• The Journal of the Acoustical Society of Korea
• /
• v.13 no.1E
• /
• pp.89-95
• /
• 1994

This paper describes development of a real-time voice dialing system which can recognize around one hundred word vocabularies in speaker independent mode. The voice recognition algorithm in this system is implemented on a DSP board with a telephone interface plugged in an IBM PC AT/486. In the DSP board, procedures for feature extraction, vector quantization(VQ), and end-point detection are performed simultaneously in every 10 msec frame interval to satisfy real-time constraints after detecting the word starting point. In addition, we optimize the VQ codebook size and the end-point detection procedure to reduce recognition time and memory requirement. The demonstration system has been displayed in MOBILAB of the Korean Mobile Telecom at the Taejon EXPO'93.

• The Transactions of the Korea Information Processing Society
• /
• v.5 no.5
• /
• pp.1367-1376
• /
• 1998

In this paper we propose an real time marker detection algorithm for motion analysis both in 2 dimensions and 3 dimensions with CCD camera and rfame grabber only which has no image processor. The main algorithm consists of the following 3 algorithms; 1) the tracing algorithm that makes it possible to predict the expected marker location by narrowing the searching boundary, 2) the searching algorithm that detects the marker in the expected boundary using Ad-hoc previous screen search technique, tornado search method rotate diagonal search method search technique, 3) the algorithm that finds the central point of the detected marker. We try to narrow the searching boundary for real time processing. Also, it is able to find the central point of the detected marker much faster than typical contour tracing algorithm.

• Journal of Broadcast Engineering
• /
• v.4 no.2
• /
• pp.119-126
• /
• 1999

We proposed an algorithm which detects scene changes in video with speediness and accuracy. It is a two-step approach. In the first step, we decide potential scene change segments using the compressed domain data extracted by temporal sampling of MPEG compressed video. In the second step, we determine the exact scene change positions using the pixel values of each frame in those segments by means of combining the intensity and edge changes. In addition we discuss the method to remove false detection generated from camera flash. Integrating the above methods, we introduce a structure that can detect scene changes speedily and accurately.

• Journal of Internet Computing and Services
• /
• v.3 no.2
• /
• pp.69-77
• /
• 2002

Recently, Digital video is one of the important information media delivered on the Internet and playing an increasingly important role in multimedia. 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 combines the RGB color histogram with the $x^2$(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 field 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 KIISE:Computing Practices and Letters
• /
• v.7 no.5
• /
• pp.440-450
• /
• 2001

This paper discusses an effective algorithm which is proposed for abrupt scene change detection in MPEG bitstream. The proposed algorithm restores DC images by decoding only DC coefficients and estimates the new motion vectors between adjacent DC images and detects scene change by similarity measure between frames. The proposed algorithm calculates similarity measure between adjacent frames, i.e motion compensated inter-frame correlation, and detects scene change by comparing this similarity measure with threshold value independent of sequences. Experimental results show that the proposed algorithm has more than 90% \`recall\` and \`precision\` in almost sequences and these results can be considered better than other algorithms using threshold value dependent of sequences.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.37 no.3
• /
• pp.31-40
• /
• 2000

A robust scheme of locating and counting the number of vehicles m urban traffic scenes, a core component of vision-based traffic monitoring systems, is presented The method is based on the evidential reasoning, where vehicle evidences m the background subtraction Image are obtained by a new locally optimum thresholding, and the evidences are merged by three heuristic rules using the geometric constraints The locally optimum thresholding guarantees the separation of bright and dark evidences of vehicles even when the vehicles are overlapped or when the vehicles have similar color to the background Experimental results on diverse traffic scenes show that the detection performance is very robust to the operating conditions such as the camera location and the weather The method may be applied even when vehicle movement is not observed since a static Image IS processed without the use of frame difference.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.5 no.5
• /
• pp.465-470
• /
• 2010

This paper proposes an adaptive video watermarking algorithm according to bit detection rate of watermark in MPEG2 system. The watermark strength is adaptively applied as BER(bit error rate) of watermark extracted from decoded frame for motion compensation. Watermark insertion uses a frequency spread spectrum method. A realtime watermark extraction is done directly in the DCT domain during MPEG decoding. The experimental simulations show that PSNR(peak signal to noise ratio) results 31.5dB for a fixed watermark strength and 33.dB for an adaptive watermark strength. Also average BER is 0.126 and less than 0.2 avaliable value.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2013.10a
• /
• pp.1003-1005
• /
• 2013

4D technology provides physical effects with the general videos or 3D videos. Implementing 4D technology, producing 4D metadata according to video play time and frame data is necessary. In this paper, we propose a method to provide physical effects by judging the temperature of video according to color information. In the proposed method, we provide physical effects to watcher by cognizing the color information in the video when a disaster such as fire is occurred. By using the method, it is expected that 4D matadata for sensing experience like heater device can be produced without programmers automatically.

• Journal of Communications and Networks
• /
• v.14 no.1
• /
• pp.34-39
• /
• 2012

Collisions between tags greatly reduce the identification speed in radio frequency identification (RFID) systems and increase communication overhead. In particular for an active RFID system, tags are powered by small batteries, and a large number of re-transmissions caused by collisions can deteriorate and exhaust the tag energy which may result in missing tags. An efficient collision-free arbitration protocol for active RFID systems is proposed in this paper. In this protocol, a new mechanism involving collision detection, collision avoidance, and fast tag access is introduced. Specifically, the pulse burst duration and busy-tone-detection delay are introduced between the preamble and data portion of a tag-to-reader (T-R) frame. The reader identifies tag collision by detecting pulses and transmits a busy tone to avoid unnecessary transmission when collision occurs. A polling process is then designed to quickly access the collided tags. It is shown that the use of the proposed protocol results in a system throughput of 0.612, which is an obvious improvement when compared to the framed-slotted ALOHA (FSA) arbitration protocol for ISO/IEC 18000-7 standard. Furthermore, the proposed protocol greatly reduces communication overhead, which leads to energy conservation.

• Smart Media Journal
• /
• v.6 no.3
• /
• pp.49-56
• /
• 2017

Images and Videos that include the human face contain a lot of information. Therefore, accurately extracting human face is a very important issue in the field of computer vision. However, in real life, human faces have various shapes and textures. To adapt to these variations, A model-based approach is one of the best ways in which unknown data can be represented by the model in which it is built. However, the model-based approach has its weaknesses when the motion between two frames is big, it can be either a sudden change of pose or moving with fast speed. In this paper, we propose an enhanced human face-tracking model. This approach included human face detection and motion estimation using Cascaded Convolutional Neural Networks, and continuous human face tracking and modeling correction steps using the Active Appearance Model. A proposed system detects human face in the first input frame and initializes the models. On later frames, Cascaded CNN face detection is used to estimate the target motion such as location or pose before applying the old model and fit new target.