System and method for evaluating motion imitation

Abstract

The present invention relates to a system and a method for evaluating motion imitation. According to an embodiment of the present invention, a user's motion performance accuracy is evaluated in real time and the result is converted into a score and provided to the user, thereby promoting motivation for motion performance and improving physical activity. In addition, according to the present invention, a user's motion performance state can be more accurately and quickly analyzed on the basis of image analysis and joint coordinate extraction technology, thereby stably evaluating a motion regardless of physical conditions and motion environments of various users such as seniors and infants.

Description

Action Imitation Evaluation System and Method

[0001] The present invention relates to a motion-following evaluation system and method.

[0002] The present application claims priority based on Korean Patent Application No. 10-2024-0159321 filed on November 11, 2024 and Korean Patent Application No. 10-2025-0027561 filed on March 4, 2025, and all contents described in the specifications and drawings of said applications are incorporated by reference into the present application.

[0003] Generally, imitation-based movement learning methods are utilized to promote the health of seniors and improve the motor skills of young children.

[0004] This motion-following method is useful for promoting physical activity by inducing users to mimic exercise movements or specific behaviors that appear on the screen.

[0005] For example, in the case of seniors, having them follow light gymnastic movements or health exercises can increase physical activity and help form a consistent exercise habit.

[0006] In addition, for infants, educational methods are used to improve body part recognition and hand-eye coordination skills by encouraging them to point to specific parts such as the eyes, nose, and mouth, or to follow simple movements.

[0007] However, conventional motion-following methods have limitations in providing accurate feedback because it is difficult to quantitatively evaluate how accurately the user followed the action during the process, and they often rely on subjective observation.

[0008] Furthermore, there is a lack of technical means to analyze, score, and provide feedback on the user's performance status in real time.

[0009] Therefore, there is a need to develop technology capable of providing objective feedback by analyzing user movements in real time, evaluating the accuracy of imitation, and quantifying it.

[0010] The technical objective of the present invention is to provide a motion imitation evaluation system and method capable of promoting motivation for motion performance and improving physical activity by evaluating the accuracy of a user's motion performance in real time, converting the results into a score, and providing it to the user.

[0011] Furthermore, the present invention aims to enable more accurate and rapid analysis of a user's movement performance status based on image analysis and joint coordinate extraction technology, thereby allowing stable movement evaluation regardless of the physical conditions and movement environments of various users, such as seniors and infants.

[0012] The objectives of the present invention are not limited thereto, and other unmentioned objectives will be clearly understood by a person skilled in the art from the description below.

[0013] An embodiment of the present invention for achieving the above technical objectives provides a motion-following evaluation system for evaluating the state of a user's motion performance, comprising: an image output unit that outputs motion guide video information to be provided to the user; an image capturing unit that captures the user's motion to acquire motion performance video information; a coordinate extraction unit that analyzes the motion performance video information to set coordinate information for the user's joint parts; a motion evaluation unit that performs frame matching between the motion guide video information and the motion performance video information based on the coordinate information to calculate a similarity; and a result output unit that displays the user's motion performance result scored according to the similarity.

[0014] Additionally, the coordinate extraction unit comprises: (a) a process of detecting the user in each frame of the motion execution video information, setting a bounding box, and identifying the location by calculating the center coordinate information of the user based on the set bounding box; (b) a process of extracting coordinate information for the detected joint parts of the user using a neural network-based human joint recognition model; (c) a process of selecting among the extracted coordinate information that has a reliability greater than or equal to a preset threshold; and (d) a process of setting the selected coordinate information as coordinate information by performing vector normalization based on the user's shoulder width or body height to correct errors according to the user's body size and shooting environment. The present invention provides a motion imitation evaluation system characterized by including: (a) a process of (b

