Apparatus and method for motion generation based on generative artificial intelligence having object and background coordinate correction function

Abstract

The present invention relates to a motion generation device and method based on generative artificial intelligence equipped with object and background coordinate correction functions. An electronic device according to one embodiment of the present invention includes a memory and a processor connected to the memory, and the processor receives input information for generating three-dimensional motion from a user terminal, generates a motion sequence of a three-dimensional character corresponding to the received input information using a Generative AI model, monitors the coordinate change of a foot joint interacting with the ground among a plurality of joints constituting the character within the motion sequence, determines a specific frame section within the motion sequence to be in a ground contact state based on the amount of coordinate change of the foot joint, and performs physical constraint correction to prevent slipping on the ground by correcting the coordinates of the foot joint for the frame section determined to be in a ground contact state.

Description

Technology Field

[0001] The present invention relates to a motion generation device and method based on generative artificial intelligence equipped with object and background coordinate correction functions. Background Technology

[0003] Unless otherwise indicated in this specification, the contents described in this section are not prior art for the claims of this application, and are not to be recognized as prior art simply because they are included in this section.

[0004] With the recent rapid growth of the virtual reality (VR), augmented reality (AR), and metaverse industries, the importance of 3D avatars and digital humans that act as proxies for or interact with users within virtual spaces is emerging. In order for these 3D characters to convey a sense of presence to users, it is essential to implement not only the appearance of the character but also natural and realistic motion.

[0005] Conventional methods for producing 3D character animation have relied on skilled professional animators manually manipulating joint movements in keyframe units or recording the movements of real people using expensive optical motion capture equipment. However, these methods have the problem of requiring enormous time and cost for production, and presenting a high barrier to entry for general users without professional personnel or specialized equipment.

[0006] Accordingly, research on Generative AI technology, which automatically generates 3D motion with only simple input such as a text prompt, and specifically Motion Diffusion Models (MDM), has been actively conducted recently. These deep learning-based generative models have the advantage of being able to probabilistically generate various motions by learning from vast amounts of motion data.

[0007] However, existing generative AI models focus primarily on ensuring visual similarity or diversity in motion data, which limits their ability to perfectly reflect the physical laws of the real world. A typical problem is the frequent occurrence of a phenomenon known as "foot skating," where the feet, which should be firmly planted on the ground, slide or float instead of remaining fixed to the ground when a generated character performs walking or running motions. This issue arises because the coordinates of the foot joints within the generated motion sequence are not physically linked to the ground coordinates but are calculated independently, significantly undermining the realism and immersion of the resulting output.

[0008] Therefore, there is an urgent need for technical means to efficiently generate motion through generative AI, while simultaneously automatically detecting and correcting physical errors in the generated result, particularly slipping caused by misalignment of contact between the ground and the character, to provide high-quality animation. Prior art literature

[0010] Korean Registered Patent No. 10-2872821 (October 14, 2025) The problem to be solved

[0011] One embodiment of the present invention provides a motion generation device and method based on generative artificial intelligence equipped with object and background coordinate correction functions.

[0012] The technical problems to be solved by the present invention are not limited to those mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art to which the present invention belongs from the description below. means of solving the problem

[0014] To achieve the above-mentioned purpose, an electronic device according to one embodiment of the present invention includes a memory and a processor connected to the memory, and the processor receives input information for generating three-dimensional motion from a user terminal, generates a motion sequence of a three-dimensional character corresponding to the received input information using a Generative AI model, monitors the coordinate change of a foot joint interacting with the ground among a plurality of joints constituting the character within the motion sequence, determines a specific frame section within the motion sequence as being in contact with the ground based on the amount of coordinate change of the foot joint, and performs physical constraint correction to prevent slipping on the ground by correcting the coordinates of the foot joint for the frame section determined as being in contact with the ground.

[0015] At this time, the input information includes a text prompt describing the movement of the character, and the processor may use a Motion Diffusion Model (MDM) as the generative artificial intelligence model, which converts the input information into vector values ​​through a text encoder and removes noise to generate the motion sequence.

