Earticle

현대사회는 모든 정보가 디지털화되고 있으므로 개인 정보의 보호가 무척이나 중요한 이슈다. 특히 미디어 산업의 폭발적 성장으로 인해 비디오에 촬영된 인물의 익명화는 매우 중대한 문제가 되었다. 전통적인 방법은 블러링이나 픽셀화를 사용하고, 최신 기술들 은 생성적 적대 신경망(GAN)을 활용하여 비디오에 촬영된 얼굴을 다시 그리는 방법으로 익명화를 달성한다. 우리는 훨씬 작은 모델 을 활용하여 실시간 연산을 통해 비디오에 촬영된 인물의 신체 전부를 익명화하는 방법을 제안한다. 기존 방법은 인물의 피부색, 의복, 소지품, 체형 등 얼굴 이외의 개인 식별 정보를 제거하는 것이 어려웠으나, 우리의 방법은 영상 내에서 이러한 정보를 모두 지울 수 있다. 또한 자세 인식 알고리즘을 활용하여 인물의 위치, 움직임, 자세 등의 정보는 표현할 수 있다. 이 알고리즘은 다양한 산업 현장에 설치된 CCTV나 IP 카메라에 적용되어 실시간으로 동작할 수 있으므로, 전신 익명화 기술의 보급에 기여할 수 있을 것이다.

In the contemporary digital era, protection of personal information has become a paramount issue. The exponential growth of the media industry has heightened concerns regarding the anonymization of individuals captured in video footage. Traditional methods, such as blurring or pixelation, are commonly employed, while recent advancements have introduced generative adversarial networks (GANs) to redraw faces in videos. In this study, we propose a novel approach that employs a significantly smaller model to achieve real-time full-body anonymization of individuals in videos. Unlike conventional techniques that often fail to effectively remove personal identification information—such as skin color, clothing, accessories, and body shape—our method successfully eradicates all such details. Furthermore, by leveraging pose estimation algorithms, our approach accurately represents information regarding individuals' positions, movements, and postures. This algorithm can be seamlessly integrated into CCTV or IP camera systems installed in various industrial settings, functioning in real-time and thus facilitating the widespread adoption of full-body anonymization technology.

In many of 5G applications, end devices with lack of computing power often need to carry out heavy computations involving multimedia data. Edge computing has emerged as a promising solution to circumvent scarce resources at end devices, with moderate delays compared to cloud computing. In this work, we study the problem of offloading video processing tasks to edge servers. To this end, we develop a deep reinforcement learning based method for selecting either local or edge server to process video frames. We demonstrate the performance of our method through experiments with video frame transform tasks.

Early development of listening ability has been viewed as vital in the field of teaching English as a Foreign/Second Language (EFL/ESL). Both researchers and teachers have thus been greatly concerned with finding some ways to improve the learner’s listening competence, or better ideas to help learners acquire listening skills in the formal classroom setting. As one way of achieving this goal, this study spells out advantages of using technologically up-to-date audio/video data processing softwares over traditional audio/video players or CD/DVD players. These traditional language teaching aids have been of use and still can be to some extent. But they have been somewhat limited in terms of easier retrieval of necessary language input at the teacher’s will. For instance, when teachers try to use some specific parts of audio-video tapes for test construction, they have to spend a lot of time and energy cutting and pasting them. In addition, the audio/video quality often decreases in the process due to some technical problems. This study shows that both Digital Sound Processors (DSP) like Adobe Audition and Visual Data Processors (VDP) like Virtual Dub can be used so as to overcome those ever-existing limitations and thus to enhance ways of teaching listening. As a case study, this study comes up with a few practical ideas of how to apply language teaching materials made by using audio/video data processing programs to teaching listening, particularly for adult EFL learners.

넓은 지역에서의 종합감시시스템 구축은 인력을 효율적으로 관리할 수 있으며 자동화된 감시망을 구축하여 비감시 지역을 줄일 수 있는 특징이 있다. 본 논문에서는 넓은 지역에서 적용할 수 있는 지능형 종합감시시스템에서 입력되는 비디오의 특징을 위치와 용도별로 분석하여 적합한 비디오 감시 알고리즘을 선택할 수 있는 방안을 제시하였다. 7가지 대표적인 상황으로서 침입, 물건 버림/없어짐, 배회, 혼잡도 측정 등이다.

In this paper, we introduced design of intelligent surveillance camera system and typical event processing scenario for urban transit. To analyze video, we studied events that frequently occur in surveillance camera system. Event processing scenario is designed for seven representative situations(designated area intrusion, object abandon, object removal in designated area, object tracking, loitering and congestion measurement) in urban transit. Our system is optimized for low hardware complexity, real time processing and scenario dependent solution.

