• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.46 no.4
• /
• pp.50-57
• /
• 2009

This paper presents a new model-based macroblock layer rate control algorithm for low bit rate video coding which generates output bits corresponding to a target bit budget. The H.264 standard uses various coding modes and optimization methods to improve the compression performance, which makes it difficult to control the generated traffic accurately in low bit rate environments. In the proposed scheme, we first estimate MAD values of macroblocks in a frame and define a target remaining bits using the estimated MAD values before encoding each macroblock. If a difference between the target value and the actual value is greater than a threshold value, the quantization parameter is adjusted to decrease the difference. It is shown by experimental results that the new algorithm can obtain more than 66% decrease of the difference between the target bits and the resulting bits for a frame with the PSNR performance better than that of the existing rate control algorithm.

• Journal of electromagnetic engineering and science
• /
• v.14 no.3
• /
• pp.284-292
• /
• 2014

This paper presents a characteristic evaluation of commercial gallium nitride (GaN) transistors having two different gate-lengths of $0.4-{\mu}m$ and $0.25-{\mu}m$ in the design of a class-S power amplifier (PA). Class-S PA is operated by a random pulse-width input signal from band-pass delta-sigma modulation and has to deal with harmonics that consider quantization noise. Although a transistor having a short gate-length has an advantage of efficient operation at higher frequency for harmonics of the pulse signal, several problems can arise, such as the cost and export license of a $0.25-{\mu}m$ transistor. The possibility of using a $0.4-{\mu}m$ transistor on a class-S PA at 955 MHz is evaluated by comparing the frequency characteristics of GaN transistors having two different gate-lengths and extracting the intrinsic parameters as a shape of the simplified switch-based model. In addition, the effectiveness of the switch model is evaluated by currentmode class-D (CMCD) simulation. Finally, device characteristics are compared in terms of current-mode class-S PA. The analyses of the CMCD PA reveal that although the efficiency of $0.4-{\mu}m$ transistor decreases more as the operating frequency increases from 955 MHz to 3,500 MHz due to the efficiency limitation at the higher frequency region, it shows similar power and efficiency of 41.6 dBm and 49%, respectively, at 955 MHz when compared to the $0.25-{\mu}m$ transistor.

• Journal of KIISE:Information Networking
• /
• v.30 no.1
• /
• pp.17-23
• /
• 2003

In this Paper. we Propose a digital watermarking scheme for digital images based on a perceptual model, the frequency masking, texture making, and luminance masking Properties of the human visual system(HVS), which have been developed in the context of image compression. We embed two types of watermark, one is pseudo random(PN) sequences, the other is a logo image. To embed the watermarks, original images are decomposed into $8\times8$ blocks, and the discrete cosine transform(DCT) is carried out for each block. Watermarks are casted in the low frequency components of DCT coefficients. The perceptual model adjusts adaptively scaling factors embedding watermarks according to the local image properties. Experimental results show that the proposed scheme presents better results than that of non-perceptual watermarking methods for image qualify without loss of robustness.

• International Journal of Automotive Technology
• /
• v.6 no.2
• /
• pp.107-117
• /
• 2005

A distributed hardware-in-the-loop simulation (HILS) platform is developed for designing an automotive engine control system. The HILS equipment consists of a widely used PC and commercial-off-the-shelf (COTS) I/O boards instead of a powerful computing system and custom-made I/O boards. The distributed structure of the HILS system supplements the lack of computing power. These features make the HILS equipment more cost-effective and flexible. The HILS uses an automatic code generation extension, REAL-TIME WORKSHOP$^{ (RTW$^{) of MATLAB$^{ tool-chain and RT-LAB$^{, which enables distributed simulation as well as the detection and generation of digital event between simulation time steps. The mean value engine model, which is used in control design phase, is imported into this HILS. The engine model is supplemented with some I/O subsystems and I/O boards to interface actual input and output signals in real-time. The I/O subsystems are designed to imitate real sensor signals with high fidelity as well as to convert the raw data of the I/O boards to the appropriate forms for proper interfaces. A lot of attention is paid to the generation of a precise crank/ earn signal which has the problem of quantization in a conventional fixed time step simulation. The detection of injection! command signal which occurs between simulation time steps are also successfully compensated. In order to prove the feasibility of the proposed environment, a simple PI controller for an air-to-fuel ratio (AFR) control is used. The proposed HILS environment and I/O systems are shown to be an efficient tool to develop various control functions and to validate the software and hardware of the engine control system.

• KIISE Transactions on Computing Practices
• /
• v.21 no.8
• /
• pp.531-536
• /
• 2015

The speech recognition is an active research area in the human computer interface (HCI). The objective of this study is to control digital devices with voices. In addition, the mouse is used as a computer peripheral tool which is widely used and provided in graphical user interface (GUI) computing environments. In this paper, we propose a method of controlling the mouse with the real-time speech recognition function of a smartphone. The processing steps include extracting the core voice signal after receiving a proper length voice input with real time, to perform the quantization by using the learned code book after feature extracting with mel frequency cepstral coefficient (MFCC), and to finally recognize the corresponding vowel using hidden markov model (HMM). In addition a virtual mouse is operated by mapping each vowel to the mouse command. Finally, we show the various mouse operations on the desktop PC display with the implemented smartphone application.