[0016] At this time, in the step of determining the ground contact state, the processor may calculate a vertical movement value based on the temporal variance of the vertical axis (Y-axis) coordinate value of the foot joint, and determine the ground contact state as a frame interval in which the vertical movement value is smaller than a preset threshold vertical value.

[0017] At this time, in the step of performing the physical constraint correction, the processor may clamp the horizontal axis (X-axis, Z-axis) coordinates of the foot joint at the start of the frame segment determined to be in contact with the ground during the frame segment, and forcibly assign the vertical axis (Y-axis) coordinates of the foot joint to match the ground level.

[0018] At this time, the processor may further perform an inverse kinematics (IK) process to recalculate the rotation and position values ​​of the character's parent joints using the corrected position of the foot joint as a constraint after the physical constraint correction is performed.

[0019] At this time, the processor derives a target frame interval for calculation that includes a predetermined number of frames before and after a specific frame (t) for deriving the vertical movement value, and can derive the vertical movement value based on the variance of the vertical axis coordinate value of the foot joint in the target frame interval for calculation.

[0020] At this time, the above vertical movement value is derived by the following mathematical formula,

[0021]

[0022] V_move represents the vertical movement value in the specific frame (t), N represents the number of operations, y_i represents the vertical axis coordinate value of the foot joint in the i-th frame of the operation target frame interval, and ay may represent the average value of the vertical axis coordinate value of the foot joint in all frames of the operation target frame interval.

[0023] At this time, the processor may set a transition section having a preset number of transition frames to prevent abrupt changes in the coordinates (vertical axis and horizontal axis) of the foot joint at the start and end points of the frame section determined to be in contact with the ground, and in the transition section, derive the final coordinates of the foot joint by performing linear interpolation between the corrected coordinates (vertical axis and horizontal axis) of the foot joint derived by the physical constraint correction and the original coordinates (vertical axis and horizontal axis) of the foot joint generated by the generative artificial intelligence model.

[0024] At this time, the above final coordinates are derived by the following mathematical formula,

[0025]

[0026] P_final refers to the final coordinates for a specific frame, P_fixed refers to the corrected coordinates, P_raw refers to the original coordinates, w refers to the interpolation weight, and the interpolation weight is derived by the following mathematical formula,

[0027]

[0028] w represents the interpolation weight for the specific frame, K represents the number of transition frames, d_start represents the number of frame differences between the frame where the transition interval starts and the specific frame, d_end represents the number of frame differences between the frame where the transition interval ends and the specific frame, and min() may be a function that selects the minimum value among the values ​​inside the parentheses. Effects of the invention

[0030] As such, according to one embodiment of the present invention, a motion generation device and method based on generative artificial intelligence equipped with object and background coordinate correction functions can be provided.

[0031] The effects obtainable from the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below. Brief explanation of the drawing

[0033] Other aspects, features, and benefits of specific preferred embodiments of the present invention, as described above, will become more apparent from the following description in conjunction with the accompanying drawings. FIG. 1 is a conceptual diagram of a generative artificial intelligence-based motion generation device equipped with object and background coordinate correction functions according to one embodiment of the present invention. FIG. 2 is a block diagram of an electronic device according to one embodiment of the present invention. FIG. 3 is a drawing showing a foot joint according to one embodiment of the present invention. FIG. 4 is a drawing showing a frame section in contact with the ground according to one embodiment of the present invention. FIG. 5 is a drawing showing an upper-level joint according to an embodiment of the present invention. FIG. 6 is a flowchart of a motion generation method based on generative artificial intelligence equipped with object and background coordinate correction functions according to an embodiment of the present invention. It should be noted that in the drawings above, similar reference numbers are used to illustrate identical or similar elements, features, and structures. Specific details for implementing the invention

[0034] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

[0035] In describing the embodiments, technical details that are well known in the art to which the present invention belongs and are not directly related to the present invention are omitted. This is intended to convey the essence of the present invention more clearly without obscuring it by omitting unnecessary explanations.

[0036] For the same reason, some components in the attached drawings have been exaggerated, omitted, or schematically depicted. Additionally, the size of each component does not entirely reflect its actual dimensions. Identical or corresponding components in each drawing have been assigned the same reference numbers.