[0015] In addition, the motion evaluation unit comprises: (a) a process of applying a dynamic time warping (DTW) technique to correct the difference in the speed of the user's motion execution, calculating the Euclidean distance between the reference motion frame sequence of the motion guide video information and the user motion frame sequence of the motion execution video information, generating an Accumulated Cost Matrix based on the calculated distance value, extracting an optimal frame alignment path (warping path) having the minimum cost from the Accumulated Cost Matrix, and aligning the frames of the user motion and the reference motion based on the extracted alignment path; (b) a process of matching identical time frames of the motion guide video information and the motion execution video information using the coordinate information based on the result of applying the dynamic time warping (DTW) technique; (c) a process of calculating the Euclidean distance between the user's joint coordinates and the pre-set reference joint coordinates (coordinate information of the reference joint extracted from the motion guide video information) in each matched frame, and measuring the similarity of each joint; and (d) a process of calculating the similarity of the entire motion by averaging the Euclidean distance values ​​calculated in each matched frame; the present invention provides a motion imitation evaluation system characterized by including

[0016] Meanwhile, another embodiment of the present invention provides a motion-following evaluation method for evaluating a user's motion performance status, comprising: (a) a step in which an image output unit of a motion-following evaluation system outputs motion guide image information to be provided to the user; (b) a step in which an image capturing unit of the motion-following evaluation system captures the user's motion to acquire motion performance image information; (c) a step in which a coordinate extraction unit of the motion-following evaluation system analyzes the motion performance image information to set coordinate information for the user's joint parts; (d) a step in which a motion evaluation unit of the motion-following evaluation system performs frame matching between the motion guide image information and the motion performance image information based on the coordinate information to calculate a similarity; and (e) a step in which a result output unit of the motion-following evaluation system displays the user's motion performance result scored according to the similarity.

[0017] Additionally, the above step (c) comprises: (c-1) detecting the user in each frame of the motion execution video information to set a bounding box, and calculating the center coordinate information of the user based on the set bounding box to identify the location; (c-2) extracting coordinate information for the detected joint parts of the user using a neural network-based human joint recognition model; (c-3) selecting among the extracted coordinate information that has a reliability greater than or equal to a preset threshold; and (c-4) setting the selected coordinate information as the coordinate information by performing vector normalization based on the user's shoulder width or body height to correct errors according to the user's body size and shooting environment. This provides a method for evaluating motion imitation.

[0018] Additionally, the above step (d) comprises: (d-1) applying a dynamic time warping (DTW) technique to correct the difference in the speed of the user's action execution, calculating the Euclidean distance between the reference action frame sequence of the action guide video information and the user action frame sequence of the action execution video information, generating an Accumulated Cost Matrix based on the calculated distance value, extracting an optimal frame alignment path (warping path) having the minimum cost from the Accumulated Cost Matrix, and aligning the frames of the user action and the reference action based on the extracted alignment path; (d-2) matching identical time frames of the action guide video information and the action execution video information using the coordinate information based on the result of applying the dynamic time warping (DTW) technique; (d-3) calculating the Euclidean distance between the user's joint coordinates and the pre-set reference joint coordinates (coordinate information of the reference joint extracted from the action guide video information) in each matched frame, and measuring the similarity of each joint; and (d-4) a step of calculating the similarity of the entire motion by averaging the Euclidean distance values ​​calculated in each matched frame; the present invention provides a motion imitation evaluation method characterized by including

[0019] According to one embodiment of the present invention, by evaluating the accuracy of a user's movement performance in real time and providing the result as a score to the user, it is possible to promote motivation for movement performance and improvement of physical activity.

[0020] Furthermore, based on image analysis and joint coordinate extraction technology, the present invention enables more accurate and rapid analysis of a user's movement performance status, thereby having the effect of allowing stable movement evaluation regardless of the physical conditions and movement environments of various users, such as seniors and infants.

[0021] FIG. 1 is a block diagram showing an operation-following evaluation system according to an embodiment of the present invention.

[0022] FIGS. 2 to 4 are flowcharts illustrating a method for evaluating operation following according to another embodiment of the present invention.

