KR-20260064767-A - APPARATUS FOR FOR ANALYSISING POSTURE BASED ON AI

KR20260064767AKR 20260064767 AKR20260064767 AKR 20260064767AKR-20260064767-A

Abstract

The present invention relates to an AI-based posture analysis device. According to one aspect of the present invention, the AI-based posture analysis device comprises: a front image input unit that receives a front image captured from the front showing the entire body of a person; a first feature point calculation unit that calculates pairs of first feature points corresponding to body items including the face, shoulders, pelvis, knees, and ankles from the front image input from the front image input unit, using a pre-trained learning model to output two feature points corresponding to the face, two feature points corresponding to the shoulders, two feature points corresponding to the knees, and two feature points corresponding to the ankles from the image captured from the front showing the entire body of a person; and based on the first feature points detected by the first feature point calculation unit for the pelvis, knees, and ankles in the front image, generates a pair of line segments by connecting a pair of first feature points for the pelvis and a pair of first feature points for the ankles with straight lines, and determines whether the pair of first feature points for the knees are located on the pair of line segments or between the pair of line segments, including normal, O-legs, and X-legs. It includes a leg shape classification unit that classifies the leg shape into one of the preset leg shapes.

Inventors

김대정

Assignees

김대정

Dates

Publication Date: 20260508
Application Date: 20241029

Claims (5)

Front image input unit that receives a front image taken from the front showing the entire body of a person; A first feature point calculation unit that calculates first feature points, each corresponding to body items including the face, shoulders, pelvis, knees, and ankles, from a frontal image input by the frontal image input unit, using a pre-trained learning model to output two feature points corresponding to the face, two feature points corresponding to the shoulders, two feature points corresponding to the knees, and two feature points corresponding to the ankles from a frontal image of a person's entire body captured from the front; and An AI-based posture analysis device comprising: a leg shape classification unit that, based on the first feature point calculation unit detecting first feature points for the pelvis, knee, and ankle respectively in a frontal image, generates a pair of line segments by connecting a pair of first feature points for the pelvis and a pair of first feature points for the ankle with a straight line, and classifies the leg shape into one of preset leg shapes including normal, O-legs, and X-legs depending on whether a pair of first feature points for the knee are located on the pair of line segments or between the pair of line segments.
In paragraph 1, A side image input unit that receives a side image taken facing one side of a person from a user so that the person's whole body is visible; and An AI-based posture analysis device further comprising: a second feature point calculation unit that calculates second feature points composed of second feature points for each of the body items including the face, shoulders, pelvis, knees, and ankles from a side image input by the side image input unit, using a pre-trained learning model to output feature points corresponding to the face, shoulders, pelvis, knees, and ankles from a side image of a person's whole body taken from the side.
In paragraph 2, An AI-based posture analysis device further comprising: a body balance point calculation unit that calculates a body balance point quantifying the degree of body misalignment relative to the user's foot based on the difference between the first feature points detected for the face, shoulder, pelvis, and knee relative to the first feature point for the ankle and the difference between the second feature points detected for the face, shoulder, pelvis, and knee relative to the second feature point for the ankle, each based on the first feature points detected for the face, shoulder, pelvis, and ankle in a frontal image by the first feature point detection unit and the second feature points detected for the face, shoulder, pelvis, and knee in a side image by the second feature point detection unit.
In paragraph 2, A horizontal line calculation unit that calculates a horizontal line by connecting pixels among the pixels included in a normal image whose y-axis coordinate values correspond to a preset first reference value; A first slope calculation unit that generates line segments for the face, line segments for the shoulder, line segments for the pelvis, line segments for the knee, and line segments for the ankle by connecting the first feature points calculated for body items including the face, shoulder, pelvis, knee, and ankle in the first feature point calculation unit according to each body item, and calculates slope values corresponding to the face, shoulder, pelvis, knee, and ankle by the angle between the horizontal line calculated in the horizontal line calculation unit and the line segments generated for each of the face, shoulder, pelvis, knee, and ankle; A vertical line calculation unit that calculates a vertical line by connecting pixels among the pixels included in the side image whose x-axis coordinate values correspond to a preset second reference value; An AI-based posture analysis device further comprising: a second slope calculation unit that generates line segments corresponding to face-shoulder, shoulder-pelvis, pelvis-knee, and knee-ankle by connecting the second feature points calculated for body items including face, shoulder, pelvis, knee, and ankle in a preset order in the second feature point calculation unit, and calculates slope values corresponding to face-shoulder, shoulder-pelvis, pelvis-knee, and knee-ankle as the angle between each generated line segment and the vertical line calculated in the vertical line calculation unit.
In paragraph 4, An AI-based posture analysis device further comprising: a body balance index calculation unit that calculates a body balance index by summing the slope values corresponding to the face, shoulder, pelvis, and knee calculated by the first slope calculation unit and the slope values corresponding to the face-shoulder, shoulder-pelvis, pelvis-knee, and knee-ankle calculated by the second slope calculation unit.