본 연구의 목적은 장기간의 상호작용적 비디오 게임이 노인의 인지정보처리에 어떤 영향을 미치는지 알아보는 데 있다. 본 연구에 참여할 피험자는 K시 D, J, K, W 노인복지관 노인 남ㆍ녀(65-70세) 250명 중 신체활동수준을 측정 하여 신체활동이 낮은 수준(낮음: 3200kcal 이하/1주)을 보인 60명이며, 피험자들 모두 사전 동의를 거쳐 자발적으로 본 실험에 참여하였다. 선정된 모든 피험자들은 난수표를 이용하여 무선 할당되어 (1) 상호작용적 비디오게임 집단(20 명) (2) 유산소운동(20명) (3) 통제집단(20명)으로 배정하였다. 본 연구의 실험설계는 3(집단)×2(사전사후)에 대해 반복 측정 이원분산분석을 실시했다. 종속변수는 인지기능 척도(주의집중력, 지연 기억력, 단기기억 능력, 즉각 기억력, 언어 유창성, 전두엽 운동기능), ERP 분석에서는 P300의 진폭과 잠재기, 반응시간과 정확률이다. 연구결과에서 인지기능과 ERP 분석에서 운동수행의 반응시간과 반응 정확률 및 진폭과 잠재기에서 상호작용적 비디오게임 집단과 유산소 운동집 단은 유의미한 통계적 차이가 없었으나, 상호작용적 비디오게임 집단과 유산소 운동(걷기운동)집단이 통제집단보다 향 상된 결과를 보였다. 이러한 연구결과는 인지적 운동인 상호작용적 비디오게임과 같은 유산소운동의 꾸준한 참여는 노 인의 인지기능 쇠퇴방지에 좋은 영향을 미칠 수 있다고 여겨진다.

The objectives of this study was to examine the effect of Interactive Video Game on cognitive information processing the elderly. Sixty elderly were attended in this study. Their ages ranged from 65 to 70, with a mean age of 67.60 years. The subjects were randomly assigned to one of three experimental conditions: (1) interactive video game group (n=20), (2) aerobic exercise group (n=20), (3) control group (n=20). The experimental design of this study was analyzed using two-way ANOVAs with repeated measures of groups and time. Cognitive function was assessed by neuroelectrical response, and ERP analysis. The results of the study showed that the interactive video game group and aerobic exercise group showed no significant statistical differences in the response time, response accuracy, amplitude and potential of the performance of the exercise in cognitive function and ERP analysis, but improved the interaction video game group and aerobic exercise (walking) group over the control group. It was concluded that long-term aerobic exercise like interactive video game is associated with attenuation of cognitive decline in the elderly.

영상 감시 시스템은 다양한 실외 환경에서 실시간으로 수집되는 다채널 영상을 처리해야 하며, 저시정 환 경에서는 영상 품질이 급격히 저하되는 문제가 발생한다. 이를 해결하기 위한 인공지능(AI) 기반 영상 복원 기술 은 고품질 영상 복원을 가능하게 하지만, 높은 연산 자원 요구로 인해 모든 채널에 일괄 적용하기에는 실시간성 확보가 어렵다. 본 논문에서는 복원 기능을 채널별로 선택적으로 적용할 수 있는 선택적 제어 구조(Selective Control Architecture)를 제안하고, 온 디바이스 환경에서 구현된 시스템의 자원 효율성과 실시간 처리 성능을 실 험적으로 분석하였다. 실험 결과, 복원 기능의 적용 채널 수에 따라 GPU 사용률과 영상 출력 속도가 크게 달라지 는 것을 확인하였으며, 선택적 제어 구조가 자원 최적화와 시스템 안정성 확보에 효과적임을 입증하였다. 본 연구 는 AI 영상 복원 기술의 실시간 시스템 통합에 있어 유연하고 실용적인 구조 설계 방안을 제시하며, 향후 모델 경량화 및 병렬 처리 구조 확장을 통한 시스템 고도화를 계획하고 있다.

Video surveillance systems must process multi-channel real-time streams in various outdoor environments, where image quality often deteriorates under low-visibility conditions such as fog or haze. While AI-based video restoration techniques have shown promise in recovering high-quality images, their high computational demands make it difficult to apply them uniformly across all channels in real-time systems. This paper proposes a Selective Control Architecture that enables channel-wise activation of AI restoration functions. The system was implemented in an on-device environment using a CPU with integrated GPU, and its performance was evaluated by measuring GPU usage and frame rate under varying numbers of active restoration channels. The experimental results demonstrate that GPU load and processing speed significantly change depending on the number of channels using AI restoration, and that the proposed architecture effectively optimizes resource usage while maintaining system stability. This study provides a practical system-level solution for integrating AI-based restoration into surveillance platforms, and future work will focus on model compression and parallel processing to enhance real-time performance and scalability.

