• Vacuum Magazine
• /
• v.3 no.3
• /
• pp.15-18
• /
• 2016

In future wearable electronic systems, 3-dimensional (3D) devices have attracted much attention due to their high density integration and low-power functionality. Among 3D devices, gate-all-around (GAA) nanowire transistor provides superior gate controllability, resulting in suppressing short channel effect and other drawbacks in 2D metal-oxide-semiconductor field-effect transistor (MOSFET). Silicon nanowires (SiNWs) are the most promising building block for GAA structure device due to their compatibility with the current Si-based ultra large scale integration (ULSI) technology. Moreover, the theoretical limit for subthreshold swing (SS) of MOSFET is 60 mV/dec at room temperature, which causes the increase in Ioff current. To overcome theoretical limit for the SS, it is crucial that research into new types of device concepts should be performed. In our present studies, we have experimentally demonstrated feedback FET (FBFET) and tunnel FET (TFET) with sub-60 mV/dec based on SiNWs. Also, we fabricated SiNW based complementary TFET (c-TFET) and SiNW complementary metal-oxide-semiconductor (CMOS) inverter. Our research demonstrates the promising potential of SiNW electronic devices for future wearable electronic systems.

• ETRI Journal
• /
• v.41 no.3
• /
• pp.371-382
• /
• 2019

Machine reading comprehension is the task of understanding a given context and finding the correct response in that context. A simple recurrent unit (SRU) is a model that solves the vanishing gradient problem in a recurrent neural network (RNN) using a neural gate, such as a gated recurrent unit (GRU) and long short-term memory (LSTM); moreover, it removes the previous hidden state from the input gate to improve the speed compared to GRU and LSTM. A self-matching network, used in R-Net, can have a similar effect to coreference resolution because the self-matching network can obtain context information of a similar meaning by calculating the attention weight for its own RNN sequence. In this paper, we construct a dataset for Korean machine reading comprehension and propose an $S^2-Net$ model that adds a self-matching layer to an encoder RNN using multilayer SRU. The experimental results show that the proposed $S^2-Net$ model has performance of single 68.82% EM and 81.25% F1, and ensemble 70.81% EM, 82.48% F1 in the Korean machine reading comprehension test dataset, and has single 71.30% EM and 80.37% F1 and ensemble 73.29% EM and 81.54% F1 performance in the SQuAD dev dataset.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.11a
• /
• pp.559-561
• /
• 2022

인공지능 시대에 들어서면서 개인 맞춤형 환경을 제공하기 위하여 사람의 감정을 인식하고 교감하는 기술이 많이 발전되고 있다. 사람의 감정을 인식하는 방법으로는 얼굴, 음성, 신체 동작, 생체 신호 등이 있지만 이 중 가장 직관적이면서도 쉽게 접할 수 있는 것은 표정이다. 따라서, 본 논문에서는 정확도 높은 얼굴 감정 식별을 위해서 Convolution Block Attention Module(CBAM)의 각 Gate와 Residual Block, Skip Connection을 이용한 Identity- CBAM Module을 제안한다. CBAM의 각 Gate와 Residual Block을 이용하여 각각의 표정에 대한 핵심 특징 정보들을 강조하여 Context 한 모델로 변화시켜주는 효과를 가지게 하였으며 Skip-Connection을 이용하여 기울기 소실 및 폭발에 강인하게 해주는 모듈을 제안한다. AI-HUB의 한국인 감정 인식을 위한 복합 영상 데이터 세트를 이용하여 총 6개의 클래스로 구분하였으며, F1-Score, Accuracy 기준으로 Identity-CBAM 모듈을 적용하였을 때 Vanilla ResNet50, ResNet101 대비 F1-Score 0.4~2.7%, Accuracy 0.18~2.03%의 성능 향상을 달성하였다. 또한, Guided Backpropagation과 Guided GradCam을 통해 시각화하였을 때 중요 특징점들을 더 세밀하게 표현하는 것을 확인하였다. 결과적으로 이미지 내 표정 분류 Task에서 Vanilla ResNet50, ResNet101을 사용하는 것보다 Identity-CBAM Module을 함께 사용하는 것이 더 적합함을 입증하였다.