[0037] The advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below but can be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined only by the scope of the claims. Throughout the specification, the same reference numerals refer to the same components.

[0038] At this point, it will be understood that each block of the process flow diagrams and combinations of the flow diagrams can be executed by computer program instructions. Since these computer program instructions can be loaded into the processor of a general-purpose computer, a special-purpose computer, or other programmable data processing equipment, the instructions executed through the processor of the computer or other programmable data processing equipment create means to perform the functions described in the flow diagram block(s). Since these computer program instructions can also be stored in computer-available or computer-readable memory that can be directed toward the computer or other programmable data processing equipment to implement the function in a specific way, the instructions stored in computer-available or computer-readable memory can also produce a manufactured item containing instruction means to perform the function described in the flow diagram block(s). Since computer program instructions can be loaded onto a computer or other programmable data processing equipment, instructions that perform a series of operation steps on the computer or other programmable data processing equipment to create a process executed by the computer can also provide steps for executing the functions described in the flowchart block(s).

[0039] Additionally, each block may represent a module, segment, or part of code containing one or more executable instructions for executing a specified logical function(s). It should also be noted that in some alternative execution examples, the functions mentioned in the blocks may occur out of order. For instance, two blocks described in succession may actually be executed substantially simultaneously, or the blocks may be executed in reverse order according to their corresponding functions.

[0040] In this embodiment, the term "part" refers to a software or hardware component such as a field-programmable gate array (FPGA) or an application-specific integrated circuit (ASIC), and the "part" performs certain roles. However, the meaning of "part" is not limited to software or hardware. The "part" may be configured to reside in an addressable storage medium or configured to run one or more processors. Accordingly, as an example, the "part" includes components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functions provided within the components and "parts" may be combined into a smaller number of components and "parts" or further separated into additional components and "parts." In addition, the components and '~parts' may be implemented to play one or more CPUs within the device or secure multimedia card.

[0041] In describing the embodiments of the present invention in detail, the primary focus will be on examples of specific systems, but the main point claimed in this specification is applicable to other communication systems and services having a similar technical background without significantly departing from the scope disclosed in this specification, and this will be possible at the judgment of a person with skilled technical knowledge in the relevant technical field.

[0042] In addition, the user terminal described below may include a communication-capable desktop computer, laptop computer, notebook, smartphone, tablet PC, mobile phone, smart watch, smart glass, e-book reader, PMP (portable multimedia player), portable game console, navigation device, digital camera, DMB (digital multimedia broadcasting) player, digital audio recorder, digital audio player, digital video recorder, digital video player, PDA (Personal Digital Assistant), etc.

[0044] FIG. 1 is a conceptual diagram of a generative artificial intelligence-based motion generation device having object and background coordinate correction functions according to one embodiment of the present invention, and FIG. 2 is a block diagram of an electronic device (100) according to one embodiment of the present invention.

[0045] An electronic device (100) according to one embodiment includes a processor (110) and a memory (120). The processor (110) can perform at least one of the methods described above. The memory (120) can store information related to the method described above or store a program in which the method described above is implemented. The memory (120) may be volatile memory or non-volatile memory. The memory (120) may be referred to as a 'database', 'storage unit', etc.

[0046] The processor (110) can execute a program and control the electronic device (100). The code of the program executed by the processor (110) can be stored in memory (120). The device (100) can be connected to an external device (e.g., a personal computer or a network) through an input / output device (not shown) and exchange data.

[0047] At this time, the processor (110) can receive input information for generating three-dimensional motion from a user terminal.

[0048] At this time, the input information may include a text prompt describing the action of the character.

[0049] For example, the text prompt may be a sentence entered by the user in the form of natural language and may include content describing specific actions, speed, direction, or emotional states to be performed by the character. As a specific example, the text prompt may instruct actions that reflect the character's emotions, such as 'jumping happily' or 'shocked and startled'.

