• 한국ITS학회:학술대회논문집
• /
• v.2007 no.10
• /
• pp.289-296
• /
• 2007

• Ceramist
• /
• v.7 no.3
• /
• pp.28-33
• /
• 2004

1987년 미국 버클리 대학의 연구진은 반도체 미세공정 기술을 이용해 머릿카락 굵기의 초소형 모터를 발표하였다(Fig. 1).1) 이는 MEMS(Micro Electro Mechanical System, 혹은 MST , Micro System Technology)라는 새로운 학문 분야의 실질적 효시로, 명실공히 마이크론 단위의 기계-전자 시스템의 구현이라는 새로운 장을 열게되었다. (중략)

• Proceedings of the Materials Research Society of Korea Conference
• /
• 2003.11a
• /
• pp.69-69
• /
• 2003

최근 trench capacitor, isolation trench, micro-electromechanical system(MEMS), micro-opto-electromechanical system(MOEMS)등의 다양한 기술에 적용될 고종횡비(HAR) 실리콘 식각기술연구가 진행되어 지고 있다. 이는 기존의 습식식각시 발생하는 결정방향에 따른 식각률의 차이에 관한 문제와 standard reactive ion etching(RIE) 에서의 낮은 종횡비와 식각률에 기인한 문제점들을 개선하기 위해 고밀도 플라즈마를 이용한 건식식각 장비를 사용하여 고종횡비(depth/width), 높은 식각률을 가지는 이방성 트랜치 구조를 얻는 것이다. 초기에는 주로 HBr chemistry를 이용한 연구가 진행되었는데 이는 식각률이 낮고 많은양의 식각부산물이 챔버와 시편에 재증착되는 문제가 발생하였다. 또한 SF6 chemistry의 사용을 통해 식각률의 향상은 가져왔지만 화학적 식각에 기인한 local bowing과 같은 이방성 식각의 문제점들로 인해 최근까지 CHF3, C2F6, C4F8, CF4등의 첨가가스를 이용하여 측벽에 Polymer layer의 식각보호막을 형성시켜 이방성 구조를 얻는 multi_step 공정이 일반화 되었다. 이에 본 연구에서는 SF6 chemistry와 소량의 02/HBr의 첨가가스를 이용한 single_step 공정을 통해 공정의 간소화 및 식각 프로파일을 개선하여 최적의 HAR 실리콘 식각공정 조건을 확보하고자 하였다.

• Proceedings of the Korean Institute of Resources Recycling Conference
• /
• 2001.05b
• /
• pp.129-132
• /
• 2001

The purpose of this research was to understand possibility of foodwaste treatment by hybrid anaerobic reactor(HAR). The Possibility of methane utility and applicability of hybrid reactor system using foodwaste as substrate was investigated. The maximum loading rate and optimized operational conditions were determined. Hybrid anaerobic reactor was filled with packing material 50% of its total volume between the tube and the outer surface. The packing material used was randomly packed open-pore synthesis activated ceramic(SAC) media as support media for microbial attachment, growth, and chemical stability protected bacteria from effect of organic acid accumulation. In this research, although foodwaste has high concentrations C $l^{[-10]}$ and S $O_{4}$$^{2-}$ concentration the possibility of foodwaste anaerobic treatment was when foodwaste is treated by anaerobic digestion, this study focused on the possibility using C $H_4$ gas made under the anaerobic treatment as an alternative energy source. Other objective of this research is to study struvite formation and crystal forms in anaerobic digester. HAR is used to investigate phosphate crystallization without the addition of chemicals.

• Tribology and Lubricants
• /
• v.33 no.6
• /
• pp.282-287
• /
• 2017

SU-8 is an epoxy-type photoresist widely used for the fabrication of high-aspect-ratio (HAR) micro-structures in micro-electro-mechanical systems (MEMS). To fabricate highly integrated structures, chemical mechanical polishing (CMP) has emerged as the preferred manufacturing process for planarizing the MEMS structure. In SU-8 CMP, an oxidizer decomposes organic impurities and particles in the CMP slurry remove the chemically reacted surface of SU-8. To fabricate HAR microstructures using the CMP process, the adhesion between SU-8 and substrate material is important to avoid the delamination of the SU-8 film caused by the mechanical-dominant material removal characteristic. In this study, the friction force during the CMP process is measured with a CMP monitoring system to detect the delamination phenomenon and investigate the delamination of the SU-8 film from the silicon substrate under various pressure conditions. The increase in applied pressure causes an increase in the frictional force and wafer-edge stress concentration. The frictional force measurement shows that the friction force changes according to the delamination phenomenon of the SU-8 film, and that it is possible to monitor the delamination phenomenon during the SU-8 CMP process. The delamination at a high applied pressure is explained by the effect of stress distribution and pad deformation. Consequently, it is necessary to control the pressure of polishing, which can avoid the delamination in SU-8 CMP.