• The Journal of the Acoustical Society of Korea
• /
• v.41 no.5
• /
• pp.518-524
• /
• 2022

In this paper, we propose a method for combining convolutional neural networks and attention mechanism to improve the recognition performance of emotional stress from Electroencephalogram (EGG) signals. In the proposed method, EEG signals are decomposed into five frequency domains, and spatial information of EEG features is obtained by applying a convolutional neural network layer to each frequency domain. As a next step, salient frequency information is learned in each frequency band using a gate transformer-based attention mechanism, and complementary frequency information is further learned through inter-frequency mapping to reflect it in the final attention representation. Through an EEG stress recognition experiment involving a DEAP dataset and six subjects, we show that the proposed method is effective in improving EEG-based stress recognition performance compared to the existing methods.

• Proceedings of the Materials Research Society of Korea Conference
• /
• 2011.10a
• /
• pp.20.1-20.1
• /
• 2011

The nano-sized quantum structure has been an attractive candidate for investigations of the fundamental physical properties and potential applications of next-generation electronic devices. Metal nano-particles form deep quantum wells between control and tunnel oxides due to a difference in work functions. The charge storage capacity of nanoparticles has led to their use in the development of nano-floating gate memory (NFGM) devices. When compared with conventional floating gate memory devices, NFGM devices offer a number of advantages that have attracted a great deal of attention: a greater inherent scalability, better endurance, a faster write/erase speed, and more processes that are compatible with conventional silicon processes. To improve the performance of NFGM, metal nanocrystals such as Au, Ag, Ni Pt, and W have been proposed due to superior density, a strong coupling with the conduction channel, a wide range of work function selectivity, and a small energy perturbation. In the present study, bismuth metal nanocrystals were self-assembled within high-k $Bi_2Mg_{2/3}Nb_{4/3}O_7$ (BMN) films grown at room temperature in Ar ambient via radio-frequency magnetron sputtering. The work function of the bismuth metal nanocrystals (4.34 eV) was important for nanocrystal-based nonvolatile memory (NVM) applications. If transparent NFGM devices can be integrated with transparent solar cells, non-volatile memory fields will open a new platform for flexible electron devices.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.16 no.3
• /
• pp.15-27
• /
• 2017

In general, increasing train dwell time leads to increasing train service frequency, and it in turn contributes to increasing the congestion level of train and platform. Therefore, the studies on train dwell time have received growing attention in the perspective of scheduling train operation. This study develops a prediction model of train dwell time to enable train operators to mitigate platform congestion by metering passenger inflow at platform gate with respect to platform congestion levels in real-time. To estimate the prediction model, three types of independent variables were applied: number of passengers to get into train, number of passengers to get out of trains, and train weights, which are collectable in real-time. The explanatory power of the estimated model was 0.809, and all of the dependent variables were statistically significant at the 99%. As a result, this model can be available for the basis of on-time train service through platform gate metering, which is a strategy to manage passenger inflow at the platform.

• Proceedings of the Korean Vacuum Society Conference
• /
• 2016.02a
• /
• pp.268-268
• /
• 2016

Chemical sensors have attracted much attention due to their various applications such as agriculture product, cosmetic and pharmaceutical components and clinical control. A conventional chemical and biological sensor is consists of fluorescent dye, optical light sources, and photodetector to quantify the extent of concentration. Such complicated system leads to rising cost and slow response time. Until now, the most contemporary thin film transistors (TFTs) are used in the field of flat panel display technology for switching device. Some papers have reported that an interesting alternative to flat panel display technology is chemical sensor technology. Recent advances in chemical detection study for using TFTs, benefits from overwhelming progress made in organic thin film transistors (OTFTs) electronic, have been studied alternative to current optical detection system. However numerous problems still remain especially the long-term stability and lack of reliability. On the other hand, the utilization of metal oxide transistor technology in chemical sensors is substantially promising owing to many advantages such as outstanding electrical performance, flexible device, and transparency. The top-gate structure transistor indicated long-term atmosphere stability and reliability because insulator layer is deposited on the top of semiconductor layer, as an effective mechanical and chemical protection. We report on the fabrication of InGaZnO TFTs with silver nanowire as the top gate electrode for the aim of chemical materials detection by monitoring change of electrical properties. We demonstrated that the improved sensitivity characteristics are related to the employment of a unique combination of nano materials. The silver nanowire top-gate InGaZnO TFTs used in this study features the following advantages: i) high sensitivity, ii) long-term stability in atmosphere and buffer solution iii) no necessary additional electrode and iv) simple fabrication process by spray.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2008.06a
• /
• pp.477-477
• /
• 2008