[0050] In addition, the text prompt can describe complex body movements beyond simple motion commands. For example, it may specify the positional relationship of specific body parts, such as 'thinking with one hand on the chin and the other hand on the hip,' or describe a gesture combining upper body and head movements, such as 'shaking the head while waving the hand to indicate refusal.' Furthermore, it may include descriptions of movements premised on a situation involving a specific sport or tool, such as 'lightly putting a golf ball from the right direction.' These text prompts are analyzed by the processor (110) and used as basic data for generating a motion sequence that enables the character to perform natural movements corresponding to the description.

[0052] In addition, the processor can generate a motion sequence of a 3D character corresponding to the received input information using a Generative AI model.

[0053] At this time, the generative artificial intelligence model may be a Motion Diffusion Model (MDM) that generates the motion sequence by converting the input information into vector values ​​through a Text Encoder and removing noise.

[0054] Specifically, the generative artificial intelligence model refers to a deep learning-based model that learns a large-scale dataset to generate new data similar to the distribution of the training data. In one embodiment of the present invention, the generative artificial intelligence model learns a dataset that maps text descriptions to corresponding human motion data, thereby identifying the semantic context of an input text prompt and generating new natural human 3D movements that correspond thereto.

[0055] In particular, the motion diffusion model (MDM) may be a model that operates based on a probabilistic diffusion process. The motion diffusion model learns a reverse diffusion process that recovers meaningful motion data from a random noise distribution. For example, the processor (110) receives a text vector converted through the text encoder as a condition and can predict and generate clean motion by gradually removing noise through iterative denoising steps starting from the initial noise data.

[0056] In this process, the motion diffusion model described above can generate motions that flow naturally over time by learning the temporal dependencies of motion data based on a Transformer architecture.

[0057] In addition, the motion sequence refers to a set of multiple frames arranged in chronological order as the output of the generative artificial intelligence model. Each frame within the motion sequence includes pose information of the 3D character at a specific point in time.

[0058] At this time, the posture information may include position coordinates in three-dimensional space and rotation information (e.g., quaternions or Euler angles) for each of the multiple joints (e.g., head, shoulders, elbows, pelvis, knees, ankles, etc.) constituting the character. Accordingly, by applying the motion sequence to the character's rig and playing it sequentially, continuous animations such as walking, running, and waving hands can be implemented.

[0059] In addition, the processor may monitor the change in coordinates of a foot joint that interacts with the ground among a plurality of joints constituting the character within the motion sequence, determine a specific frame segment within the motion sequence to be in a state of contact with the ground based on the amount of change in coordinates of the foot joint, and perform physical constraint correction to prevent slipping on the ground by correcting the coordinates of the foot joint for the frame segment determined to be in a state of contact with the ground. This will be described in more detail later.

[0061] FIG. 3 is a drawing showing a foot joint according to one embodiment of the present invention, and FIG. 4 is a drawing showing a frame section in a ground contact state according to one embodiment of the present invention.

[0062] Referring to FIGS. 3 and 4, the processor, in the step of determining the ground contact state, calculates a vertical movement value based on the temporal variance of the vertical axis (Y-axis) coordinate value of the foot joint, and can determine the ground contact state as a frame interval in which the vertical movement value is smaller than a preset threshold vertical value.

[0063] Specifically, when a character performs a standing or walking motion on the ground in three-dimensional space, there should be almost no physical change in the vertical height (Y-axis) of the foot joints while the feet are in contact with the ground. Therefore, the processor can analyze the variance value, which indicates how severely the vertical axis coordinates of the foot joints fluctuate over time, and determine that the foot is in a state of being firmly in contact with the ground in the section where the value is stably maintained below a certain level (critical vertical value). This is intended to exclude unstable states in the generated motion where the foot does not adhere to the ground and shakes slightly or lifts, and to identify a section of firm support.

[0064] Meanwhile, the specific formula and process for calculating the above vertical movement value will be described later.

[0065] In addition, the processor may, in the step of performing the physical constraint correction, clamp the horizontal axis (X-axis, Z-axis) coordinates of the foot joint at the start of the frame segment determined to be in contact with the ground during the frame segment, and forcibly assign the vertical axis (Y-axis) coordinates of the foot joint to match the ground level.