[0023] [Explanation of the symbol]

[0024] 100: Action Imitation Evaluation System

[0025] 101 : Video Output Section

[0026] 103 : Video Recording Department

[0027] 105 : Coordinate extraction unit

[0028] 107 : Motion Evaluation Department

[0029] 109 : Result Output Section

[0030]

[0031] i1 : Operation guide video information

[0032] i3 : Action execution video information

[0033] i5 : Coordinate information

[0034] Hereinafter, some embodiments of the present invention will be described in detail with reference to the exemplary drawings. It should be noted that in assigning reference numerals to the components of each drawing, the same components are given the same reference numeral whenever possible, even if they are shown in different drawings. Furthermore, in describing the present invention, if it is determined that a detailed description of related known components or functions could obscure the essence of the invention, such detailed description is omitted.

[0035] In addition, terms such as first, second, A, B, (a), (b), etc., may be used when describing the components of the present invention. These terms are intended only to distinguish the components from other components, and the essence, order, or sequence of the components is not limited by the terms. Where it is stated that a component is "connected," "combined," or "connected" to another component, it should be understood that the component may be directly connected or connected to the other component, but that another component may also be "connected," "combined," or "connected" between each component.

[0036] FIG. 1 is a block diagram showing a motion-following evaluation system according to one embodiment of the present invention. FIG. 2 to 4 are flowcharts showing a motion-following evaluation method according to another embodiment of the present invention.

[0037] As illustrated in these drawings, an embodiment of the present invention is a motion-following evaluation system (100) for evaluating the state of a user's motion performance, comprising: a video output unit (101) for outputting motion guide video information (i1) to be provided to a user; a video capture unit (103) for capturing a user's motion and acquiring motion performance video information (i3); a coordinate extraction unit (105) for analyzing motion performance video information (i3) and setting coordinate information (i5) for the user's joint parts; a motion evaluation unit (107) for calculating similarity by performing frame matching between motion guide video information (i1) and motion performance video information (i3) based on the coordinate information (i5); and a result output unit (109) for displaying the user's motion performance result scored according to the similarity.

[0038] Below, each component is described in detail.

[0039] Video output unit (101)

[0040] The video output unit (101) outputs operation guide video information (i1) to be provided to the user.

[0041] The video output unit (101) can be provided as various devices, such as a display device, a projector, a smartphone, a tablet, etc.

[0042]

[0043] Video recording unit (103)

[0044] The video recording unit (103) captures the user's actions and obtains action execution video information (i3).

[0045] The video recording unit (103) can be provided with various devices, for example, including a camera.

[0046]

[0047] Coordinate extraction unit (105)

[0048] The coordinate extraction unit (105) analyzes the motion execution video information (i3) and sets the coordinate information (i5) for the user's joint area.

[0049] More specifically, the coordinate extraction unit (105) can perform a process including: (a) detecting a user in each frame of motion execution video information (i3) to set a bounding box, and calculating the user's center coordinate information based on the set bounding box to identify the location; (b) extracting coordinate information for the detected user's joint parts using a neural network-based human joint recognition model; (c) selecting among the extracted coordinate information that has a reliability greater than or equal to a preset threshold; and (d) setting the selected coordinate information as coordinate information (i5) by performing vector normalization based on the user's shoulder width or body height to correct errors according to the user's body size and shooting environment.

[0050] Explain each process in more detail.

[0051] 1. (a) process

[0052] (a) The process is to detect a user in each frame of the action execution video information (i3), set a bounding box, and calculate the user's center coordinate information based on the set bounding box to identify the location.

[0053] In this process, deep learning-based object detection models (YOLO, Faster R-CNN, MediaPipe, etc.) can be utilized, and the area where the user is located is designated as a bounding box.

[0054] The position (area) occupied by the user on the screen is defined through the coordinates of the bounding box (e.g., (X_min, Y_min, X_max, Y_max)).

[0055] The center coordinate information (center point, Reference Point) is calculated based on the specified bounding box.

[0056] For example, the center coordinate information can be set based on the center coordinates (X_center, Y_center) of the bounding box.

[0057] X_center = (X_min + X_max) / 2

[0058] Y_center = (Y_min + Y_max) / 2

[0059]

[0060] 2. (b) Process

[0061] (b) Process is the process of extracting coordinate information of the detected user's joint parts using a neural network-based human joint recognition model.

