• Journal of Information Processing Systems
• /
• v.17 no.3
• /
• pp.556-570
• /
• 2021

The existing video expression recognition methods mainly focus on the spatial feature extraction of video expression images, but tend to ignore the dynamic features of video sequences. To solve this problem, a multi-mode convolution neural network method is proposed to effectively improve the performance of facial expression recognition in video. Firstly, OpenFace 2.0 is used to detect face images in video, and two deep convolution neural networks are used to extract spatiotemporal expression features. Furthermore, spatial convolution neural network is used to extract the spatial information features of each static expression image, and the dynamic information feature is extracted from the optical flow information of multiple expression images based on temporal convolution neural network. Then, the spatiotemporal features learned by the two deep convolution neural networks are fused by multiplication. Finally, the fused features are input into support vector machine to realize the facial expression classification. Experimental results show that the recognition accuracy of the proposed method can reach 64.57% and 60.89%, respectively on RML and Baum-ls datasets. It is better than that of other contrast methods.

• The Transactions of the Korea Information Processing Society
• /
• v.7 no.2S
• /
• pp.621-628
• /
• 2000

A retrieval performance of VOD sever is estimated by how quickly it services popular videos to users and how many users it is able to service. Each video data is placed on heterogeneous disks and placement techniques are various, retrieval performance is under the control of these elements, so that a retrieval performance is affected by placement policy. In this paper, we place video data considering their characteristics, especially, we place videos distributively according to their popularity. To verify our policy, we make various environment of experiment, estimate a placement policy using popularity of videos and a contrary policy, and compare them.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2000.10a
• /
• pp.751-754
• /
• 2000

In this paper, we propose an efficient method to detect abrupt shot changes in compressed MPEG video data by using reference ratios among video frames. The reference ratios among video frames imply the degree of similarities among adjacent frames by prediction coded type of each frames. A shot change is detected if the similarity degrees of a frame and its adjacent frames are low. This paper proposes an efficient shot change detection algorithm by using Fuzzy c-means(FCM) clustering algorithm. The FCM clustering uses the shot change probabilities evaluated in the mask matching of reference ratios and difference measure values based on frame reference ratios.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2005.11a
• /
• pp.151-154
• /
• 2005

본 논문에서는 스케일러블 멀티미디어 콘텐츠에 대한 조건적 접근제어가 가능한 암호화 방법을 제안한다. 현재 표준화가 진행중인 스케일러블 비디오 코딩방법인 JSVM(Joint Scalable Video Model)은 부호화한 동영상에 대해 공간, 시간, 품질의 스케일러빌리티(Scalability)를 지원하는데, 각 스케일러 빌리티를 고려한 조건적인 접근제어기술은 스케일러빌리티에 따라 사용자를 제한해야 하는 경우를 위해 필수적인 기술이다. 제안하는 방법은 공간, 시간, 품질의 세가지 스케일러빌리티를 지원하도록 부호화(Encoding)후 구성되는 NAL(Network Abstract Layer)을 지원하는 스케일러빌리티에 따라 구분하고, 구분된 NAL 의 종류에 따라 암호화 key 를 다르게 제공하는 방법을 통해 사용자의 접근제어 수준에 맞게 암호화 key 를 조합하는 방법을 적용하였다. 실험 결과 제안한 방법은 JSVM 에서 공간, 시간, 품질의 스케일러빌리티가 보장되고, 이때 생성되는 Key 의 조합으로 조건적 접근제어(Conditional access control)가 가능함을 확인하였다.

• Journal of Information Processing Systems
• /
• v.14 no.5
• /
• pp.1272-1285
• /
• 2018

The paper presents a novel embedded multifunctional media sever (EMMS) for mobile devices to receive various media programs. Being different from other contemporary system research, the paper mainly studies how to design a reception solution for terrestrial digital television (DTV) on mobile devices and how to enable mobile devices can receive DTV program, enjoy video-on-demand (VOD), achieve video surveillance and relay Internet video program via local Wi-Fi simultaneously. In the system design, we integrate broadcasting-terrestrial DTV tuner, streaming media re-transmission system, VOD disk, video camera and access interface to the Internet into EMMS, which can either receive terrestrial DTV radio signals and demodulate out digital transport stream (TS), or can read streaming media bit-stream from VOD disk, surveillance camera or access interface to the Internet. The experimental results show the proposed system is stable and quality-efficient. Comparing with the other systems, the proposed system has the least packet loss rate and response time.

• Journal of Information Processing Systems
• /
• v.7 no.1
• /
• pp.103-110
• /
• 2011

In this paper, an adaptive motion vector smoothing scheme based on weighted vector median filtering is proposed in order to eliminate the motion outliers more effectively for improving the quality of side information in frame-based distributed video coding. We use a simple motion vector outlier reliability measure for each block in a motion compensated interpolated frame and apply weighted vector median filtering only to the blocks with unreliable motion vectors. Simulation results show that the proposed adaptive motion vector smoothing algorithm improves the quality of the side information significantly while maintaining low complexity at the encoder in frame-based distributed video coding.

• The Transactions of the Korea Information Processing Society
• /
• v.6 no.5
• /
• pp.1410-1417
• /
• 1999

In this paper, we propose the method of extracting the caption regions from news video and the method of recognizing the captions that can be used mainly for content-based indexing and retrieving the MPEG-2 compressed news for NOD(News On Demand). The proposed method can reduce the searching time on detecting caption frames with minimum MPEG-2 decoding, and effectively eliminate the noise in caption regions by deliberately devised preprocessing. Because the kind of fonts that are used for captions is not various in the news video, an enhanced template matching method is used for recognizing characters. We could obtain good recognition result in the experiment of sports news video by the proposed methods.

• 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 Korea Information Processing Society Conference
• /
• 2004.05a
• /
• pp.707-710
• /
• 2004

In the case of terrestrial DMB(Digital Multimedia Broadcasting) system that offers mobile multimedia broadcasting services, transmission error must be considered. Although DMB transmission system provides the error protection functions of convolution coding and Reed-Solomon (204,188,t=8) coding, additional error resilience video coding methods are needed to satisfy the requirement of BER lower than $10^{-8}$. In this thesis, we propose and evaluate effective error resilience coding schemes using the MPEG-4 redundant slice for MPEG-4 video services in the DMB environment. In this scheme, we analyze the drift error caused by transmission error based on the random noise concept and the redundant slice selection algorithm that selects the most influential slice in the view of the drift error increment.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.648-651
• /
• 2023

Video anomaly detection aims to detect abnormal events. Motivated by the power of transformers recently shown in vision tasks, we propose a novel transformer-based network for video anomaly detection. To capture long-range information in video, we employ a multi-scale transformer as an encoder. A convolutional decoder is utilized to predict the future frame from the extracted multi-scale feature maps. The proposed method is evaluated on three benchmark datasets: USCD Ped2, CUHK Avenue, and ShanghaiTech. The results show that the proposed method achieves better performance compared to recent methods.