[0066] In this case, the above-mentioned physical constraint correction may refer to a technical process that ensures the realism of the movement by retrospectively correcting motion data generated probabilistically by an artificial intelligence model when it violates the physical laws of the real world (e.g., friction, gravity, etc.).

[0067] That is, the processor clamps the horizontal position (X, Z axis) of the foot to the starting point coordinate value of the section so that it does not change during the entire section in which the foot is determined to be in contact with the ground, thereby fundamentally eliminating the 'foot skating' phenomenon in which the foot slides while ignoring ground friction.

[0068] In addition, the processor can eliminate the visual discrepancy of the foot penetrating the ground or floating slightly in the air by aligning the vertical height (Y-axis) of the foot with the ground height (e.g., 0), and can implement realistic movement in which the character is stably grounded on the ground.

[0070] FIG. 5 is a drawing showing an upper-level joint according to an embodiment of the present invention.

[0071] Referring to FIG. 5, the processor may further perform an inverse kinematics (IK) process to recalculate the rotation and position values ​​of the parent joints of the character using the corrected position of the foot joint as a constraint after the physical constraint correction is performed.

[0072] In this case, the above-mentioned Inverse Kinematics (IK) is a technique used in 3D computer graphics and robotics, and refers to a mathematical algorithm that calculates inversely the rotation angles and positions of the lower joints required to reach a target position of an end-effector (in this invention, 'foot') in a chain-like structure once that target position is determined. This is a concept opposite to Forward Kinematics, which determines the position of the end by first determining the angles of the upper joints.

[0073] In addition, the above-mentioned parent joints refer to joints that are connected above a specific joint (in the direction of the root or pelvis) in the character's skeleton hierarchy and affect the movement of that joint. For example, if the end joint is the 'foot', the parent joints may include the 'knee' joint and the 'hip' joint.

[0074] In other words, if the coordinates of the foot joint are clamped to the ground height by the physical constraint correction performed earlier, a geometric discrepancy may occur between the existing leg posture before correction and the corrected foot position.

[0075] At this time, the processor sets the corrected foot position as a target constraint and performs an inverse kinematics process, thereby automatically readjusting the joint angles of the entire leg so that the knee naturally bends or straightens in accordance with the fixed foot position. This prevents distortion where only the foot is attached to the ground while the leg is abnormally stretched or bent, and enables the completion of a natural final pose in which the character is stably standing on the ground.

[0077] In addition, the processor can derive a target frame section for calculation that includes a predetermined number of frames before and after a specific frame (t) for deriving the vertical movement value as described above, and derive the vertical movement value based on the variance of the vertical axis coordinate value of the foot joint in the target frame section for calculation.

[0078] Looking at it in more detail, the above vertical movement value can be derived by the following mathematical formula 1.

[0079] [Mathematical Formula 1]

[0080]

[0081] At this time, V_move represents the vertical movement value in the specific frame (t), N represents the number of operations, y_i represents the vertical axis coordinate value of the foot joint in the i-th frame of the operation target frame interval, and ay may represent the average value of the vertical axis coordinate value of the foot joint in all frames of the operation target frame interval.

[0082] At this time, the above mathematical formula 1 is designed to quantify the vertical stability of the foot joint by adopting a statistical variance calculation method. Specifically, the formula calculates how much the height of the foot deviates from the average height of the interval during a total of 2N+1 frame intervals covering the past (tN) and future (t+N) based on the current time point (t).

[0083] At this time, it is designed to accumulate and reflect the magnitude of the movement itself regardless of the direction of change (up or down) by squaring the difference between the height of each frame and the average height.

[0084] In other words, the fact that the vertical movement value, which is the result of this formula, is calculated to be close to 0 means that the height of the foot joint remained constant and hardly changed during the calculation period, which suggests that there is a very high probability that the foot is physically in contact with the ground and supporting body weight. On the other hand, the fact that the value is calculated to be large means that the height of the foot fluctuated erratically from the average value, so it can be used as an indicator that the foot is moving or floating in the air.

