Kee-seong Cho

Efficient Traffic Prediction Algorithm of Mutimedia Traffic for Scheduling the Wireless Network Resources

Real-time variable bit rate (VBR) video traffic generated from multimedia applications, such as MPEG-coded video, is expected to be large portions of the traffic in future wired/wireless networks. The nature of VBR traffic and its quality of service (QoS) constraints increase a number of challenges on network resource managements and operational utilizations. Thus, accurate traffic prediction based dynamic resource allocation and scheduling can significantly improve network performance substantially while satisfying QoS requirements. In this paper, we propose a novel prediction algorithm of MPEG-coded real-time VBR video traffic for IEEE 802.11e wireless LANs (WLANs). Based on statistical property of traffic stream, our algorithm performs a prediction of the next frame size for i-, p-, and b-frames. Simulation study using real- world MPEG-4 coded video traces shows that the proposed algorithm achieves much better performance than ANFIS LMS, and NN methods. In addition, the applicability of our prediction algorithm to IEEE 802.11e WLAN is discussed.

Cross-layered Hop-by-Hop Congestion Control for Multihop Wireless Networks

In this paper, we present a novel congestion control algorithm (HRC), which uses hop-by-hop approach and rate-based transmission, for IEEE 802.11 multihop wireless networks. Since medium contention is most dominant cause of packet loss in multihop wireless networks, our algorithm considers the degree of medium contention level experienced at a given node for data rate control at the transport layer. The main purpose of our algorithm is to maximize the network throughput by data rate control while minimizing the packet losses caused by medium contention. The simulation study using GloMoSim shows that our algorithm achieves up to 40% more the network throughput than TCP and TCP-selective acknowledgment (SACK)

Event-based sport video segmentation using multimodal analysis

Through CATV and OTT (Over the Top) media services, vast amount of sport video contents is broadcasted. For viewers, it is necessary to summarize these contents. Semantic event annotation is helpful for video summarization. In this paper, we proposed an approach of event-based sport video segmentation. Our approach uses webcast text data and several types of video information extracted from the sport video. Through multimodal analysis, the approach calculates an appropriate time interval for each event from the original video. By implementing the approach, it is observed that our implemented system provides acceptable video segmentation results.

Extraction of visual information in basketball broadcasting video for event segmentation system

Video analysis is an essential process to segment and summarize sports videos automatically. In this paper, we propose fast and simple computer vision algorithms which can be employed to an event segmentation system for basketball broadcasting videos. In our approach, camera panning is estimated by the optical flow estimation and flow segmentation algorithms. For recognizing shot classes and clock digits, Convolutional Neural Network (CNN) is used. By the experiment, it is observed that our algorithms operate in real time and are accurate to be adapted to the event segmentation system.

Analyzing and detecting method of Android malware via disassembling and visualization

In light of their rapid growth, there is a pressing need to develop analysis and decision solutions whether or not. However, most of protections are limited understanding of these mobile malware and sophisticated analyzing. In this paper, we propose a method of analyzing and deciding malware on the basis of similarity with existing malware families on the popular platform, Android. We focus on the checking visual similarity among Android malwares and deciding the degree of similarity with other malware families to help distributing to inspector appropriately.

An approach of webcast event clustering for sport video event annotation

Event detection using webcast text is an important function for providing event based video segmenting services of sport videos. In this paper, an approach of webcast event clustering for sport video event annotation is proposed. In our approach, one of unsupervised learning algorithms, called pLSA (probabilistic latent semantic analysis), is used to classify event types. To decide an appropriate number of event types, lexical pattern analysis is used. By the experimental results, it is observed that our approach could be used to extract event keywords from the basketball webcast texts.

Enhanced Speaker Verification Using GMM-Supervector Based Modified Adaptive GMM Training

In this paper, an enhanced speaker verification is proposed by exploring a novel modified adaptive Gaussian mixture model (GMM) training. Based weight factor of observation called the observation reliability; we propose to apply a modified Expectation maximization (EM) algorithm, combined with a modified Maximum a posteriori (MAP) estimation to train the modified adaptive GMM model. Using this proposed model, we generate GMM-supervectors which are combined with SVM for verification task. We evaluate performance of speaker verification system based the proposed approaches on utterances from Korean movie database (“You came from the stars”). Experimental results demonstrate that our proposed approaches can outperform the standard GMM-UBM and GMM-supervector approaches in noise conditions.

Business Models for SongTan Ubiquitous City

This paper presents the main business models in ubiquitous city (u-City), the future city strategically integrated with the ubiquitous computing technology. Based on the concept of u-City, the chief services related to the u-City that has been actively developed are analysed, and then the best main four business models (u-Tour, u-ITS, u-SCM and B2B) for SongTan city are presented. The four business models are selected from making an analysis for the present situation of the tour industries, industrial complex, traditional market in SongTan. With the business models, the strategies to realize the active business transactions, high revenues, high value-adding industries, self-sufficient business through the construction of u-City are also presented. Finally, the promoting plans for local autonomy party, citizen party and consortium of industries/universities/researches to give an comfortable u-Life to citizen in SongTan are described.

Multi-object tracker using kemelized correlation filter based on appearance and motion model

The objective of this study is to determine a tracking method using kernelized correlation filter based on objects appearance and motion model used to track multi-object. This system largely consists of 4 modules: motion model, background subtraction, hijacking handling and occlusion handling. Lab colour model is applied to subtract background, and histogram of oriented gradient (HoG) is used to extract object feature. If occlusion among objects occurs, we use a method that tracks again after removing the overlapping objects in consideration of the depth between objects: The head of the closer object is being taken from a camera positioned below the head of the distant object. Thus, among occluded tracking objects, we find that the most upper located object is considered as the furthest object in captured camera image. If hijacking among objects is occurred, it has been solved by removing the overlapping region of the bounding box between two objects that maintain their relative positions for a period of time. These results indicate that this method may allow a solution for tracking of multi-object to be more robust to real-world tracking environments.

Multi-view video acquisition system using object tracking-based multi-PTZ camera control

We present a novel multi-view video acquisition system using object tracking-based multi-PTZ camera control. Proposed system captures multi-view videos for an object-of-interest by controlling multi-PTZ cameras to focus on it. Control information of each PTZ camera is calculated by homography and a tracked position of the object-of-interest in a top-view image. Moreover, object tracking is performed in the top-view image captured by a camera mounted at the top of the site of interest. Test results using four PTZ cameras and two top-view cameras at the basketball court showed successful acquisition of multi-view video for an object-of-interest within 800 milliseconds.