본 연구는 화상대화를 하는 동안 자기지각 여부에 따른 자기초점주의와 사후처리의 변화양상을 확인하고 자 했다. 80명의 성인을 대상으로 고 사회불안집단과 저 사회불안집단을 선별하였다. 화상대화 시 자기 얼 굴제시 여부로 무선 배정하여 최종 4집단을 구성하였다. 참가자들은 5분 동안 낯선 이성과 화상대화에 참 여하였으며 화상대화 전, 직후, 24시간 후 각각 설문에 응하였다. 화상대화 전과 직후 얼굴지각, 자기초점 주의, 사후처리 수준의 차이를 확인한 결과, 얼굴지각에 따른 자기초점주의 수준의 차이는 유의하지 않았 다. 고 사회불안집단에서는 저 사회불안집단보다 화상대화 직후 및 24시간 후 사후처리 수준이 모두 높게 나타났다. 이에 대한 상관분석 결과, 고 사회불안집단에서 얼굴 확인 정도가 부정적 평가 예상 정도 및 사 후처리와 정적상관을 보였으며, 저 사회불안집단에서는 얼굴 확인 정도가 긍정적 평가 예상 정도, 상태 자 기초점주의 및 상태 외부초점주의와 정적상관이 나타났다. 사회불안이 높은 집단에서 얼굴지각과 사후처리 사이의 기제를 확인하기 위해 매개분석을 실시하였다. 그 결과, 고 사회불안집단의 경우에만 자신의 얼굴 을 지각하면 부정적 평가에 대한 예상을 하게 되며 이는 화상대화 24시간 후의 사후처리에 영향을 미치는 것으로 나타났다. 본 연구는 COVID-19로 인해 화상회의 및 수업이 증가한 상황에서 사람들이 자신의 얼굴 을 보는 것에 불편감을 느끼는 이유에 대한 함의를 제공하며, 사회불안장애 특정적인 사후처리가 대면뿐만 아니라 온라인 사회적 상호작용 후에도 나타남을 확인하였다.

Video conversations have increased due to COVID-19. This study investigated the relationship between perception of self-face, self-focused attention (SFA), and post-event processing (PEP) in social anxiety (SA). Eighty highly and low socially anxious participants were randomly assigned to a self-face (n = 40) or non-self-face condition (n = 40) and completed baseline measures of social anxiety, depression, trait SFA, state SFA, and trait PEP. After SFA was manipulated by presenting self-face, participants engaged in a 5-min unstructured conversation with a confederate in Zoom, followed by a manipulation check. State PEP was assessed online after the conversation and the next day (24h). Results showed that presented self-face does not increase SFA in both high and low SA groups. The presented self-face only in the high SA group allowed participants to anticipate negative evaluation, affecting state-PEP after 24 hours. These results provide explain why people with SA feel uncomfortable seeing their faces in a video conversation and suggest that there is a need to find a new method to manipulate SFA.

Kinect is a motion-sensing device which was originally developed for the Xbox 360 gaming console. This recently developed low-cost sensor detects the body position, motion, and voice; it consists of a microphone, a RGB camera, and a depth sensor. Kinect is PC-centric sensor which allows developers to develop real-life applications with human gestures and body motions. This paper presents an approach to interpret the indoor room objects in order to match the objects features in depth images captured from an RGBD video database. The dataset consists of color and depth image pairs gathered in real-time indoor home environment. The objects features are matched in depth image pairs with the feature association method to detect stable features at different time instances.

In recent years, various multimedia services have become available and the demand for three-dimensional television (3DTV) is growing rapidly. Since 3DTV is considered as the next generation broadcasting service that can deliver photo-realistic and immersive experiences, a number of advanced 3D video processing techniques have been studied. In this paper, after reviewing the overall 3D video system, we explain several challenging technical issues of 3D video processing.

This paper initially proposes a technique for identifying a moving object in a video clip of stationary background for real time content based multimedia communication systems [2]. It deals with identifying an object of interest. Dynamic objects are identified using both background elimination and background registration techniques. Post processing techniques are applied to reduce the noise. The background elimination method uses concept of least squares to compare the accuracies of the current algorithm with the already existing algorithms. The background registration method uses background subtraction which improves the adaptive background mixture model and makes the system learn faster and more accurately, as well as adapt effectively to changing environments.

ports video metadata extraction system based on the content of basic goal use an automated or semi-automated interactive means to obtain video data as complete features and attributes for efficient retrieval mechanism. For fast access to video information needed, sports video ornamental create conditions. Firstly, video-based layered metadata description model, we discuss the structure of the video processing technology, and an increase in the time domain and airspace video object motion information on this basis. Low-level visual features for video and high-level semantic features presents a particular field of video information for video implicit hierarchical division method. Video automated visual feature extraction, semantic feature places marked attracted achieve human-computer interaction. Focus on the sports information descriptors and visual content descriptors, descriptor structure video. Video data based on hierarchical structure model and video features standard video content description model.

