• The Journal of Korean Institute of Next Generation Computing
• /
• v.15 no.5
• /
• pp.31-42
• /
• 2019

With the recent development of deep learning technology, remarkable achievements have been made in many research areas of computer vision. Deep learning has also made dramatic improvement in two-dimensional or three-dimensional human pose estimation based on a single image, and many researchers have been expanding the scope of this problem. The human pose estimation is one of the most important research fields because there are various applications, especially it is a key factor in understanding the behavior, state, and intention of people in image or video analysis. Based on this background, this paper surveys research trends in estimating human poses based on a single image. Because there are various research results for robust and accurate human pose estimation, this paper introduces them in two separated subsections: 2D human pose estimation and 3D human pose estimation. Moreover, this paper summarizes famous data sets used in this field and introduces various studies which utilize human poses to solve their own problem.

• Proceedings of the Korean Information Science Society Conference
• /
• 2005.11b
• /
• pp.841-843
• /
• 2005

비디오 데이터 분석은 감시, 검색, 스포츠 경기 자동 요약 등 많은 분야에서 사용되는 기술이다. 그러나 감시 카메라나 스포츠 경기 비디오와 같이 사람의 영역이 저해상도인 환경에서는 포즈 추정, 모델과의 매칭이 어렵기 때문에 제스처 인식 연구는 많이 이루어지고 있지 못하다. 본 논문에서는 카메라가 Pan/Tilt/Zoom 동작을 하고 사람이 빠르게 움직이는 방송용 테니스 비디오에서, 사람을 추출하고, Curvature Scale Space를 기반으로 한 특징을 추출하여 학습된 포즈 모델과 매칭하는 방법과, 차원의 축소를 통해 일련의 포즈들을 학습된 제스처와 매칭하는 방법을 제안한다. 50개의 방송용 테니스 경기 비디오 장면에 대하여 서브 제스처 추출을 수행한 결과, 서브 포즈에 대하여 모델과 매칭이 잘 되고, 매칭이 되지 않는 포즈를 포함하는 시퀀스에 대해서도 강인한

• Journal of KIISE:Software and Applications
• /
• v.35 no.10
• /
• pp.600-607
• /
• 2008

This paper proposes a robust human detection method irrespective of their pose variation using the multiple oriented 2D elliptical filters (MO2DEFs). The MO2DEFs can detect the humans regardless of their poses unlike existing object oriented scale adaptive filter (OOSAF). To overcome OOSAF's limitation, we introduce the MO2DEFs whose shapes look like the oriented ellipses. We perform human detection by applying four different 2D elliptical filters with specific orientations to the 2D spatial-depth histogram and then by taking the thresholds over the filtered histograms. In addition, we determine the human pose by using convolution results which are computed by using the MO2DEFs. We verify the human candidates by either detecting the face or matching head-shoulder shapes over the estimated rotation. The experimental results showed that the accuracy of pose angle estimation was about 88%, the human detection using the MO2DEFs outperformed that of using the OOSAF by $15{\sim}20%$ especially in case of the posed human.

• Journal of KIISE:Software and Applications
• /
• v.36 no.10
• /
• pp.836-850
• /
• 2009

Human posture recognition has attracted tremendous attention in ubiquitous environment, performing arts and robot control so that, recently, many researchers in pattern recognition and computer vision are working to make efficient posture recognition system. However the most of existing studies is very sensitive to human variations such as the rotation or the translation of body. This is why the feature, which is extracted from the feature extraction part as the first step of general posture recognition system, is influenced by these variations. To alleviate these human variations and improve the posture recognition result, this paper presents 3D Star Skeleton and Principle Component Analysis (PCA) based feature extraction methods in the multi-view environment. The proposed system use the 8 projection maps, a kind of depth map, as an input data. And the projection maps are extracted from the visual hull generation process. Though these data, the system constructs 3D Star Skeleton and extracts the rotation invariant feature using PCA. In experimental result, we extract the feature from the 3D Star Skeleton and recognize the human posture using the feature. Finally we prove that the proposed method is robust to human variations.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2015.10a
• /
• pp.1696-1697
• /
• 2015

최근 포즈 인식 기술의 연구는 HCI, 인공지능 등의 분야에서 활발히 진행되고 있다. 하지만 대부분의 연구는 센서와 사용 환경으로부터 많은 영향을 받는다. 본 논문에서는 이러한 종속성을 최소화하여 범용성 있는 포즈 인식 방법을 제안한다. 이 방법을 통해 사람의 다양한 포즈로부터 획득한 특징 데이터를 최적화함으로써 다양한 포즈를 실시간에 인식할 수 있다.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.23 no.5
• /
• pp.423-430
• /
• 2013

In typical face recognition systems, the frontal view of face is preferred to reduce the complexity of the recognition. Thus individuals may be required to stare into the camera, or the camera should be located so that the frontal images are acquired easily. However these constraints severely restrict the adoption of face recognition to wide applications. To alleviate this problem, in this paper, we address the problem of tracking and recognizing faces in video captured with no environmental control. The face tracker extracts a sequence of the angle/size normalized face images using IVT (Incremental Visual Tracking) algorithm that is known to be robust to changes in appearance. Since no constraints have been imposed between the face direction and the video camera, there will be various poses in face images. Thus the pose is identified using a PCA (Principal Component Analysis)-based pose classifier, and only the pose-matched face images are used to identify person against the pre-built face DB with 5-poses. For face recognition, PCA, (2D)PCA, and $(2D)^2PCA$ algorithms have been tested to compute the recognition rate and the execution time.

