• Journal of information and communication convergence engineering
• /
• v.20 no.3
• /
• pp.189-194
• /
• 2022

As non-face-to-face activities have become commonplace, online video conferencing platforms have become popular collaboration tools. However, existing video conferencing platforms have a structure in which one side unilaterally exchanges information, potentially increase the fatigue of meeting participants. In this study, we designed a video conferencing platform utilizing virtual reality (VR), a metaverse technology, to enable various interactions. A virtual conferencing space and realistic VR video conferencing content authoring tool support system were designed using Meta's Oculus Quest 2 hardware, the Unity engine, and 3D Max software. With the Photon software development kit, voice recognition was designed to perform automatic text translation with the Watson application programming interface, allowing the online video conferencing participants to communicate smoothly even if using different languages. It is expected that the proposed video conferencing platform will enable conference participants to interact and improve their work efficiency.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2022.11a
• /
• pp.85-87
• /
• 2022

최근 CCTV 출입 기록, 휴대폰 보안, 스마트 매장 등에서 얼굴 인식을 통해 개인을 식별하는 기술이 널리 사용되고 있다. 카메라의 각도, 조명, 사람의 움직임 등 얼굴 인식에 많은 외부 환경이 영향을 미치고 있지만 그중에서도 실제 영상에서 얼굴이 차지하는 영역이 작아 저해상도 얼굴 인식에 어려움을 겪고 있다. 이러한 문제점을 해결하고자 본 논문에서는 이미지 해상도가 얼굴 인식에 끼치는 영향을 알아보고 이미지 초해상화를 통해 얼굴 인식 성능을 개선하고자 한다. 쌍선형, 양3차 회선 보간법과 딥러닝 기반의 이미지 초해상화 모델인 RCAN을 이용하여 업스케일링한 데이터셋에 대해 학습한 ArcFace를 통해 얼굴 검증 평가를 진행하였다. 고해상도 이미지는 얼굴 인식 성능을 향상시키며, RCAN을 사용한 이미지 초해상화가 보간법을 사용한 방법보다 더 좋은 성능을 보였다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2010.11a
• /
• pp.588-591
• /
• 2010

본 논문에서는 CCTV 감시 환경에서 얼굴 이미지를 이용하여 영상에 저장된 범죄 용의자 또는 특정한 시간대에 출입한 사람들을 검색할 수 있는 시스템을 설계 및 구현하였다. 제안된 시스템은 감시 영상을 효율적으로 검색하기 위하여 사람의 얼굴이 나타난 장면을 기반으로 감시 영상을 분할하였으며, 최근 얼굴 인식 분야에서 성공적인 업적을 보여주고 있는 신호 처리 분야의 SRC 를 이용하여 얼굴 검색 모듈을 구성하였다. 자체 제작한 KUFD(Korea University Face Database)와 CCTV 환경의 얼굴 인식 기반 검색 시스템 환경을 캠퍼스 내에서 모의 구축하여 제안된 시스템의 성능을 실험적으로 검증하였다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2011.11a
• /
• pp.290-292
• /
• 2011

본 논문에서는 얼굴 요소의 자연적 특징과 PCA(Principal Component Analysis)를 융합한 얼굴인식 알고리즘을 소개한다. 지금까지 PCA 를 비롯한 다양한 얼굴인식 알고리즘이 소개되었지만, 얼굴영상을 하나의 '신호'혹은 '벡터'로 간주하여 이를 수학적 접근법으로 풀이하는 방법이 대부분이었다. 이에 본 논문에서는 템플릿 정합 기법을 이용하여 눈썹, 눈, 턱 등을 형태에 따라 분류하는 특징 분류기를 통하여 그룹을 나누고, 각 그룹별로 PCA 분류를 진행하는 2 단계 알고리즘을 구현하였다. 이를 CMU-PIE 데이터베이스를 이용해 검증하고, 실험 결과를 논의하였다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.11a
• /
• pp.651-654
• /
• 2021

Face anti-spoofing (FAS) techniques play a significant role in the defense of facial recognition systems against spoofing attacks. Existing FAS methods achieve the great performance depending on annotated additional modalities. However, labeling these high-cost modalities need a lot of manpower, device resources and time. In this work, we proposed to use self-transforming modalities instead the annotated modalities. Three different modalities based on frequency domain and temporal domain are applied and analyzed. Intuitive visualization analysis shows the advantages of each modality. Comprehensive experiments in both the CNN-based and transformer-based architecture with various modalities combination demonstrate that self-transforming modalities improve the vanilla network a lot. The codes are available at https://github.com/chenmou0410/FAS-Challenge2021.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.598-600
• /
• 2023

