the first recommended course list corresponding to the recommended instructor from the course information (S171). Specifically, the lecture recommendation unit 140 may generate a first recommended lecture list by extracting the list of lectures taught by the recommending instructor from the lecture information stored in the database unit 120.

Through this, the lecture recommendation unit 140 can provide the student terminal 200 with information about a lecture given by an instructor whose skills have been verified regarding the student's incorrect exercise posture. However, this explanation is merely illustrative and may be configured with some modifications.

Subsequently, the lecture recommendation unit 140 may derive a list of recommended lectures based on two or more parameters among the student's preference, time schedule, location proximity, and cost, and the first recommended lecture list (S173). For example, the lecture recommendation unit 140 considers the student's preference, time schedule, and location proximity in the first recommended lecture list so that the student terminal 200 can take each lecture included in the first recommended lecture list. A list of recommended lectures can be derived by determining whether it is possible to move to the location.

The student's preference is information received from the student terminal 200 and may include the exercise class the student wants to take, the exercise posture the student wants to receive guidance on, and matters to be considered when selecting the course.

The time schedule may include a time schedule and a lecture time schedule that takes into account blank time when the student has no schedule. Additionally, the time schedule may include the student's gap time and location information of the schedule planned before and after the gap time.

Location proximity refers to the lecture location corresponding to each lecture included in the recommended lecture list, the scheduled lecture location received in advance from the student terminal 200, the location information of the student terminal 200, and the location information of the schedule planned before and after the student's gap time. This is information that determines whether the student terminal 200 can be moved to the lecture location or scheduled lecture location in response to the lecture time schedule based on this.

For example, the lecture recommendation unit 140 recommends the student terminal 200 based on the lecture location corresponding to each lecture, the scheduled lecture location, the location information of the student terminal 200, and the location information of the schedule planned before and after the student's gap time. The travel time to move to the lecture location or the scheduled lecture can be calculated from the location information of the schedule planned before and after the student's gap time. If the lecture is a non-face-to-face or online lecture, the lecture location may be a location set in advance by the student.

The lecture recommendation unit 140 may calculate the first travel time to each lecture location from the student's location information just before the gap time. The lecture recommendation unit 140 may add the first travel time from the start of the blank time and compare it with the lecture start time. If the time added to the first travel time from the time of the blank time is later than the start time of the first lecture, the lecture recommendation unit 140 may remove information about the first lecture from the list of first recommended lectures.

Additionally, the lecture recommendation unit 140 may calculate a second travel time to move from the lecture location of the first lecture to the location information of the planned schedule after the blank time. Therefore, the time added to the first travel time from the start of the gap time is earlier than the start time of the first lecture, and the time added to the second travel time from the end of the first lecture of the first lecture is later than the schedule planned after the gap time. In this case, the lecture recommendation unit 140 may remove information about the first lecture from the first recommended lecture list. However, this explanation is merely illustrative, and the method of extracting lectures that students can take based on the environment in which they can take the course may be configured with some modification.

Additionally, the lecture recommendation unit 140 may derive a list of recommended lectures by additionally considering cost. The cost may be the amount received as a limit from the student terminal 200.

Additionally, the student terminal 200 may transmit course registration information generated based on the provided recommended course list to the data receiver 110.

Hereinafter, the place reservation unit 150 according to some embodiments of the present invention will be described with reference to FIGS. 17 to 19.

FIG. 17 is a block diagram for explaining the venue reservation unit of FIG. 2 in detail. FIG. 18 is a flowchart for explaining in detail the operation of the place reservation unit of FIG. 17. FIG. 19 is a diagram for explaining a method of generating the lecture facility list of FIG. 17.

Referring to FIG. 17, the venue reservation unit 150 may include a first lecture facility extraction module 151, a second lecture facility extraction module 153, and a final lecture facility extraction module 155.