• Proceedings of the KIEE Conference
• /
• 2008.07a
• /
• pp.1909-1910
• /
• 2008

본 논문에서는 실린더 모델을 이용하여 머리의 다양한 포즈 변화와 조명 변화에 대해 강인한 얼굴 인식을 제안하고자 한다. 실린더 모델은 사람의 머리가 실린더 모양과 유사하고 그 표면은 얼굴에 해당된다고 가정한다. 실린더 모델은 6가지의 모션 파라메터를 따라 움직이며 Lucas-Kanade 알고리즘에 의해 모션 파라메터의 양을 결정한다. 강인한 동작을 위해 템플릿을 지속적으로 바꿔주는 동적 템플릿(dynamic template)방법과 그에 따른 에러가 누적되는 것을 막기 위해 re-registration방법을 사용한다. 조명 문제를 해결하기 위해 템플릿에서 조명 주성분 벡터를 추출하여 제거하는 방법으로 조명 효과를 제거한다. 실험에서는 다양한 포즈 변화와 조명 변화가 반영된 얼굴 데이터베이스를 구축하고 추출한 텍스쳐 맵(texture map image)을 SVM에 적용함으로서 포즈, 조명 변화에 강인한 얼굴인식을 보인다.

• Journal of KIISE
• /
• v.41 no.11
• /
• pp.927-934
• /
• 2014

In this paper, we propose an effective method for recognizing daily human activities from a stream of three dimensional body poses, which can be obtained by using Kinect-like RGB-D sensors. The body pose data provided by Kinect SDK or OpenNI may suffer from both the view variance problem and the scale variance problem, since they are represented in the 3D Cartesian coordinate system, the origin of which is located on the center of Kinect. In order to resolve the problem and get the view-invariant and scale-invariant features, we transform the pose data into the spherical coordinate system of which the origin is placed on the center of the subject's hip, and then perform on them the scale normalization using the length of the subject's arm. In order to represent effectively complex internal structures of high-level daily activities, we utilize Hidden state Conditional Random Field (HCRF), which is one of probabilistic graphical models. Through various experiments using two different datasets, KAD-70 and CAD-60, we showed the high performance of our method and the implementation system.

• Smart Media Journal
• /
• v.12 no.7
• /
• pp.59-67
• /
• 2023

Recently, research on behavior analysis tracking human posture and movement has been actively conducted. In particular, OpenPose, an open-source software developed by CMU in 2017, is a representative method for estimating human appearance and behavior. OpenPose can detect and estimate various body parts of a person, such as height, face, and hands in real-time, making it applicable to various fields such as smart healthcare, exercise training, security systems, and medical fields. In this paper, we propose a method for classifying four exercise movements - Squat, Walk, Wave, and Fall-down - which are most commonly performed by users in the gym, using OpenPose-based deep learning models, DNN and CNN. The training data is collected by capturing the user's movements through recorded videos and real-time camera captures. The collected dataset undergoes preprocessing using OpenPose. The preprocessed dataset is then used to train the proposed DNN and CNN models for exercise movement classification. The performance errors of the proposed models are evaluated using MSE, RMSE, and MAE. The performance evaluation results showed that the proposed DNN model outperformed the proposed CNN model.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.951-952
• /
• 2023

소위 '인생네컷'이라 불리는 셀프사진관은 MZ 세대의 새로운 놀이 문화로 떠오르며 사용자 수가 나날이 증가하고 있다. 그러나 짧은 시간 내에 다양한 포즈를 취해야 하는 셀프사진관 특성상 촬영이 낯선 사람에게는 여전히 진입장벽이 존재한다. 더불어 매번 비슷한 포즈와 사진 결과물에 기존 사용자는 점차 흥미를 잃어가는 문제점도 발생하고 있다. 이에 본 연구에서는 셀프사진관 사용자의 관계를 분류하는 모델을 개발하여 관계에 따른 적합하고 다양한 포즈를 추천하는 프레임워크를 제안한다. 사용자의 관계를 'couple', 'family', 'female_friend', 'female_solo', 'male_friend', 'male_solo' 총 6 개로 구분하였고 실제 현장과 유사하도록 단색 배경의 이미지를 우선으로 학습 데이터를 수집하여 모델의 성능을 높였다. 모델 학습 단계에서는 모델의 성능을 높이기 위해 여러 CNN 기반의 모델을 전이학습하여 각각의 정확도를 비교하였다. 결과적으로 195 장의 test_set 에서 accuracy 0.91 의 성능 평가를 얻었다. 본 연구는 객체 인식보다 객체 간의 관계를 학습시켜 관계성을 추론하고자 하는 것을 목적으로, 연구 결과가 희박한 관계 분류에 대한 주제를 직접 연구하여 추후의 방향성이나 방법론과 같은 초석을 제안할 수 있다. 또한 관계 분류 모델을 CCTV 에 활용하여 미아 방지 혹은 추적과 구조 등에 활용하여 국가 치안을 한층 높이는 데 기대할 수 있다.