• International Journal of Internet, Broadcasting and Communication
• /
• v.15 no.3
• /
• pp.94-102
• /
• 2023

Anomaly detection in human movements can improve safety in indoor workplaces. In this paper, we design a framework for detecting anomalous trajectories of humans in indoor spaces based on a variational autoencoder (VAE) with Bi-LSTM layers. First, the VAE is trained to capture the latent representation of normal trajectories. Then the abnormality of a new trajectory is checked using the trained VAE. In this step, the anomaly score of the trajectory is determined using the trajectory reconstruction error through the VAE. If the anomaly score exceeds a threshold, the trajectory is detected as an anomaly. To select the anomaly threshold, a new metric called D-score is proposed, which measures the difference between recall and precision. The anomaly threshold is selected according to the minimum value of the D-score on the validation set. The MIT Badge dataset, which is a real trajectory dataset of workers in indoor space, is used to evaluate the proposed framework. The experiment results show that our framework effectively identifies abnormal trajectories with 81.22% in terms of the F1-score.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.31 no.8
• /
• pp.85-90
• /
• 2003

To control spacecraft including satellite, we should understand precisely the performance of propulsion system and the program logic with appropriate format for satellite operations. In this study, the thruster performance functions was generated by using the best curve fitting for performance data from bi-propellant thrusters. Detailed thruster performance data are, in general, company proprietary information, therefore real firing tests were performed to understand the basic characteristics of the performance curve. Experimental rocket motor utilize liquid oxygen and kerosine as propellant and designed average thrust was 100 pound.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2007.06a
• /
• pp.439-440
• /
• 2007

To enhance the electron injection from the cathode of organic light-emitting diodes (OLEDs), We have studied characteristics of device that electron injection layer(EIL) is inserted between emissive layer and cathode. We fabricated bi-layer cathode $Li_2O$(x nm)/Al(100nm) and LiF(x nm)/Al(100nm) using LiF and $Li_2O$ as an electron injection layer. We analyzed the current efficiency, luminance efficiency, and external quantum efficiency of the device by varying the thickness of $Li_2O$ and LiF to be 0.5nm, 1nm, or 3nm. Using the EIL, we have obtained the efficiency of 7cd/A and the luminance of $20,000cd/m^2$. There is an improvement of efficiency by more than 3 times than the device without the $Li_2O$ layer.

• Journal of the Korean Institute of Electrical and Electronic Material Engineers
• /
• v.23 no.12
• /
• pp.936-941
• /
• 2010

In this study, we have investigated the effects of Mn dopant on the bulk trap levels and grain boundary characteristics of $Bi_2O_3$-based ZnO (ZB) varistor using admittance spectroscopy and dielectric functions (such as $Z^*,\;Y^*,\;M^*,\;\varepsilon^*$, and $tan\delta$). Admittance spectra and dielectric functions show two bulk traps of $Zn_i^{..}$ (0.20 eV) and $V^{\bullet}_o$ (0.29~0.33 eV) in ZnO-$Bi_2O_3-Mn_3O_4$ (ZBM). The barrier of grain boundaries in ZBM could be electrochemically single type. However, its thermal stability was slightly disturbed by ambient oxygen because the apparent activation energy of grain boundaries was changed from 0.79 eV at lower temperature to 1.08 eV at higher temperature. The grain boundary capacitance $C_{gb}$ was decreased slightly with temperature as 1.3~1.8 nF but resistance $R_{gb}$ decreased exponentially. The relaxation time distribution can result from the heterogeneity of the barriers constituting the varistor. It is revealed that Mn dopant in ZB reduced the heterogeneity of the barrier in grain boundaries and stabilized the barrier against the ambient temperature.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.7
• /
• pp.1759-1772
• /
• 2023

Chinese named entity recognition (NER) is a challenging work that seeks to find, recognize and classify various types of information elements in unstructured text. Due to the Chinese text has no natural boundary like the spaces in the English text, Chinese named entity identification is much more difficult. At present, most deep learning based NER models are developed using a bidirectional long short-term memory network (BiLSTM), yet the performance still has some space to improve. To further improve their performance in Chinese NER tasks, we propose a new NER model, IDCNN-BiLSTM-Highway, which is a combination of the BiLSTM, the iterated dilated convolutional neural network (IDCNN) and the highway network. In our model, IDCNN is used to achieve multiscale context aggregation from a long sequence of words. Highway network is used to effectively connect different layers of networks, allowing information to pass through network layers smoothly without attenuation. Finally, the global optimum tag result is obtained by introducing conditional random field (CRF). The experimental results show that compared with other popular deep learning-based NER models, our model shows superior performance on two Chinese NER data sets: Resume and Yidu-S4k, The F1-scores are 94.98 and 77.59, respectively.

• Journal of the Korean Institute of Electrical and Electronic Material Engineers
• /
• v.30 no.5
• /
• pp.284-287
• /
• 2017

The electrocaloric effect in $0.94(Bi_{0.5}Na_{0.5})TiO_3+0.06KNbO3+0.9wt%$ G.F.ferroelectricceramics was observed in terms of the temperature change (${\Delta}T$) of the fabricated ceramics, Curie temperature $T_c$, and applied electric field. The specimens were fabricated by a conventional solid-state reaction. $T_c$ appeared near $165{\sim}170^{\circ}C$. The P-E hysteresis showed a tendency to slim down with a temperature increase and finally was slimmest near $150^{\circ}C$. With the increase of temperature, the polarization revealed a gradual decrease, and a sharp decline near $T_c$. When an electric field of 45 kV/cm was applied, the largest polarization was shown. The maximum value of the temperature change (${\Delta}T=0.31^{\circ}C$) was obtained at $165^{\circ}C$ under an applied electric field of 45 kV/cm.