CCTV가 범죄 예방 및 수사에 사용되는데, 수사를 위해 저화질 CCTV 영상에서 특정인의 얼굴 인식엔 어려움을 겪어 CCTV 본연의 역할의 희석된다. 따라서 본 논문은 저화질 영상을 고화질로 변환하여 얼굴 인식의 정확성을 높일 수 있는 알고리즘을 연구하는 것을 목적으로 한다. 기존에 연구된 인공지능 기반의 업스케일링 알고리즘을 분석하여 K-FACE 데이터셋에 적절한 모델을 제안한다. 이를 위해 2020년 이전과 이후의 AI 업스케일링 관련 연구를 비교 분석한다. 향후 제시된 모델을 대상으로 동일한 환경내에서 K-FACE 데이터셋을 학습시켜 통일된 기준의 지표 산출이 필요하다.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.46 no.1
• /
• pp.112-120
• /
• 2009

Face membership authentication is to decide whether an incoming person is an enrolled member or not using face recognition, and basically belongs to two-class classification where support vector machine (SVM) has been successfully applied. The previous SVMs used for face membership authentication have been trained and tested using image feature vectors extracted from member face images of each class (enrolled class and unenrolled class). The SVM so trained using image feature vectors extracted from members in the training set may not achieve robust performance in the testing environments where configuration and size of each class can change dynamically due to member's joining or withdrawal as well as where testing face images have different illumination, pose, or facial expression from those in the training set. In this paper, we propose an effective class discriminating feature vector-based SVM for robust face membership authentication. The adopted features for training and testing the proposed SVM are chosen so as to reflect the capability of discriminating well between the enrolled class and the unenrolled class. Thus, the proposed SVM trained by the adopted class discriminating feature vectors is less affected by the change in membership and variations in illumination, pose, and facial expression of face images. Through experiments, it is shown that the face membership authentication method based on the proposed SVM performs better than the conventional SVM-based authentication methods and is relatively robust to the change in the enrolled class configuration.

• Journal of KIISE
• /
• v.42 no.2
• /
• pp.235-241
• /
• 2015

Locally linear embedding (LLE) [1] is a type of manifold algorithms, which preserves inner product value between high-dimensional data when embedding the high-dimensional data to low-dimensional space. LLE closely embeds data points on the same subspace in low-dimensional space, because the data points have significant inner product values. On the other hand, if the data points are located orthogonal to each other, these are separately embedded in low-dimensional space, even though they are in close proximity to each other in high-dimensional space. Meanwhile, it is well known that the facial images of the same person under varying illumination lie in a low-dimensional linear subspace [2]. In this study, we suggest an improved LLE method for face recognition problem. The method maximizes the characteristic of LLE, which embeds the data points totally separately when they are located orthogonal to each other. To accomplish this, all of the subspaces made by each class are forced to locate orthogonally. To make all of the subspaces orthogonal, the simultaneous Diagonalization (SD) technique was applied. From experimental results, the suggested method is shown to dramatically improve the embedding results and classification performance.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.5A
• /
• pp.717-725
• /
• 2000

Within the last several years, there has been a large number of algorithms developed for face recognition. The majority of these algorithms have been view- and projection-based algorithms. Our definition of projection is not restricted to projecting the image onto an orthogonal basis the definition is expansive and includes a general class of linear transformation of the image pixel values. The class includes correlation, principal component analysis, clustering, gray scale projection, and matching pursuit filters. In this paper, we perform a detailed analysis of this class of algorithms by evaluating them on the FERET database of facial images. In our experiments, a projection-based algorithms consists of three steps. The first step is done off-line and determines the new basis for the images. The bases is either set by the algorithm designer or is learned from a training set. The last two steps are on-line and perform the recognition. The second step projects an image onto the new basis and the third step recognizes a face in an with a nearest neighbor classifier. The classification is performed in the projection space. Most evaluation methods report algorithm performance on a single gallery. This does not fully capture algorithm performance. In our study, we construct set of independent galleries. This allows us to see how individual algorithm performance varies over different galleries. In addition, we report on the relative performance of the algorithms over the different galleries.

• Science of Emotion and Sensibility
• /
• v.17 no.1
• /
• pp.53-62
• /
• 2014

Despite the huge potential of the practical application of emotion recognition technologies, the enhancement of the technologies still remains a challenge mainly due to the difficulty of recognizing emotion. Although not perfect, human emotions can be recognized through human images and sounds. Emotion recognition technologies have been researched by extensive studies that include image-based recognition studies, sound-based studies, and both image and sound-based studies. Studies on emotion recognition through facial expression detection are especially effective as emotions are primarily expressed in human face. However, differences in user environment and their familiarity with the technologies may cause significant disparities and errors. In order to enhance the accuracy of real-time emotion recognition, it is crucial to note a mechanism of understanding and analyzing users' personality traits that contribute to the improvement of emotion recognition. This study focuses on analyzing users' personality traits and its application in the emotion recognition system to reduce errors in emotion recognition through facial expression detection and improve the accuracy of the results. In particular, the study offers a practical solution to users with subtle facial expressions or low degree of emotion expression by providing an enhanced emotion recognition function.