[0062] Neural network-based human joint recognition models used in this process may include, for example, OpenPose, MediaPipe Pose, and HRNet (High-Resolution Network).

[0063]

[0064] 3. (c) Process

[0065] (c) The process is to select from the extracted coordinate information those whose confidence level is greater than or equal to a preset threshold.

[0066] 'Reliability' is a value provided when extracting coordinate information about a user's joint parts using a neural network-based human joint recognition model (a probability value indicating how accurately the coordinate information was predicted), and is expressed as a value between 0 and 1.

[0067] Meanwhile, the threshold can be set to 0.6 to 0.8.

[0068]

[0069] 4. (d) Process

[0070] (d) The process is to set the selected coordinate information by performing vector normalization based on the user's shoulder width or body height to correct errors according to the user's body size and shooting environment.

[0071] The coordinates of each joint can be expressed as a 2D vector (X,Y) or a 3D vector (X,Y,Z).

[0072] For example, if the coordinate information is head coordinates: (X_h,Y_h), shoulder coordinates: (X_s,Y_s), wrist coordinates: (X_w,Y_w), and knee coordinates: (X_k,Y_k), all coordinate information can be represented as a single vector as follows.

[0073] V = [(X_h,Y_h), (X_s,Y_s), (X_w,Y_w), (X_k,Y_k)]

[0074] Meanwhile, the user's shoulder width can be calculated using the following formula.

[0075] Formula: Shoulder width (D_shoulder) = sqrt((X_rs - X_ls)² + (Y_rs - Y_ls)²)

[0076] (X_rs, Y_rs: right shoulder coordinates, X_ls, Y_ls: left shoulder coordinates)

[0077] In addition, the user's body height can be calculated using the following formula.

[0078] Formula : Body Height (D_body) = sqrt((X_h - X_f)² + (Y_h - Y_f)²)

[0079] (X_h, Y_h: head coordinates, X_f, Y_f: foot coordinates)

[0080] Meanwhile, vector normalization can be calculated using the following formula.

[0081] Formula: V' = V / D

[0082] V' = Normalized joint coordinate vector

[0083] (V = Original joint coordinate vector, D = Reference distance (shoulder width or body height))

[0084]

[0085] Motion evaluation unit (107)

[0086] The motion evaluation unit (107) calculates similarity by performing frame matching between motion guide video information (i1) and motion execution video information (i3) based on coordinate information (i5).

[0087] More specifically, the motion evaluation unit (107) performs the following steps: (a) applying a dynamic time warping (DTW) technique to correct the difference in the speed of the user's motion execution, calculating the Euclidean distance between the reference motion frame sequence of the motion guide video information (i1) and the user motion frame sequence of the motion execution video information (i3), generating an Accumulated Cost Matrix based on the calculated distance value, extracting an optimal frame alignment path (warping path) with the minimum cost from the Accumulated Cost Matrix, and aligning the frame of the user motion and the frame of the reference motion based on the extracted alignment path; (b) matching the same time frame of the motion guide video information (i1) and the motion execution video information (i3) using coordinate information (i5) based on the result of applying the dynamic time warping (DTW) technique; (c) a process of measuring the similarity of each joint by calculating the Euclidean distance between the user’s joint coordinates and the preset reference joint coordinates (coordinate information of the reference joint extracted from the motion guide video information (i1)) in each matched frame; and (d) a process of calculating the similarity of the entire motion by averaging the Euclidean distance values ​​calculated in each matched frame; may be included.

[0088] Explain each process in more detail.

[0089] 1. (a) process

[0090] (a) The process involves applying a dynamic time warping (DTW) technique to compensate for the difference in the speed of the user's action execution, calculating the Euclidean distance between the reference action frame sequence of the action guide video information (i1) and the user action frame sequence of the action execution video information (i3), generating an Accumulated Cost Matrix based on the calculated distance value, extracting an optimal frame alignment path (warping path) with the minimum cost from the Accumulated Cost Matrix, and aligning the frames of the user action and the reference action based on the extracted alignment path.