[0085] Through this, ground contact, which is difficult to determine based solely on instantaneous position values, can be determined more robustly by considering the time-series context.

[0087] In addition, the processor may set a transition section having a preset number of transition frames to prevent abrupt changes in the coordinates (vertical axis and horizontal axis) of the foot joint at the start and end points of the frame section determined to be in contact with the ground, and in the transition section, derive the final coordinates of the foot joint by performing linear interpolation between the corrected coordinates (vertical axis and horizontal axis) of the foot joint derived by the physical constraint correction and the original coordinates (vertical axis and horizontal axis) of the foot joint generated by the generative artificial intelligence model.

[0088] At this time, the above final coordinates can be derived by the following mathematical formula 2.

[0089] [Mathematical Formula 2]

[0090]

[0091] In this case, P_final refers to the final coordinates for a specific frame, P_fixed refers to the corrected coordinates, P_raw refers to the original coordinates, and w may refer to the interpolation weight.

[0092] In addition, the above interpolation weights can be derived by the following mathematical formula 3.

[0093] [Mathematical Formula 3]

[0094]

[0095] At this time, w represents the interpolation weight for the specific frame, K represents the number of transition frames, d_start represents the number of frame differences between the frame where the transition interval starts and the specific frame, d_end represents the number of frame differences between the frame where the transition interval ends and the specific frame, and min() may be a function that selects the minimum value among the values ​​inside the parentheses.

[0096] At this time, the above-mentioned transition section refers to a buffer section set at the boundary point where one enters from a state where physical constraint correction is not applied (a section where the foot moves freely) to a state where correction is applied (a section where the foot is completely fixed to the ground), or conversely, when correction is released.

[0097] If the foot position is forcibly fixed immediately from a specific point in time without separate transition processing, a popping phenomenon may occur in the animation where the foot bounces as if teleporting, causing visual discontinuity. Therefore, to prevent such abrupt coordinate changes, the present invention gradually changes the coordinates by providing a section of a certain length (number of transition frames).

[0098] To this end, the above mathematical formula 2 represents a linear interpolation logic for determining the final position of the foot joint within this transition interval. The formula mixes the 'original coordinates' generated by the artificial intelligence model and the 'corrected coordinates' forcibly fixed according to physical laws at an appropriate ratio using weights (w). Through this, it induces convergence to a position with applied physical constraints while maintaining the characteristics of the original movement.

[0099] In addition, the above mathematical formula 3 is designed to dynamically calculate the 'interpolation weight (w),' which is the ratio of mixing two coordinates. The formula calculates how far the current frame is from both ends (start point or end point) of the transition interval, and is designed so that the weight approaches 1 as it moves toward the center of the interval.

[0100] That is, at the edges of the transition zone, the ratio of the original coordinates is increased to allow for a natural entry, and at the center where ground contact is certain, the ratio of the corrected coordinates is maximized to perfectly fix the foot to the ground. Through this dynamic change of weights, the present invention can simultaneously secure physical accuracy, such as 'foot slip prevention,' and visual quality, such as 'naturalness of movement.'

[0102] FIG. 6 is a flowchart of a motion generation method based on generative artificial intelligence equipped with object and background coordinate correction functions according to an embodiment of the present invention.

[0103] Referring to FIG. 6, a generative artificial intelligence-based motion generation method having object and background coordinate correction functions according to one embodiment of the present invention can receive input information for 3D motion generation from a user terminal (S101).

[0104] In addition, a motion generation method based on generative artificial intelligence having object and background coordinate correction functions according to one embodiment of the present invention can generate a motion sequence of a 3D character corresponding to the received input information using a generative artificial intelligence (Generative AI) model (S103).

[0105] In addition, a motion generation method based on generative artificial intelligence having object and background coordinate correction functions according to one embodiment of the present invention can monitor changes in the coordinates of a foot joint interacting with the ground among a plurality of joints constituting the character within the motion sequence (S105).

[0106] In addition, a motion generation method based on generative artificial intelligence having object and background coordinate correction functions according to one embodiment of the present invention can determine a specific frame section within the motion sequence to be in a ground contact state based on the amount of change in coordinates of the foot joint (S107).