• Journal of Chest Surgery
• /
• v.45 no.5
• /
• pp.287-294
• /
• 2012

Background: Biologic valved grafts are important in cardiac surgery, and although several types of graft are currently available, most commercial xenografts tend to cause early disfiguration due to intimal proliferation and calcification. We studied the graft failure patterns on non-fixed and glutaraldehyde-fixed pulmonary xenograft in vivo animal experiment. Materials and Methods: Pulmonary valved conduits were obtained from the right ventricular outflow tract of eleven miniature pigs. The grafts were subjected to 2 different preservation methods; with or without glutaraldehyde fixation: glutaraldehyde fixation (n=7) and non-glutaraldehyde fixation (n=4). The processed explanted pulmonary valved grafts of miniature pig were then transplanted into eleven goats. Calcium quantization was achieved in all of the explanted xenograft, hemodynamic, histopathologic and radiologic evaluations were performed in the graft which the transplantation period was over 300 days (n=7). Results: Grafts treated with glutaraldehyde fixation had more calcification and conduit obstruction in mid-term period. Calcium deposition also appeared much higher in the glutaraldehyde treated graft compared to the non-glutaraldehyde treated graft (p<0.05). Conclusion: The present study suggests that xenografts prepared using glutaraldehyde fixation alone appeared to have severe calcification compared to the findings of non-glutaraldehyde treated xenografts and to be managed with proper anticalcification treatment and novel preservation methods. This experiment gives the useful basic chemical, histologic data of xenograft failure model with calcification for further animal study.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.24 no.1
• /
• pp.1-8
• /
• 2024

On-device AI technology, which can operate AI models at the edge devices to support real-time processing and privacy enhancement, is attracting attention. As intelligent IoT is applied to various industries, services utilizing the on-device AI technology are increasing significantly. However, general deep learning models require a lot of computational resources for inference and learning. Therefore, various lightweighting methods such as quantization and pruning have been suggested to operate deep learning models in embedded edge devices. Among the lightweighting methods, we analyze how to lightweight and apply deep learning models to edge computing devices, focusing on pruning technology in this paper. In particular, we utilize dynamic and static pruning techniques to evaluate the inference speed, accuracy, and memory usage of a lightweight AI vision model. The content analyzed in this paper can be used for intelligent video control systems or video security systems in autonomous vehicles, where real-time processing are highly required. In addition, it is expected that the content can be used more effectively in various IoT services and industries.

• Journal of Broadcast Engineering
• /
• v.27 no.1
• /
• pp.31-43
• /
• 2022

Traditional video compression has developed so far based on hybrid compression methods through motion prediction, residual coding, and quantization. With the rapid development of technology through artificial neural networks in recent years, research on image compression and video compression based on artificial neural networks is also progressing rapidly, showing competitiveness compared to the performance of traditional video compression codecs. In this paper, a new method capable of improving the performance of such an artificial neural network-based video compression model is presented. Basically, we take the rate-distortion optimization method using the auto-encoder and entropy model adopted by the existing learned video compression model and shifts some components of the latent information that are difficult for entropy model to estimate when transmitting compressed latent representation to the decoder side from the encoder side, and finally compensates the distortion of lost information. In this way, the existing neural network based video compression framework, MFVC (Motion Free Video Compression) is improved and the BDBR (Bjøntegaard Delta-Rate) calculated based on H.264 is nearly twice the amount of bits (-27%) of MFVC (-14%). The proposed method has the advantage of being widely applicable to neural network based image or video compression technologies, not only to MFVC, but also to models using latent information and entropy model.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.32 no.9C
• /
• pp.849-860
• /
• 2007

Because the H.264/AVC standard adopts the variable length coding algorithm, the rate of encoded video bitstream fluctuates a lot as time flows, though its compression efficiency is superior to that of existing standards. When a video is transmitted in real-time over networks with fixed low-bandwidth, it is necessary to control the bit rate which is generated from encoder. Many existing rate control algorithms have been adopting the quadratic rate-distortion model which determines the target bits for each frame. We propose a new rate control algorithm for H.264/AVC video transmission over networks with fixed bandwidth. The proposed algorithm predicts quantization parameter adaptively to reduce video distortion using the quadratic rate-distortion model, which uses the target bit rate and the mean absolute difference for current frame considering pixel difference between macroblocks in the previous and the current frame. On video samples with high motion and scene change cases, experimental results show that (1) the proposed algorithm adapts the encoded bitstream to limited channel capacity, while existing algorithms abruptly excess the limit bit rate; (2) the proposed algorithm improves picture quality with $0.4{\sim}0.9dB$ in average.

• KIISE Transactions on Computing Practices
• /
• v.20 no.11
• /
• pp.604-609
• /
• 2014

Decoding processes in portable media players have a high computational cost, resulting in high power consumption by the CPU. If decoding computations are reduced, the power consumed by the CPU is also be reduced, but such a choice generally results in a degradation of the video quality for the users, so it is essential to address this tradeoff. We proposed a new CPU power management scheme that can make use of the scalability property available in the H.164/SVC standard. We first proposed a new video quality model that makes use of a video quality metric(VQM) in order to efficiently take into account the different quantization factors in the SVC. We then propose a new dynamic voltage scaling(DVS) scheme that can selectively combine the previous decoding times and frame sizes in order to accurately predict the next decoding time. We then implemented a scheme on a commercial smartphone and performed a user test in order to examine how users react to the VQM difference. Real measurements show that the proposed scheme uses up to 34% fewer energy than the Linux DVFS governor, and user tests confirm that the degradation in the quality is quite tolerable.