Referring to FIGS. 17 and 18 , the first lecture facility extraction module 151 may load the scheduled lecture location, preset distance, and affiliated lecture facility information stored in advance in the database unit (S310). At this time, the first lecture facility extraction module 151 may load data from the database unit after the lecture recommendation unit 140 extracts the list of recommended lectures or receives course registration information from the student terminal 200. . The pre-stored course schedule is information received from the student terminal 200 and is location information about the course location desired by the student.

Subsequently, the first lecture facility extraction module 151 may extract a list of first lecture facilities available for rent located within a preset distance based on the preset course attendance location from the lecture facility information (S320). Lecture facility information is information about affiliated lecture facilities, and is information about places where exercise classes can be held or exercise facilities can be rented.

Specifically, referring to FIG. 19, the first lecture facility extraction module 151 may extract facility information about lecture facilities located within a preset distance (r) based on the preset course attendance destination (C). For example, referring to FIG. 19, the first lecture facility extraction module 151 can extract lecture facilities ①, ②, and ③ located within a preset distance (r) based on the lecture scheduled location (C).

The second lecture facility extraction module 153 may extract a second lecture facility list suitable for the lecture information based on the first lecture facility list and the lecture information taken by the student (S330). For example, if the type of exercise corresponding to the lecture information is ballet, the second lecture facility extraction module 153 extracts a place where ballet can be performed from the first lecture facility list and extracts a second lecture facility list.

The final lecture facility extraction module 155 may generate a final lecture facility list based on the rental status information of each lecture facility included in the second lecture facility list (S340). For example, when the second lecture facility list is extracted as a place where ballet can be performed, the final lecture facility extraction module 155 determines whether the lecture facility included in the second lecture facility list can be rented and selects the final lecture facility. Create a list. Specifically, the final lecture facility extraction module 155 may determine whether any of the lecture facilities included in the second lecture facility list can be rented at a time corresponding to the lecture information.

Hereinafter, the student management unit 160 according to some embodiments of the present invention will be described with reference to FIG. 20.

FIG. 20 is a block diagram for explaining the student management unit of FIG. 2 in detail.

Referring to FIG. 20, the student management unit 160 may include a course management module 161 and an exercise feedback module 163.

The course management module 161 generates schedule notification information based on at least one of the course history, the student's schedule information corresponding to the course history, lecture information, traffic information, and weather information, and the student terminal corresponding to the course history ( 200), schedule notification information can be provided. Course attendance history is accumulated course registration information.

The schedule notification information includes lecture information for future lectures based on the time provided to the student terminal 200, departure time based on at least one of traffic information and weather information, recommended means of transportation, and Recommended clothing information may be included.

Additionally, the schedule notification information may include the remaining number of lectures and scheduled lecture content based on course registration information. However, this description is merely illustrative, and the embodiments are not limited to this description.

The exercise feedback module 163 loads the first posture analysis information and Inbody information associated with the student from the database unit 120, generates exercise feedback information based on the first posture analysis information and Inbody information, and provides the student terminal 200 can be provided.

The exercise feedback information may include an exercise achievement rate based on the student's accumulated first exercise posture analysis information, an improvement rate regarding the degree of improvement in the exercise posture according to the lecture, and guideline information for improving the exercise posture. Additionally, exercise feedback information may include recommended exercise types and lecture information based on InBody information. The exercise feedback information may be configured with some modifications.

Additionally, the exercise feedback module 163 may compare and analyze the change value of the InBody information and the first posture analysis information to calculate change information for body changes and posture changes according to the lecture.

Additionally, the student management unit 160 may rank students for each exercise type based on the first posture analysis information of each of the plurality of students. Through this, the server 100 can provide compensation benefits according to competition between students based on predetermined compensation guidelines to each student through the student terminal 200 corresponding to each student.

The student management unit 160 may provide corresponding data to the student terminal 200 each time each piece of information is updated or created, and may be configured with some modification.

Figure 21 is a diagram for explaining the hardware implementation of a system that performs a method for recommending and managing exercise classes according to some embodiments of the present invention.

