• Annual Conference of KIPS
• /
• 2022.11a
• /
• pp.55-56
• /
• 2022

Hypertension is a severe health problem and increases the risk of other health issues, such as heart disease, heart attack, and stroke. In this research, we propose a machine learning-based prediction method for the risk of chronic hypertension. The proposed method consists of four main modules. In the first module, the linear interpolation method fills missing values of the integration of gas and meteorological datasets. In the second module, the OrdinalEncoder-based normalization is followed by the Decision tree algorithm to select important features. The prediction analysis module builds three models based on k-Nearest Neighbors, Decision Tree, and Random Forest to predict hypertension levels. Finally, the features used in the prediction model are explained by the DeepSHAP approach. The proposed method is evaluated by integrating the Korean meteorological agency dataset, natural gas leakage dataset, and Korean National Health and Nutrition Examination Survey dataset. The experimental results showed important global features for the hypertension of the entire population and local components for particular patients. Based on the local explanation results for a randomly selected 65-year-old male, the effect of hypertension increased from 0.694 to 1.249 when age increased by 0.37 and gas loss increased by 0.17. Therefore, it is concluded that gas loss is the cause of high blood pressure.

• KIPS Transactions on Software and Data Engineering
• /
• v.8 no.11
• /
• pp.441-448
• /
• 2019

Previous researches on automatic spacing of Korean sentences has been researched to correct spacing errors by using n-gram based statistical techniques or morpheme analyzer to insert blanks in the word boundary. In this paper, we propose an end-to-end automatic word spacing by using deep neural network. Automatic word spacing problem could be defined as a tag classification problem in unit of syllable other than word. For contextual representation between syllables, Bi-LSTM encodes the dependency relationship between syllables into a fixed-length vector of continuous vector space using forward and backward LSTM cell. In order to conduct automatic word spacing of Korean sentences, after a fixed-length contextual vector by Bi-LSTM is classified into auto-spacing tag(B or I), the blank is inserted in the front of B tag. For tag classification method, we compose three types of classification neural networks. One is feedforward neural network, another is neural network language model and the other is linear-chain CRF. To compare our models, we measure the performance of automatic word spacing depending on the three of classification networks. linear-chain CRF of them used as classification neural network shows better performance than other models. We used KCC150 corpus as a training and testing data.

• Journal of the Korean Society for Precision Engineering
• /
• v.26 no.8
• /
• pp.72-78
• /
• 2009

We developed a line standards measurement system using an optical microscope and measured two kinds of line standards. It consists of three main parts: an optical microscope module including a CCD camera, a stage system with a linear encoder, and a measurement program for a microscopic image processing. The magnification of microscope part was calibrated using one-dimensional gratings and the angular motion of stage was measured to estimate the Abbe error. The threshold level in line width measurement was determined by comparing with certified values of a line width reference specimen, and its validity was proved through the measurement of another line width specimen. The expanded uncertainty (k=2) was about 100 nm in the measurements of $1{\mu}m{\sim}10{\mu}m$ line width. In the comparison results of line spacing measurement, two kinds of values were coincide within the expanded uncertainty, which were obtained by the one-dimensional measuring machine in KRISS and the line standards measurement system. The expanded uncertainty (k=2) in the line spacing measurement was estimated as $\sqrt{(0.098{\mu}m)^2+(1.8{\times}10^{-4}{\times}L)^2}$. Therefore, it will be applied effectively to the calibration of line standards, such as line width and line spacing, with the expanded uncertainty of several hundreds nanometer.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.7C
• /
• pp.962-970
• /
• 2004

This paper presents an efficient implementation of the 2.4 kbps EHSX(Enhanced Harmonic Stochastic Excitation) speech coder on a TMS320C6701$^{TM}$ floating-point digital signal processor. The EHSX speech codec is based on a harmonic and CELP(Code Excited Linear Prediction) modeling of the excitation signal respectively according to the frame characteristic such as a voiced speech and an unvoiced speech. In this paper, we represent the optimization methods to reduce the complexity for real-time implementation. The complexity in the filtering of a CELP algorithm that is the main part for the EHSX algorithm complexity can be reduced by converting program using floating-point variable to program using fixed-point variable. We also present the efficient optimization methods including the code allocation considering a DSP architecture and the low complexity algorithm of harmonic/pitch search in encoder part. Finally, we obtained the subjective quality of MOS 3.28 from speech quality test using the PESQ(perceptual evaluation of speech quality), ITU-T Recommendation P.862 and could get a goal of realtime operation of the EHSX codec.c.

• KIPS Transactions on Software and Data Engineering
• /
• v.6 no.12
• /
• pp.573-580
• /
• 2017

Deep Neural Network (DNN) is a large layered neural network which is consisted of a number of layers of non-linear units. Deep Learning which represented as DNN has been applied very successfully in various applications. However, many issues in DNN have been identified through past researches. Among these issues, generalization is the most well-known problem. A Recent study, Dropout, successfully addressed this problem. Also, Dropout plays a role as noise, and so it helps to learn robust feature during learning in DNN such as Denoising AutoEncoder. However, because of a large computations required in Dropout, training takes a lot of time. Since Dropout keeps changing an inter-layer representation during the training session, the learning rates should be small, which makes training time longer. In this paper, using mutation operation, we reduce computation and improve generalization performance compared with Dropout. Also, we experimented proposed method to compare with Dropout method and showed that our method is superior to the Dropout one.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2006.11a
• /
• pp.127-132
• /
• 2006