For more than three decades, the gate dielectrics in CMOS devices are $SiO_2$ because of its blocking properties of current in insulated gate FET channels. As the dimensions of feature size have been scaled down (width and the thickness is reduced down to 50 urn and 2 urn or less), gate leakage current is increased and reliability of $SiO_2$ is reduced. Many metal oxides such as $TiO_2$, $Ta_2O_4$, $SrTiO_3$, $Al_2O_3$, $HfO_2$ and $ZrO_2$ have been challenged for memory devices. These materials posses relatively high dielectric constant, but $HfO_2$ and $Al_2O_3$ did not provide sufficient advantages over $SiO_2$ or $Si_3N_4$ because of reaction with Si substrate. Recently, $HfO_2$ have been attracted attention because Hf forms the most stable oxide with the highest heat of formation. In addition, Hf can reduce the native oxide layer by creating $HfO_2$. However, new gate oxide candidates must satisfy a standard CMOS process. In order to fabricate high density memories with small feature size, the plasma etch process should be developed by well understanding and optimizing plasma behaviors. Therefore, it is necessary that the etch behavior of $HfO_2$ and plasma parameters are systematically investigated as functions of process parameters including gas mixing ratio, rf power, pressure and temperature to determine the mechanism of plasma induced damage. However, there is few studies on the the etch mechanism and the surface reactions in $BCl_3$ based plasma to etch $HfO_2$ thin films. In this work, the samples of $HfO_2$ were prepared on Si wafer with using atomic layer deposition. In our previous work, the maximum etch rate of $BCl_3$/Ar were obtained 20% $BCl_3$/ 80% Ar. Over 20% $BCl_3$ addition, the etch rate of $HfO_2$ decreased. The etching rate of $HfO_2$ and selectivity of $HfO_2$ to Si were investigated with using in inductively coupled plasma etching system (ICP) and $BCl_3/Cl_2$/Ar plasma. The change of volume densities of radical and atoms were monitored with using optical emission spectroscopy analysis (OES). The variations of components of etched surfaces for $HfO_2$ was investigated with using x-ray photo electron spectroscopy (XPS). In order to investigate the accumulation of etch by products during etch process, the exposed surface of $HfO_2$ in $BCl_3/Cl_2$/Ar plasma was compared with surface of as-doped $HfO_2$ and all the surfaces of samples were examined with field emission scanning electron microscopy and atomic force microscope (AFM).

• Journal of Electrical Engineering and Technology
• /
• v.7 no.4
• /
• pp.601-605
• /
• 2012

Insulated Gate Bipolar Transistors(IGBTs) have received wide attention because of their high current conduction and good switching characteristics. To reduce the power loss of IGBT, the onstate voltage drop should be lowered and the switching time should be shortened. However, there is trade-off between the breakdown voltage and the on-state voltage drop. The FLoatingIsland(FLI) structure can lower the on-state voltage drop without reducing breakdown voltage. In this paper, The FLI IGBT shows an on-state voltage drop that is 22.5% lower than the conventional IGBT, even though the breakdown voltages of each IGBT are almost identical.

• Proceedings of the KIEE Conference
• /
• 2011.07a
• /
• pp.1306-1307
• /
• 2011

Recently, the small scale photovoltaic (PV) electronic devices are drawing attention as the upcoming PV generation system. The PV system is commonly used in small scale PV applications such as LED lighting and cell phone. This paper proposes photovoltaic output sensorless (POS) maximum power point tracking (MPPT) control for a small scale charging system of PV-nickel metal hydride battery using field-programmable gate array (FPGA) controller. A converter is connected to a small scale PV cell and battery, and performs the POS MPPT at the battery terminal current instead of being at the PV cell output voltage and current. The FPGA controller and converter operate based on POS MPPT method. The experimental results show that the nickel metal hydride battery is charged by the maximum PV output power.