[0091] Dynamic Time Warping (DTW) is an algorithm that performs optimal frame alignment by comparing two sequences aligned with different time axes.

[0092] An example of the process for applying the Dynamic Time Warping (DTW) technique is as follows.

[0093] (1) Setting the joint coordinate sequence of the reference motion (motion guide video) and the user motion (motion execution video)

[0094] Standard operation :

[0095] User Actions :

[0096] Here, S_i and U_j are joint coordinate vectors extracted from each frame

[0097] (2) Calculation of Euclidean Distance for each frame

[0098] - Calculate the difference in joint coordinates (Euclidean distance) between two frames S_i and U_j

[0099] - Express similarity in the form of a distance matrix (D(i,j)).

[0100]

[0101] Here, (Xi,Yi) is the coordinate of the i-th frame of the motion guide video, and (Xj,Yj) is the coordinate of the j-th frame of the motion execution video.

[0102] (3) Generating an Accumulated Cost Matrix for Optimal Frame Alignment

[0103] - Find the optimal matching path that minimizes cumulative costs

[0104] - The cumulative cost matrix C(i,j) is defined as a value accumulated including the minimum cost up to the previous frame.

[0105]

[0106] Here,

[0107] C(i-1,j) -> Path coming from above (compare with the previous frame of the reference operation)

[0108] C(i,j-1) -> Path coming from the left (compare with the previous frame of the user action)

[0109] C(i-1,j-1) -> Path coming from the diagonal (optimal matching path of previous frames)

[0110]

[0111] 2. (b) Process

[0112] (b) The process is to match the same time frames of the motion guide video information and the motion execution video information using coordinate information based on the results of applying the dynamic time warping (DTW) technique.

[0113] More specifically, process (b) is a process of correcting duplicate or unmatched frames based on the results of applying the dynamic time warping (DTW) technique to establish a 1:1 correspondence relationship between the reference action frame and the user action frame, and using the coordinate information to match the same time frame of the reference action frame of the action guide video information and the user action frame of the action execution video information.

[0114]

[0115] 3. (c) Process

[0116] (c) The process is to measure the similarity of each joint by calculating the Euclidean distance between the user's joint coordinates and the pre-set reference joint coordinates (coordinate information of the reference joint extracted from the motion guide video information) in each matched frame.

[0117] Here, the Euclidean distance formula (D) for measuring similarity is as follows.

[0118]

[0119] (Xs,Ys) : Joint coordinates of the reference motion (Reference joint coordinates)

[0120] (Xu,Yu) : Joint coordinates of user action (user joint coordinates)

[0121] Meanwhile, the similarity of each joint can be calculated using the following formula.

[0122]

[0123] D: Euclidean distance of the corresponding joint

[0124] Dmax: Preset maximum distance value (e.g., 50 pixels)

[0125] The higher the result value, the higher the similarity (100% = perfect match)

[0126]

[0127] 4. (d) Process

[0128] (d) The process is to calculate the similarity of the overall motion by averaging the Euclidean distance values ​​calculated from each matched frame.

[0129] The similarity of the overall operation can be calculated using the following formula.

[0130]

[0131] N: Total number of target joints

[0132]

[0133] Result output unit (109)

[0134] The result output unit (109) displays the user's action performance results, which are scored according to similarity.

[0135] The result output unit (109) can be provided as various devices, such as, for example, a display device, a projector, a smartphone, a tablet, etc.

[0136]

[0137] Meanwhile, another embodiment of the present invention is a motion-following evaluation method for evaluating a user's motion performance state, comprising: (a) a step in which an image output unit (101) of a motion-following evaluation system (100) outputs motion guide image information (i1) to be provided to a user; (b) a step in which an image capturing unit (103) of a motion-following evaluation system (100) captures a user's motion to obtain motion performance image information (i3); (c) a step in which a coordinate extraction unit (105) of a motion-following evaluation system (100) analyzes the motion performance image information (i3) to set coordinate information (i5) for the user's joint parts; (d) a step in which a motion evaluation unit (107) of a motion-following evaluation system (100) performs frame matching between motion guide image information (i1) and motion performance image information (i3) based on the coordinate information (i5) to calculate similarity. and (e) the result output unit (109) of the action imitation evaluation system (100) is characterized by including the step of displaying the user's action performance result, which is scored according to similarity.

