• Journal of the Institute of Electronics and Information Engineers
• /
• v.50 no.4
• /
• pp.3-8
• /
• 2013

In CDMA-based systems, recently, researches on chip-level equalization have been studied in order to improve receiving performance when supporting high-rate data services. In this paper, we consider a chip-level LMMSE (linear minimum mean-squared error) receiver for D-TxAA (dual stream transmit antenna array) based single stream HSDPA MIMO systems using precoding weights. First, we will derive precoding weights for maximizing the total instantaneous received power. We will also analyze the effects of both transmit delay of precoding weights and mobile velocity on chip-level LMMSE receivers, which is verified through computer simulations in various mobile channel environments.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2005.06a
• /
• pp.10-14
• /
• 2005

This report deals with a feeding system of the Coaxal grinding machine, processing optical ferrule. This report also examines the applicability of using the feeding system for the Coaxial grinding machine, by mean of conducting performance estimation. The results are as follow; Repeatability of regulating wheel is $17{\mu}m$, R/W rotation accuracy is between $30\;\~\;40{\mu}m$. This means 'Rotation accuracy' is lower than the concentricity level. Backlash generation level at the feeding system of the grinding wheel is under $1{\mu}m$, thereby positioning accuracy is controlled within $2{\mu}m$ In terms of repeatability, you can find occasional error at the returning process from the starting point. This error is resulted from the measurement tolerance of the starting point sensor. We will get the repeatability level under control by $1{\mu}m$, through improving the soft-ware used and up-grading the sensor at the starting point.

• The Korean Journal of Applied Statistics
• /
• v.19 no.1
• /
• pp.111-120
• /
• 2006

In a two-stage two-level balanced nested design, type I error rates for the parametric tests and the rank transformed tests for the main effects and the nested effects are in overall similar to each other. Furthermore, powers for the rank transformed statistic for the main effects and the nested effects in a two-stage two-level balanced nested design are generally superior to powers for the parametric statistic When the effect size and the sample size are increased, we can find that powers increase for the parametric statistic and the rank transformed statistic are dramatically improved. Especially for the case of the fixed effects in the asymmetric distributions such as an exponential distribution, powers for the rank transformed tests are quite high rather than powers for the parametric tests.

• The Journal of the Acoustical Society of Korea
• /
• v.22 no.3
• /
• pp.194-201
• /
• 2003

In speech recognition, confidence measuring is to decide whether it can be accepted as the recognized results or not. The confidence is measured by integrating frames into phone and word level. In case of word recognition, the confidence measuring verifies the results of recognition and Out-Of-Vocabulary (OOV). Therefore, the post-processing could improve the performance of recognizer without accepting it as a recognition error. In this paper, we measure the confidence modifying log likelihood ratio (LLR) which was the previous confidence measuring. It accumulates only those which the log likelihood ratio is negative when integrating the confidence to phone level from frame level. When comparing the verification performance for the results of word recognizer with the previous method, the FAR (False Acceptance Ratio) is decreased about 3.49% for the OOV and 15.25% for the recognition error when CAR (Correct Acceptance Ratio) is about 90%.

• Journal of Korean Society for Atmospheric Environment
• /
• v.24 no.5
• /
• pp.604-612
• /
• 2008

The ozone concentration is one of the important environmental issue for measurement of the atmospheric condition of the country. This study focuses on applying the Autoregressive Error (ARE) model for analyzing the ozone data at middle part of the Gyeonggi-Do, Anyang monitoring site in Korea. In the ARE model, eight meteorological variables and four pollution variables are used as the explanatory variables. The eight meteorological variables are daily maximum temperature, wind speed, amount of cloud, global radiation, relative humidity, rainfall, dew point temperature, and water vapor pressure. The four air pollution variables are sulfur dioxide $(SO_2)$, nitrogen dioxide $(NO_2)$, carbon monoxide (CO), and particulate matter 10 (PM10). The result shows that ARE models both overall and monthly data are suited for describing the oBone concentration. In the ARE model for overall ozone data, ozone concentration can be explained about 71% to by the PM10, global radiation and wind speed. Also the four types of ARE models for high level of ozone data (over 80 ppb) have been analyzed. In the best ARE model for high level of ozone data, ozone can be explained about 96% by the PM10, daliy maximum temperature, and cloud amount.