Quantization in video coding plays an important role in controlling the bit-rate of compressed video bit-streams. It has been used as an important control means to adjust the amount of bit-streams to at]owed bandwidth of delivery networks and storage. Due to the dependent nature of video coding, dependent quantization has been proposed and applied for MPEG-2 video coding to better maintain the quality of reconstructed frame for given constraints of target bit-rate. Since Scalable Video Coding (SVC) being currently standardized exhibits highly dependent coding nature not only between frames but also lower and higher scalability layers where the dependent quantization can be effectively applied, in this paper, we propose a dependent quantization scheme for SVC and compare its performance in visual qualities and bit-rates with the current JSVM reference software for SVC. The proposed technique exploits the frame dependences within each GOP of SVC scalability layers to formulate dependent quantization. We utilize Lagrange optimization, which is widely accepted in R-D (rate-distortion) based optimization, and construct trellis graph to find the optimal cost path in the trellis by minimizing the R-D cost. The optimal cost path in the trellis graph is the optimal set of quantization parameters (QP) for frames within a GOP. In order to reduce the complexity, we employ pruning procedure using monotonicity property in the trellis optimization and cut the frame dependency into one GOP to decrease dependency depth. The optimal Lagrange multiplier that is used for SVC is equal to H.264/AVC which is also used in the mode prediction of the JSVM reference software. The experimental result shows that the dependent quantization outperforms the current JSVM reference software encoder which actually takes a linear increasing QP in temporal scalability layers. The superiority of the dependent quantization is achieved up to 1.25 dB increment in PSNR values and 20% bits saving for the enhancement layer of SVC.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.45 no.12
• /
• pp.13-19
• /
• 2008

In this paper, we propose a new CL-OSTBC (Closed Loop Orthogonal Space-Time Block Code) scheme for four transmit antennas and compare the scheme with existing closed loop schemes on the performance of BER (Bit Error Rate). In the proposed scheme, a transmitter receives channel feedback information and combines modulated symbols by the symbol combiner, and transmits the symbols encoded by the space-time block encoder. As a result, the proposed scheme achieves full-rate and maximal channel gains by more efficient utilization of the channel feedback information. Moreover, the scheme can reduce computation complexity by using a linear detector. Simulation results on the BER performance show that the proposed CL-OSTBC scheme outperforms existing CL-OSTBC schemes.

• International Journal of Precision Engineering and Manufacturing
• /
• v.8 no.2
• /
• pp.26-31
• /
• 2007

A novel scanning probe measurement system was developed to enable precise profile measurements of microaspheric surfaces. An air-bearing stylus with a microprobe was used to perform the surface profile scanning. The new system worked in a contact mode and had the capability of measuring micro-aspheric surfaces with large tilt angles and complex profiles. Due to limitations resulting from the contact mode, such as possible damage caused by the contact force and lateral resolution restrictions from the curvature of the probe tip, several system improvements were implemented. An air bearing was used to suspend the shaft of the probe to reduce the contact force, enabling fine adjustments of the contact force by changing the air pressure. The movement of the shaft was measured by a linear encoder with a scale attached to the actual shaft to avoid Abbe errors. A $50-{\mu}m-diameter$ glass sphere was bonded to the tip of the probe to improve the lateral resolution of the system. The maximum contact force of the probe was 10 mN. The shaft was capable of holding the probe continuously if the contact force was less than 40 mN, and the resolution of the probe could be as high as 10 nm, The performance of the new scanning probe measurement system was verified by experimental data.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.26 no.11A
• /
• pp.1925-1938
• /
• 2001

In this paper, a new block-matching algorithm for standard video encoder is presented. The slice competition method is proposed as a new scheme, as opposed to a coarse-to-fine approach. The order of calculating the SAD(Sum of Absolute Difference) to fad the best matching block is changed from a raster order to a dispersed one. Based on this scheme, the increasing SAD curve during its calculation is more linear than that of other curves. Then, the candidates of low probability can be removed in the early stage of calculation. And new MV prediction technique with an adaptive search range scheme also assists the proposed block-matching algorithm. As a result, an average of 13% improvement in computational power is recorded by only the proposed MV prediction technique. Synthetically, the computational power is reduced by 3977∼77% than that of the conventional BMAs. The average MAD is always low in various sequences. The results are also very close to the MAD of the full search block-matching algorithm.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.22 no.2
• /
• pp.123-128
• /
• 2022

The torque of the SRM((Switched Reluctance Motor)) is proportional to the inductance slope, so it has a non-linear torque characteristic, and has a disadvantage in that the torque pulsation is large and noise is severe. In particular, the biggest obstacle to the commercialization of SRM is the pulsating torque generated from the rotating shaft, which has various adverse effects not only on the device itself but also on the peripheral devices. Therefore, various methods for reducing the pulsating torque have been published by domestic and foreign researchers, and there is a study result that the hysteresis controller has an advantage in that it can flow a smooth current compared to the chopping control. However, in determining the hysteresis band, if the band is too small, it has a disadvantage in that it may cause a switching loss due to many switching and an unstable initial start when the encoder is used. Therefore, in this paper, a variable hysteresis controller that can reduce torque ripple in a steady state while having a more stable and fast speed response through the change of the hysteresis band according to the speed error.