[0138] In addition, in a motion imitation evaluation method according to another embodiment of the present invention, step (c) may include: (c-1) detecting a user in each frame of motion execution video information (i3) to set a bounding box, and calculating the user's center coordinate information based on the set bounding box to identify the location; (c-2) extracting coordinate information for the detected user's joint parts using a neural network-based human joint recognition model; (c-3) selecting among the extracted coordinate information that has a reliability greater than or equal to a preset threshold; and (c-4) setting the selected coordinate information as coordinate information (i5) by performing vector normalization based on the user's shoulder width or body height to correct errors according to the user's body size and shooting environment.

[0139] In addition, in the motion imitation evaluation method according to another embodiment of the present invention, the step (d) comprises: (d-1) applying a dynamic time warping (DTW) technique to correct the difference in the speed of the user's motion execution, calculating the Euclidean distance between the reference motion frame sequence of the motion guide video information (i1) and the user motion frame sequence of the motion execution video information (i3), generating an Accumulated Cost Matrix based on the calculated distance value, extracting an optimal frame alignment path (warping path) having the minimum cost from the Accumulated Cost Matrix, and aligning the frame of the user motion and the frame of the reference motion based on the extracted alignment path; (d-2) matching the same time frame of the motion guide video information (i1) and the motion execution video information (i3) using coordinate information (i5) based on the result of applying the dynamic time warping (DTW) technique; (d-3) a step of measuring the similarity of each joint by calculating the Euclidean distance between the user’s joint coordinates and the preset reference joint coordinates (coordinate information of the reference joint extracted from the motion guide video information) in each matched frame; and (d-4) a step of calculating the similarity of the entire motion by averaging the Euclidean distance values ​​calculated in each matched frame; may be included.

[0140] In the motion imitation evaluation method according to another embodiment of the present invention, the technical features and functions of each subject are the same as those described in the preceding motion imitation evaluation system, so a detailed description is omitted.

[0141]

[0142] As described above, according to one embodiment of the present invention, by evaluating the accuracy of a user's movement performance in real time and providing the result as a score to the user, it is possible to promote motivation for movement performance and improvement of physical activity.

[0143] Furthermore, based on image analysis and joint coordinate extraction technology, the present invention enables more accurate and rapid analysis of a user's movement performance status, thereby having the effect of allowing stable movement evaluation regardless of the physical conditions and movement environments of various users, such as seniors and infants.

[0144] Although it has been described above that all components constituting an embodiment of the present invention are combined or operate as a single unit, the present invention is not necessarily limited to such an embodiment. That is, within the scope of the purpose of the present invention, all components may be selectively combined in one or more ways to operate.

[0145] The foregoing description is merely an illustrative explanation of the technical concept of the present invention, and those skilled in the art to which the present invention pertains will be able to make various modifications and variations within the scope of the essential characteristics of the present invention. Accordingly, the embodiments disclosed in the present invention are intended to explain, not limit, the technical concept of the present invention, and the scope of the technical concept of the present invention is not limited by these embodiments. The scope of protection of the present invention shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of the present invention.

