• Journal of Information Processing Systems
• /
• v.14 no.1
• /
• pp.150-161
• /
• 2018

We propose a novel deep neural network model for detecting human activities in untrimmed videos. The process of human activity detection in a video involves two steps: a step to extract features that are effective in recognizing human activities in a long untrimmed video, followed by a step to detect human activities from those extracted features. To extract the rich features from video segments that could express unique patterns for each activity, we employ two different convolutional neural network models, C3D and I-ResNet. For detecting human activities from the sequence of extracted feature vectors, we use BLSTM, a bi-directional recurrent neural network model. By conducting experiments with ActivityNet 200, a large-scale benchmark dataset, we show the high performance of the proposed DeepAct model.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2004.05a
• /
• pp.505-509
• /
• 2004

본 논문에서는 비디오 스트림 서버에서 의미 기반 검색을 가능하게 하기 위하여 대용량 스트림 데이터를 효과적으로 표현하고 저장하는 기법을 제시한다. 비디오 스트림 내의 각 프레임을 다차원 공간상의 점으로 사상함으로써 비디오 스트림은 다차원 시퀀스(multidimensional sequence)로 표현되고, 이 시퀀스는 다시 비디오 세그먼트로 분할된다. 분할된 세그먼트로부터 정적인 특성과 연속된 프레임의 움직임을 나타내는 트랜드 벡터(trend vector)등의 의미 정보를 추출하여 모델링 함으로서 스트림 데이터를 효과적으로 표현한다. 또한 제안된 기법은 효율적인 검색을 위하여 비디오 세그먼트를 인덱싱하고 저장하는 방법을 제공함으로써 공간 사용의 효율성을 높이고 신속한 검색을 가능하게 한다.

• Proceedings of the Korea Multimedia Society Conference
• /
• 2001.11a
• /
• pp.708-712
• /
• 2001

디지털 데이터의 저작권 보호를 위하여 최근 디지털 워터마킹에 관하여 많은 연구가 이루어지고 있다. 디지털 워터마킹이란 인간의 시각이나 청각이 인지할 수 없는 범위 내에서 디지털 데이터의 값을 약간 변경함에 따라 워터마크라고 하는 저작권 정보를 몰래 삽입하는 방법이다. 본 논문에서는 비 압축 비디오 계열(uncompressed video sequence)에 디지털 워터마크를 삽입하기 위한 방법을 보여준다. 대역 확산 기법을 이용하여 비디오에 pseudo-noise signal을 눈에 보이지 않도록 추가하고, 여러 조작에 대해 강하도록 구성하였다. 성능을 평가하기 위하여 비디오에 워터마크를 삽입하고 그 유효성을 보인다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.28 no.12C
• /
• pp.1200-1208
• /
• 2003

Affective computing plays an important role in intelligent Human Computer Interactions(HCI). To detect emotional events, it is desirable to construct a computing model for extracting emotion related features from video. In this paper, we propose a statistical model based on the probabilistic distribution of low level features in video shots. The proposed method extracts low level features from video shots and then from a GMM(Gaussian Mixture Model) for them to detect emotional shots. As low level features, we use color, camera motion and sequence of shot lengths. The features can be modeled as a GMM by using EM(Expectation Maximization) algorithm and the relations between time and emotions are estimated by MLE(Maximum Likelihood Estimation). Finally, the two statistical models are combined together using Bayesian framework to detect emotional events in video.

• International Journal of Internet, Broadcasting and Communication
• /
• v.12 no.2
• /
• pp.90-97
• /
• 2020

Recently, with the growth of mixed reality industrial infrastructure, relevant convergence research has been proposed. For real-time mixed reality services such as remote video conferencing, the research on real-time acquisition-process-transfer methods is required. This paper aims to implement an AR remote rendering method of volumetric 3D video data. We have proposed and implemented two modules; one, the parsing module of the volumetric 3D video to a game engine, and two, the server rendering module. The result of the experiment showed that the volumetric 3D video sequence data of about 15 MB was compressed by 6-7%. The remote module was streamed at 27 fps at a 1200 by 1200 resolution. The results of this paper are expected to be applied to an AR cloud service.

• Proceedings of the Korea Institute of Convergence Signal Processing
• /
• 2006.06a
• /
• pp.45-48
• /
• 2006

This paper presents an approach to visual information based temporal indexing of video sequences. The objective of this work is the integration of an automatic face detection and a matching system for video indexing. The face detection is done using color information. The matching stage is based on the Principal Component Analysis (PCA) followed by the Minimax Probability Machine (MPM). Using PCA one feature vector is calculated for each face which is detected at the previous stage from the video sequence and MPM is applied to these feature vectors for matching with the training faces which are manually indexed after extracting from video sequences. The integration of the two stages gives good results. The rate of 86.3% correctly classified frames shows the efficiency of our system.

• Korean Chemical Engineering Research
• /
• v.55 no.1
• /
• pp.135-140
• /
• 2017

This paper is on a study for discrimination on relative sequence as a most actively discussed topic in forensic document fields. This paper describes the application of the visual spectral comparator and infinite focus microscope as observation methods for overlapping region of printing and writing lines. As a result, we could categorize overlapping region images and identify the sequence of printing and writing lines by various inks.

• The Transactions of the Korea Information Processing Society
• /
• v.6 no.8
• /
• pp.2031-2041
• /
• 1999

In this paper, we present the HVIS (Hybrid Video Information System) which bolsters up meaning retrieval of all the various users by integrating feature-based retrieval and annotation-based retrieval of unformatted formed and massive video data. HVIS divides a set of video into video document, sequence, scene and object to model the metadata and suggests the Two layered Hybrid Object-oriented Metadata Model(THOMM) which is composed of raw-data layer for physical video stream, metadata layer to support annotation-based retrieval, content-based retrieval, and similarity retrieval. Grounded on this model, we presents the video query language which make the annotation-based query, content-based query and similar query possible and Video Query Processor to process the query and query processing algorithm. Specially, We present the similarity expression to appear degree of similarity which considers interesting of user. The proposed system is implemented with Visual C++, ActiveX and ORACLE.

• Journal of the Korea Society of Computer and Information
• /
• v.11 no.4 s.42
• /
• pp.147-154
• /
• 2006

Effective and efficient representation of color features of multiple video frames is an important vet challenging task for visual information management systems. This paper Proposes a Video Browsing Service(VBS) that provides both the video content retrieval and the video browsing by the real-time user interface on Web. For the scene segmentation and key frame extraction of video sequence, we proposes an efficient scene change detection method that combine the RGB color histogram with the X2 (Chi Square) histogram. Resulting key frames are linked by both physical and logical indexing. This system involves the video editing and retrieval function of a VCR's. Three elements that are the date, the need and the subject are used for video browsing. A Video Browsing Service is implemented with MySQL, PHP and JMF under Apache Web Server.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.12 no.8
• /
• pp.3697-3705
• /
• 2011

There is the temporal correlation of the video sequence between the motion vector of current block and the motion vector of previous block. In this paper, we propose a new prediction search algorithm for block matching using the temporal and spatial correlation of the video sequence and local statistics of neighboring motion vectors. The proposed ANBA(Adaptive Neighboring-Block Search Algorithm) determines the location of a better starting point for the search of an exact motion vector using the point of the smallest SAD(Sum of Absolute Difference) value by the predicted motion vectors of neighboring blocks around the same block of the previous frame and the current frame and use a previous motion vector. Simulation results show that PSNR(Peak-to-Signal Noise Ratio) values are improved up to the 1.06dB as depend on the video sequences and improved about 0.01~0.64dB over MVFAST and PMVFAST.