In order to apply the video analytic algorithms in metro railway system, the input images should be classified according to the camera installation environment and its role to handle corresponding events. Before we implement the integrated surveillance system at subway stations, we investigated the required camera’s role at each location to employ the video analytics. Resultantly, desirable event processing scenarios are suggested.

모듈통합형 항공전자시스템은 연방형의 LRU (line replaceable unit)의 기능을 하나의 LRM (line replaceable module)에서 제공하고, 하나의 cabinet에 여러 개의 LRM을 탑재한다. IMA core 시스템의 VPM (video processing module)은 LRM으로써 ARINC 818 ADVB (avionics digital video bus)의 bridge 및 gateway 역할을 한다. ARINC 818은 광 대역폭, 적은 지연시간, 비 압축 디지털영상 전송을 위해 개발된 규격이다. VPM의 FPGA IP core는 ARINC 818 to DVI 또는 DVI to ARINC 818 처리와 video decoder, overlay 기능을 가진다. 본 논문에서는 VPM 하드웨어 구현에 대해 다루고, VPM 기능과 IP core 성능 검증 결과를 보인다.

The integrated modular avionics (IMA) system has quite a number of line repalceable moduels (LRMs) in a cabinet. The LRM performs functions like line replaceable units (LRUs) in federated architecture. The video processing module (VPM) acts as a video bus bridge and gateway of ARINC 818 avionics digital video bus (ADVB). The VPM is a LRM in IMA core system. The ARINC 818 video interface and protocol standard was developed for high-bandwidth, low-latency and uncompressed digital video transmission. FPGAs of the VPM include video processing function such as ARINC 818 to DVI, DVI to ARINC 818 convertor, video decoder and overlay. In this paper we explain how to implement VPM's Hardware. Also we show the verification results about VPM functions and IP core performance.

The object of this research is to develop an Interactive Computerized Graphic Program for graphic output of velocity, acceleration and motion range of body-task reference point (e.g., C.O.G., joint location, etc.). Human motions can be reproduced by scanning (rate = 60Hz) the vidicon image, and the results are stored in an Apple II P.C. memory. The results of this study can be extended to simulation and reproduction of human motions for optimal task design.

This paper presents human perception-based video coding method using an online learning framework. In this work, we analyze the relationship between human attention regions and video quality, and also consider human memory. We classify the motion patterns based on the analysis. Then, we devise a motion pattern classification method using Hedge algorithm. Along with the motion patterns, we smooth out the specific regions or sharpen details of the regions using the regional priorities. The preprocessed sequences are applied to the video codec. The performance is excellent on the overall quality as well as the regional quality.

This paper presents the overview of 3D video processing technologies for 3DTV such as 3D content generation, 3D video codec and video processing techniques for 3D displays. Some experimental results for 3D contents generation are shown in 3D mixed reality and 2D/3D conversion.

The problem of video scheduling is analyzed in the framework of divisible load scheduling. A divisible load can be divided into any number of fractions (parts) and can be processed/computed independently on the processors in a distributed computing system/network, as there are no precedence relationships. In the video scheduling, a frame can be split into any number of fractions (tiles) and can be processed independently on the processors in the network, and then the results are collected to recompose the single processed frame. The divisible load arrives at one of the processors in the network (root processor) and the results of the computation are collected and stored in the same processor. In this problem communication delay plays an important role. Communication delay is the time to send/distribute the load fractions to other processors in the network. and the time to collect the results of computation from other processors by the root processors. The objective in this scheduling problem is that of obtaining the load fractions assigned to each processor in the network such that the processing time of the entire load is a minimum. We derive closed-form expression for the processing time by taking Into consideration the communication delay in the load distribution process and the communication delay In the result collection process. Using this closed-form expression, we also obtain the optimal number of processors that are required to solve this scheduling problem. This scheduling problem is formulated as a linear pro-gramming problem and its solution using neural network is also presented. Numerical examples are presented for ease of understanding.

Multi-view video has recently gained much attraction from academic and commercial fields because it can deliver the immersive viewing of natural scenes. This paper presents multi-view video processing being composed of intermediate view reconstruction (IVR), graphics composition, and multi-view video viewer. First we generate virtual views between multi-view cameras using depth and texture images of the input videos. Then we mix graphic objects to the generated view images. The multi-view video viewer is developed to examine the reconstructed images and composite images. As well, it can provide users with some special effects of multi-view video. We present experimental results that validate our proposed method and show that graphic objects could become the inalienable part of the multi-view video.