Claims

1. As a motion imitation evaluation system that evaluates the state of a user's action performance, A video output unit that outputs operation guide video information to be provided to the above user; A video capturing unit that captures the actions of the above-mentioned user to acquire action execution video information; A coordinate extraction unit that analyzes the above motion execution video information and sets coordinate information for the joint parts of the user; A motion evaluation unit that calculates similarity by performing frame matching between the motion guide video information and the motion execution video information based on the above coordinate information; and A result output unit that displays the result of the user's action performance, which is scored according to the similarity above; A motion imitation evaluation system characterized by including 2. In Paragraph 1, The above coordinate extraction unit is, (a) A process of detecting the user in each frame of the above action execution video information, setting a bounding box, and calculating the center coordinate information of the user based on the set bounding box to identify the location; (b) A process of extracting coordinate information of the detected joint parts of the user using a neural network-based human joint recognition model; (c) A process of selecting from the extracted coordinate information that the confidence level is greater than or equal to a preset threshold; and (d) A process of setting the selected coordinate information as coordinate information by performing vector normalization based on the user's shoulder width or body height to correct errors according to the user's body size and shooting environment; A motion imitation evaluation system characterized by including 3. In Paragraph 1, The above operation evaluation unit is, (a) A process of applying a dynamic time warping (DTW) technique to compensate for the difference in the speed of the user's action execution, calculating the Euclidean distance between the reference action frame sequence of the action guide video information and the user action frame sequence of the action execution video information, generating an Accumulated Cost Matrix based on the calculated distance value, extracting an optimal frame alignment path (warping path) having the minimum cost from the Accumulated Cost Matrix, and aligning the frames of the user action and the reference action based on the extracted alignment path; (b) A process of matching the same time frame of the motion guide video information and the motion execution video information using the coordinate information based on the result of applying the above dynamic time warping (DTW) technique; (c) a process of measuring the similarity of each joint by calculating the Euclidean distance between the user’s joint coordinates and the preset reference joint coordinates (coordinate information of the reference joint extracted from the motion guide video information) in each matched frame; and (d) A process of calculating the similarity of the overall motion by averaging the Euclidean distance values ​​calculated in each matched frame; A motion imitation evaluation system characterized by including 4. As a motion-following evaluation method for evaluating the user's action performance status, (a) A step in which the video output unit of the motion imitation evaluation system outputs motion guide video information to be provided to the user; (b) A video capturing unit of the above-described action imitation evaluation system captures the user's action and obtains action performance video information; (c) A coordinate extraction unit of the above-described motion imitation evaluation system analyzes the motion performance video information and sets coordinate information for the joint parts of the user; (d) A motion evaluation unit of the motion imitation evaluation system calculates similarity by performing frame matching between the motion guide video information and the motion execution video information based on the coordinate information; and (e) A result output unit of the above-described action imitation evaluation system displays the user's action performance result, which is scored according to the similarity; A method for evaluating movement imitation characterized by including 5. In Paragraph 4, The above step (c) is, (c-1) A step of detecting the user in each frame of the above action execution video information, setting a bounding box, and calculating the center coordinate information of the user based on the set bounding box to identify the location; (c-2) A step of extracting coordinate information of the detected joint parts of the user using a neural network-based human joint recognition model; (c-3) A step of selecting from the extracted coordinate information that the confidence level is greater than or equal to a preset threshold; and (c-4) A step of setting the selected coordinate information as coordinate information by performing vector normalization based on the user's shoulder width or body height to correct errors according to the user's body size and shooting environment; A method for evaluating movement imitation characterized by including 6. In Paragraph 4, The above step (d) is, (d-1) A step of applying a dynamic time warping (DTW) technique to compensate for the difference in the speed of the user's action execution, calculating the Euclidean distance between the reference action frame sequence of the action guide video information and the user action frame sequence of the action execution video information, generating an Accumulated Cost Matrix based on the calculated distance value, extracting an optimal frame alignment path (warping path) having the minimum cost from the Accumulated Cost Matrix, and aligning the frame of the user action and the frame of the reference action based on the extracted alignment path; (d-2) A step of matching the same time frame of the motion guide video information and the motion execution video information using the coordinate information based on the result of applying the above dynamic time warping (DTW) technique; (d-3) A step of measuring the similarity of each joint by calculating the Euclidean distance between the user’s joint coordinates and the preset reference joint coordinates (coordinate information of the reference joint extracted from the motion guide video information) in each matched frame; and (d-4) A step of calculating the similarity of the overall motion by averaging the Euclidean distance values ​​calculated in each matched frame; A method for evaluating movement imitation characterized by including

Images