• 정보저장시스템학회:학술대회논문집
• /
• 2005.10a
• /
• pp.119-123
• /
• 2005

We report the random pattern characteristics of the super resolution near field structure(Super-RENS) write once read-many(WORM) disc at a blue laser optical system(laser wavelength 405nm, numerical aperture 0.85) and the Super-RENS read only memory(ROM) disc at a blue laser optical system(laser wavelength 659nm, numerical aperture 0.65). We used the WORM disc of which carrier-to-noise ratio (CNR) of 75nm is 47dB and ROM disc of which carrier-to-noise ratio (CNR) of 173nm is 45dB. We controlled the equalization (EQ) characteristics and used advanced partial-response maximum likelihood (PRML) technique. We obtained bit error rate (bER) of 10-3 level at 50GB WORM disc and bite error rate of 10-4 level at 50GB level ROM disc. This result shows high feasibility of Super-RENS technology for practical use.

• Journal of the Korean Data and Information Science Society
• /
• v.19 no.4
• /
• pp.1153-1164
• /
• 2008

The ozone data is one of the important environmental data for measurement of the atmospheric condition of the country. In this article, the Autoregressive Error (ARE) model have been considered for analyzing the ozone data at the northern part of the Gyeonggi-Do, Uijeongbu monitoring site in Korea. The result showed that both overall and monthly ARE models are suited for describing the ozone concentration. In the ARE model, seven meteorological variables and four pollution variables are used as the as the explanatory variables for the ozone data set. The seven meteorological variables are daily maximum temperature, wind speed, relative humidity, rainfall, dew point temperature, steam pressure, and amount of cloud. The four air pollution explanatory variables are Sulfur dioxide(SO2), Nitrogen dioxide(NO2), Cobalt(CO), and Promethium 10(PM10). Also, the high level ozone data (over 80ppb) have been analyzed four ARE models, General ARE, HL ARE, PM10 add ARE, Temperature add ARE model. The result shows that the General ARE, HL ARE, and PM10 add ARE models are suited for describing the high level of ozone data.

• Journal of Communications and Networks
• /
• v.18 no.3
• /
• pp.333-339
• /
• 2016

A multi-stage turbo equalization scheme based on the bit-level combining (BLC) is proposed for multiple-input multiple-output (MIMO) systems with hybrid automatic repeat request (HARQ). In the proposed multi-stage turbo equalization scheme, the minimum mean-square-error equalizer at each iteration calculates the extrinsic log-likelihood ratios for the transmitted bits in a subpacket and the subpackets are sequentially replaced at each iteration according to the HARQ rounds of received subpackets. Therefore, a number of iterations are executed for different subpackets received at several HARQ rounds, and the transmitted bits received at the previous HARQ rounds as well as the current HARQ round can be estimated from the combined information up to the current HARQ round. In addition, the proposed multi-stage turbo equalization scheme has the same computational complexity as the conventional bit-level combining based turbo equalization scheme. Simulation results show that the proposed multi-stage turbo equalization scheme outperforms the conventional BLC based turbo equalization scheme for MIMO systems with HARQ.

• Transactions of the Society of Information Storage Systems
• /
• v.2 no.1
• /
• pp.38-42
• /
• 2006

We report the random pattern characteristics of the super resolution near field structure(Super-RENS) write once read-many(WORM) disc at a blue laser optical system(laser wavelength 405nm, numerical aperture 0.85) and the Super-RENS read only memory(ROM) disc at a blue laser optical system(laser wavelength 659nm, numerical aperture 0.65). We used the WORM disc of which carrier-to-noise ratio(CNR) of 75nm is 47dB and ROM disc of which carrier-to-noise ratio(CNR) of 173nm is 45dB. We controlled the equalization(EQ) characteristics and used advanced partial-response maximum likelihood(PRML) technique. We obtained bit error rate(bER) of 10-3 level at 50GB WORM disc and bite error rate of 10-4 level at 50GB level ROM disc. This result shows high feasibility of Super-RENS technology for practical use.

• International journal of advanced smart convergence
• /
• v.8 no.1
• /
• pp.141-146
• /
• 2019

These days, there are many applications using neural networks as parts of their system. On the other hand, adversarial examples have become an important issue concerining the security of neural networks. A classifier in neural networks can be fooled and make it miss-classified by adversarial examples. There are many research to encounter adversarial examples by using denoising methods. Some of them using GAN (Generative Adversarial Network) in order to remove adversarial noise from input images. By producing an image from generator network that is close enough to the original clean image, the adversarial examples effects can be reduced. However, there is a chance when adversarial noise can survive the approximation process because it is not like a normal noise. In this chance, we propose a research that utilizes high-level representation in the classifier by combining GAN network with a trained U-Net network. This approach focuses on minimizing the loss function on high representation terms, in order to minimize the difference between the high representation level of the clean data and the approximated output of the noisy data in the training dataset. Furthermore, the generated output is checked whether it shows minimum error compared to true label or not. U-Net network is trained with true label to make sure the generated output gives minimum error in the end. At last, the remaining adversarial noise that still exist after low-level approximation can be removed with the U-Net, because of the minimization on high representation terms.