[0107] In addition, a motion generation method based on generative artificial intelligence having object and background coordinate correction functions according to one embodiment of the present invention can perform physical constraint correction to prevent slipping on the ground by correcting the coordinates of the foot joints for the frame section determined to be in contact with the ground (S109).

[0108] In addition, a generative AI-based motion generation method having an object and background coordinate correction function according to one embodiment of the present invention can be configured in the same way as a generative AI-based motion generation device having an object and background coordinate correction function disclosed in FIGS. 1 to 5.

[0110] The embodiments described above may be implemented as hardware components, software components, and / or combinations of hardware and software components. For example, the devices, methods, and components described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are also possible.

[0111] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0112] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0113] Although the embodiments have been described above with reference to the limited drawings, those skilled in the art can apply various technical modifications and variations based on the above. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or if the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0114] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below.

Claims

Claim 1 In an electronic device, memory; and a processor connected to said memory; The processor comprises: receiving input information for generating 3D motion from a user terminal; generating a motion sequence of a 3D character corresponding to the received input information using a Generative AI model; monitoring the coordinate change of a foot joint interacting with the ground among a plurality of joints constituting the character within the motion sequence; determining a specific frame segment within the motion sequence to be in a ground contact state based on the amount of coordinate change of the foot joint; and performing physical constraint correction to prevent slipping on the ground by correcting the coordinates of the foot joint for the frame segment determined to be in a ground contact state; the input information includes a text prompt describing the movement of the character; the processor uses a Motion Diffusion Model (MDM) as the Generative AI model, which generates the motion sequence by converting the input information into vector values ​​and removing noise through a text encoder; and the processor, in the step of determining the ground contact state, the vertical axis (Y-axis) coordinate value of the foot joint Calculate a vertical movement value based on temporal variance, determine a frame segment in which the vertical movement value is smaller than a preset threshold vertical value as the ground contact state, and in the step of performing the physical constraint correction, the processor clamps the horizontal axis (X-axis, Z-axis) coordinates of the foot joint at the start of the frame segment determined as the ground contact state during the corresponding frame segment, and forcibly assigns the vertical axis (Y-axis) coordinates of the foot joint to align with the ground level, and after the physical constraint correction is performed,Inverse Kinematics (IK) processes are further performed to recalculate rotation and position values ​​of the parent joints of the character using the corrected position of the foot joint as a constraint, and the processor derives a target frame interval for computation that includes a preset number of frames before and after a specific frame (t) for deriving the vertical movement value, and derives the vertical movement value based on the variance of the vertical axis coordinate value of the foot joint in the target frame interval for computation, wherein the vertical movement value is derived by the following mathematical formula, V_move represents the vertical movement value at the specific frame (t), N represents the number of operations, y_i represents the vertical axis coordinate value of the foot joint at the i-th frame in the operation target frame section, ay represents the average value of the vertical axis coordinate value of the foot joint across all frames in the operation target frame section, and the processor sets a transition section having a preset number of transition frames to prevent abrupt changes in the coordinates (vertical axis and horizontal axis) of the foot joint at the start and end points of the frame section determined to be in contact with the ground, and in the transition section, performs linear interpolation between the corrected coordinates (vertical axis and horizontal axis) of the foot joint derived by the physical constraint correction and the original coordinates (vertical axis and horizontal axis) of the foot joint generated by the generative artificial intelligence model to derive the final coordinates of the foot joint, and the final coordinates are derived by the following mathematical formula, P_final refers to the final coordinates for a specific frame, P_fixed refers to the corrected coordinates, P_raw refers to the original coordinates, w refers to the interpolation weight, and the interpolation weight is derived by the following mathematical formula, An electronic device characterized in that w represents the interpolation weight for the specific frame, K represents the number of transition frames, d_start represents the number of frame differences between the frame where the transition interval starts and the specific frame, d_end represents the number of frame differences between the frame where the transition interval ends and the specific frame, and min() is a function that selects the minimum value among the values ​​inside the parentheses. Claim 2 delete Claim 3 delete

Images