Description

AI-based posture analysis device The present invention relates to a technology for analyzing a user's posture from a video of the user captured using AI. Traditional posture analysis is conducted by experts who visually inspect the posture. This conventional method has limitations, as analysis results can vary depending on the expert's experience or condition, and it is difficult to provide an objective evaluation based on numerical data. Furthermore, existing posture analysis technology, which analyzes posture based on measurements obtained by attaching motion sensors to the human body, enables objective analysis by experts compared to conventional methods, but it has the problem of being cumbersome and uncomfortable to wear. With the development of deep learning-based video analysis technology, various technologies for analyzing a user's posture through video analysis are being introduced. This deep learning-based posture analysis (Human Pose Estimation) technology is being actively applied in fields such as posture correction, sports, and medicine due to its ease of application compared to existing posture analysis technologies that use motion sensors. According to the Health Insurance Review and Assessment Service, more than 40% of patients who visited hospitals for spinal diseases in 2022 were teenagers, and the number of cases diagnosed with spinal diseases at a young age is increasing due to the increased use of smartphones and computers. Meanwhile, according to market trends for personal health record software provided by Yahoo Finance, the personal health record software market is growing and is predicted to continue growing until 2033. With the increasing diagnosis of spinal disorders highlighting the importance of proper posture and the rise of app-based personal health management, it is anticipated that the need for technology capable of conveniently analyzing posture via video will further increase. FIG. 1 is a block diagram of an AI-based posture analysis device according to one embodiment of the present invention. FIG. 2 is an example diagram illustrating the calculation of a tilt value in a frontal image in an AI-based posture analysis device according to an embodiment of the present invention. FIG. 3 is an example diagram illustrating the calculation of a tilt value in a side image in an AI-based posture analysis device according to an embodiment of the present invention. FIG. 4 is an exemplary diagram showing a first result area generated by an AI-based posture analysis device according to one embodiment of the present invention. FIG. 5 is an exemplary diagram showing a second result area generated by an AI-based posture analysis device according to one embodiment of the present invention. FIG. 6 is an exemplary diagram showing a third result area generated by an AI-based posture analysis device according to an embodiment of the present invention. FIG. 7 is an exemplary diagram showing a fourth result area generated by an AI-based posture analysis device according to an embodiment of the present invention. 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 may 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 claims. Meanwhile, the terms used in this specification are for describing the embodiments and are not intended to limit the present invention. In this specification, the singular form includes the plural form unless specifically stated otherwise in the text. Referring to FIG. 1, an AI-based posture analysis device (10) according to one embodiment of the present invention may be configured to include a front image input unit (101), a side image input unit (102), a first feature point calculation unit (111), a horizontal line calculation unit (112), a second feature point calculation unit (113), a vertical line calculation unit (114), a first tilt calculation unit (121), a second tilt calculation unit (122), a face asymmetry judgment unit (123), a leg shape classification unit (124), a body balance point calculation unit (125), a pelvic tilt classification unit (126), a back knee judgment unit (127), a body balance index calculation unit (128), a first result area generation unit (131), a second result area generation unit (132), a third result area generation unit (133), a fourth result area generation unit (134), and a result report generation unit (140). The front image input unit (101) may receive a front image taken from a user facing the front of a person so that the person's whole body is visible. The side image input unit