Referring to FIG. 21, the server 100 according to some embodiments of the present invention may be implemented as an electronic device 1000. The electronic device 1000 may include a controller 1010, an input/output device 1020, I/O, a memory device 1030, an interface 1040, and a bus 1050. . The controller 1010, input/output device 1020, memory device 1030, and/or interface 1040 may be coupled to each other through a bus 1050. At this time, the bus 1050 corresponds to a path through which data is moved.

Specifically, the controller 1010 includes a Central Processing Unit (CPU), Micro Processor Unit (MPU), Micro Controller Unit (MCU), Graphic Processing Unit (GPU), microprocessor, digital signal processor, microcontroller, and application processor (AP). , application processor) and logic elements capable of performing similar functions.

The input/output device 1020 may include at least one of a keypad, a keyboard, a touch screen, and a display device.

The memory device 1030 may store data and/or programs.

The interface 1040 may perform a function of transmitting data to or receiving data from a communication network. Interface 1040 may be wired or wireless. For example, the interface 1040 may include an antenna or a wired or wireless transceiver. Although not shown, the memory device 1030 is an operating memory for improving the operation of the controller 1010 and may further include high-speed DRAM and/or SRAM. The memory device 1030 may store programs or applications therein.

The server 100 according to embodiments of the present invention may be a system formed by connecting a plurality of electronic devices 1000 to each other through a network. In this case, each module or combination of modules may be implemented as the electronic device 1000. However, this embodiment is not limited to this.

Additionally, the server 100 is a workstation, a data center, an internet data center (IDC), a direct attached storage (DAS) system, a storage area network (SAN) system, and a network attached storage (NAS). It may be implemented as at least one of a system, a RAID (redundant array of inexpensive disks, or redundant array of independent disks) system, and an EDMS (Electronic Document Management) system, but the present embodiment is not limited thereto.

Additionally, the server 100 may transmit data to the student terminal 200 and the instructor terminal 300 through the network. Networks may include networks based on wired Internet technology, wireless Internet technology, and short-distance communication technology. Wired Internet technology may include, for example, at least one of a local area network (LAN) and a wide area network (WAN).

Wireless Internet technologies include, for example, Wireless LAN (WLAN), DMNA (Digital Living Network Alliance), Wibro (Wireless Broadband), Wimax (World Interoperability for Microwave Access: Wimax), and HSDPA (High Speed Downlink Packet). Access), HSUPA (High Speed Uplink Packet Access), IEEE 802.16, Long Term Evolution (LTE), LTE-A (Long Term Evolution-Advanced), Wireless Mobile Broadband Service (WMBS) and 5G NR (New Radio) technology. However, this embodiment is not limited to this.

Short-range communication technologies include, for example, Bluetooth, Radio Frequency Identification (RFID), Infrared Data Association (IrDA), Ultra-Wideband (UWB), ZigBee, and Near Field Communication. At least one of NFC), Ultrasound Communication (USC), Visible Light Communication (VLC), Wi-Fi, Wi-Fi Direct, and 5G NR (New Radio) may include. However, this embodiment is not limited to this.

The server 100 that communicates through a network can comply with technical standards and standard communication methods for mobile communication. For example, standard communication methods include GSM (Global System for Mobile communication), CDMA (Code Division Multi Access), CDMA2000 (Code Division Multi Access 2000), and EV-DO (Enhanced Voice-Data Optimized or Enhanced Voice-Data Only). , at least one of Wideband CDMA (WCDMA), High Speed Downlink Packet Access (HSDPA), High Speed Uplink Packet Access (HSUPA), Long Term Evolution (LTE), Long Term Evolution-Advanced (LTEA), and 5G New Radio (NR) may include. However, this embodiment is not limited to this.

The above description is merely an illustrative explanation of the technical idea of the present embodiment, and those skilled in the art will be able to make various modifications and variations without departing from the essential characteristics of the present embodiment. Accordingly, the present embodiments are not intended to limit the technical idea of the present embodiment, but rather to explain it, and the scope of the technical idea of the present embodiment is not limited by these examples. The scope of protection of this embodiment should be interpreted in accordance with the claims below, and all technical ideas within the equivalent scope should be interpreted as being included in the scope of rights of this embodiment.