• Journal of the Earthquake Engineering Society of Korea
• /
• v.10 no.3 s.49
• /
• pp.47-55
• /
• 2006

Seismic performance and retrofit of reinforced concrete (RC) two-column piers widely used at roadway bridges in Korea was experimentally evaluated. Ten two-column piers that were 400 mm in diameter and 2,000 mm in height were constructed. These piers were subjected to hi-directional cyclic loadings under a constant axial load of $0.1f_{ck}A_g$. Test parameters were the confinement steel ratio, loading pattern, lap splice of longitudinal reinforcing bars, and retrofitting method. Specimens with lap-spliced longitudinal bars were retrofitted with steel jacket, pre-stressing steel wire, and steel band. Test result showed that while the specimens subjected to bi-directional lateral cyclic loadings which consisted of two main amplitudes in the transverse axis and two sub amplitudes in longitudinal axis, referred to as a T-series cyclic loadings, exhibited plastic hinges both at the top and bottom parts of the column, the specimens subjected to bi-directional lateral cyclic loadings in an opposite way, referred to as a L-series cyclic loadings, exhibited a plastic hinge only at the bottom of the column. The displacement ductility of the specimen under the T-series loadings was bigger than that of the specimen under the L-series loadings. Specimen retrofitted with pre-stressing steel wires exhibited poor ductility due to the upward shift of the plastic hinge region because of over-reinforcement, but specimens retrofitted with steel jacket and steel band showed the required displacement ductility. Steel band can be an effective retrofitting scheme to improve the seimsic performance of RC bridge piers, considering its practical construction.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2000.11a
• /
• pp.563-566
• /
• 2000

The effect of CuO and CuO-B $i_2$ $O_3$ additives on microwave dielectric properties of (P $b_{0.45}$C $a_{0.55}$)[F $e_{0.5}$N $b_{0.5}$)$_{0.9}$S $n_{0.1}$] $O_3$were investigated to decrease the sintering temperature for usage of Low Temperature Co-firing Ceramics (LTCC). The (P $b_{0.45}$C $a_{0.55}$)[F $e_{0.5}$N $b_{0.5}$)$_{0.9}$S $n_{0.1}$] $O_3$ceramics was sintered at 11$65^{\circ}C$. In order to decrease the sintering temperature, CuO and Cuo-B $i_2$ $O_3$ were added in the (Pb,Ca)[(Fe,Nb)Sn] $O_3$ with CuO-B $i_2$ $O_3$. For the addition of 0.4 wt.% CuO, the sintered density and the dielectric constant of the ceramics were revealed the maximum values of the 6.06g/c $m^2$ and 83 respectively and temperature coefficient of resonance frequency ($\tau$$_{f}$) shifted to the positive value. As increasing B $i_2$ $O_3$to the (Pb,Ca)[(Fe,Nb)Sn] $O_3$ with CuO-B $i_2$ $O_3$with 0.2 wt.% CuO, the sintered density, the $\varepsilon$$_{r}$ and the Q was decreased, and $\tau$$_{f}$ was minimized at 0.2 wt.% CuO, and 0.2 wt.% B $i_2$ $O_3$. For this composition, dielectric properties were $\varepsilon$$_{r}$ of 81, Q. $f_{0}$ of 4400 GHz, and $\tau$$_{f}$ of 5 ppm/$^{\circ}C$ at sintering temperature of 100$0^{\circ}C$. the relationship between the microstructure and properties of ceramics was studied by X-ray diffraction(XRD), scanning electron microscopy(SEM).copy(SEM).oscopy(SEM).copy(SEM).EM).

• Annual Conference on Human and Language Technology
• /
• 2017.10a
• /
• pp.147-150
• /
• 2017

Deep learning의 개발에 따라 개체명 인식에도 neural network가 적용된 연구가 활발히 일어나고 있다. 영어권 개체명 인식에서는 F1 score 90%을 웃도는 성능을 내는 연구들이 나오고 있다. 하지만 한국어는 영어와 언어적 특질이 많이 달라 이를 그대로 적용시키는 데는 어려움이 있어 영어권 개체명 인식기에 비해 비교적 낮은 성능을 보인다. 본 논문에서는 "하다" 접사의 동사형이 보존된 워드 임베딩을 사용하고 한국어 개체명의 특징을 담은 one-hot 벡터를 추가하여 한국어의 특질에 보다 적합한 데이터를 deep learning 기술에 적용하였다.

• 한국어정보학회:학술대회논문집
• /
• 2017.10a
• /
• pp.147-150
• /
• 2017

Deep learning의 개발에 따라 개체명 인식에도 neural network가 적용된 연구가 활발히 일어나고 있다. 영어권 개체명 인식에서는 F1 score 90%을 웃도는 성능을 내는 연구들이 나오고 있다. 하지만 한국어는 영어와 언어적 특질이 많이 달라 이를 그대로 적용시키는 데는 어려움이 있어 영어권 개체명 인식기에 비해 비교적 낮은 성능을 보인다. 본 논문에서는 "하다" 접사의 동사형이 보존된 워드 임베딩을 사용하고 한국어 개체명의 특징을 담은 one-hot 벡터를 추가하여 한국어의 특질에 보다 적합한 데이터를 deep learning 기술에 적용하였다.