Claims (7)

In the exercise recommendation and management method performed on a server that operates in connection with the student terminal used by the student and the instructor terminal used by the instructor,
Converting a normal exercise image pre-stored in the server into an input sequence for learning using a pre-processing module;
learning a posture analysis module based on the learning input sequence and the corresponding learning posture analysis information;
An exercise image is received from the student terminal and the instructor terminal, and the exercise image is preprocessed using the preprocessing module to generate an input sequence for a plurality of joint position changes, wherein the exercise image is the student's first exercise image. and including a second exercise video of the instructor;
The input sequence is applied to the posture analysis module to generate posture analysis information for the exercise image, wherein the posture analysis information includes first posture analysis information associated with the first exercise image, and posture analysis information associated with the second exercise image. 2 including posture analysis information;
Comparing the first posture analysis information and the second posture analysis information to calculate posture accuracy for each part, and deriving a list of recommended lectures using the calculated posture accuracy; and
Comprising the step of providing the recommended lecture list to the student terminal,
The step of deriving the list of recommended lectures is,
(a) Based on the posture accuracy for each part, the instructor with the maximum posture accuracy for each part is derived, and first matching instructor information is generated by matching the student and the instructor for each part, but for the first part generating the first matching instructor information by excluding instructors whose posture accuracy for the first part is lower than the user's posture accuracy for the first part;
(b) deriving a recommended instructor based on the first matching instructor information and preset region weights;
(c) extracting lectures corresponding to the recommended instructor from pre-stored lecture information and deriving the recommended lecture list;
(d) extracting a first recommended lecture list corresponding to the recommended instructor from pre-stored lecture information; and
(e) deriving the recommended lecture list based on the first recommended lecture list and at least one of the time schedule of the student and the instructor and the location proximity to the lecture location of the student,
In step (e),
calculating a first travel time to travel to the lecture location of each lecture included in the first recommended lecture list based on the student's location information just before the gap time;
calculating a second travel time to move from the lecture location of each lecture to a scheduled location after the blank time; and
Based on whether the learner can arrive at the lecture location corresponding to each lecture by the start time according to the first travel time and whether the learner can arrive at the planned schedule after the blank time according to the second travel time An exercise recommendation and management method including the step of deriving the list of recommended lectures.
According to claim 1,
The step of generating the input sequence is,
Recognizing the position of a major joint in each frame of a plurality of frames included in the exercise image;
Applying joint recognition information about the positions of the main joints to each frame;
generating a plurality of joint vectors based on changes in the joint recognition information; and
An exercise recommendation and management method further comprising generating the input sequence by arranging the plurality of joint vectors in sequential frame order.
According to claim 1,
The posture analysis module includes an autoencoder including a neural network and an abnormality determination unit that determines whether the input sequence is abnormal,
The step of generating the posture analysis information is,
Generating an output sequence by applying the input sequence to the autoencoder; and
Exercise recommendation and management method further comprising generating posture analysis information based on the difference between the input sequence and the output sequence in the abnormality determination unit.
delete According to claim 1,
An exercise recommendation and management method wherein the part weight is a preset value for at least one body part according to each exercise type.
According to claim 1,
The step of deriving the list of recommended lectures is,
An exercise recommendation and management method further comprising deriving the recommended lecture list based on one or more parameters of the posture accuracy, student preference received from the student terminal, and lecture cost.
According to claim 1,
Loading pre-stored course schedule location, pre-set distance, and affiliated course facility information from a database unit included in the server;
extracting a list of first lecture facilities available for rent located within the preset distance based on the scheduled lecture location;
extracting a second lecture facility list suitable for the lecture information from the first lecture facility list based on the first lecture facility list and information on the lectures taken by the student; and
Exercise recommendation and management method further comprising generating a final lecture facility list based on rental status information of each lecture facility included in the second lecture facility list.

Images