• Proceedings of the Korean Vacuum Society Conference
• /
• 2011.02a
• /
• pp.284-284
• /
• 2011

Recently, the increasing degree of device integration in the fabrication of Si semiconductor devices, etching processes of nano-scale materials and high aspect-ratio (HAR) structures become more important. Due to this reason, etch selectivity control during etching of HAR contact holes and trenches is very important. In this study, The etch selectivity and etch rate of TEOS oxide layer using ACL (amorphous carbon layer) mask are investigated various process parameters in CH2F2/C4F8/O2/Ar plasma during etching TEOS oxide layer using ArF/BARC/SiOx/ACL multilevel resist (MLR) structures. The deformation and etch characteristics of TEOS oxide layer using ACL hard mask was investigated in a dual-frequency superimposed capacitively coupled plasma (DFS-CCP) etcher by different fHF/ fLF combinations by varying the CH2F2/ C4F8 gas flow ratio plasmas. The etch characteristics were measured by on scanning electron microscopy (SEM) And X-ray photoelectron spectroscopy (XPS) analyses and Fourier transform infrared spectroscopy (FT-IR). A process window for very high selective etching of TEOS oxide using ACL mask could be determined by controlling the process parameters and in turn degree of polymerization. Mechanisms for high etch selectivity will discussed in detail.

• Journal of the Korean Society of Industry Convergence
• /
• v.24 no.6_1
• /
• pp.693-698
• /
• 2021

As technology advances, there is increasing need for research in different fields where this technology is applied. On of the most researched topic in computer vision is Human activity recognition (HAR), which has widely been implemented in various fields which include healthcare, video surveillance and education. We therefore present in this paper a human activity recognition system based on scale and rotation while employing the Kinect depth sensors to obtain the human skeleton joints. In contrast to previous approaches that use joint angles, in this paper we propose that each limb has an angle with the X, Y, Z axes which we employ as feature vectors. The use of the joint angles makes our system scale invariant. We further calculate the body relative direction in the egocentric coordinates in order to provide the rotation invariance. For the system parameters, we employ 8 limbs with their corresponding angles each having the X, Y, Z axes from the coordinate system as feature vectors. The extracted features are finally trained and tested with the Long short term memory (LSTM) Network which gives us an average accuracy of 98.3%.

• International Journal of Computer Science & Network Security
• /
• v.21 no.7
• /
• pp.359-364
• /
• 2021

In recent years Hadoop usage has been increasing day by day. The need of development of the technology and its specified outcomes are eagerly waiting across globe to adopt speedy access of data. Need of computers and its dependency is increasing day by day. Big data is exponentially growing as the entire world is working in online mode. Large amount of data has been produced which is very difficult to handle and process within a short time. In present situation industries are widely using the Hadoop framework to store, process and produce at the specified time with huge amount of data that has been put on the server. Processing of this huge amount of data having small files & its storage optimization is a big problem. HDFS, Sequence files, HAR, NHAR various techniques have been already proposed. In this paper we have discussed about various existing techniques which are developed for accessing and storing small files efficiently. Out of the various techniques we have specifically tried to implement the HDFS- HAR, NHAR techniques.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.39 no.2
• /
• pp.30-38
• /
• 2002

In this paper, we propose a new method for image coding which efficiently use the relationship between the properties of spatial image and its wavelet transform. Firstly, an original image is decomposed into several layers by the wavelet transform, and simultaneously decomposed into 2$^n$$\times$2$^n$blocks. Each block is classified into two regions according to their standard deviation, i.e., low activity region(LAR) and high activity region(HAR). The region with low frequency in spatial domain does not only appears as zero regions in wavelet frequency domain like HL, LH, and HH but also gives little influence to the quality of reconstructed image. The other side, the high frequency regions are related to significant coefficients which gives much influence to image reconstruction. In this paper, we propose a image coding method to obtain high compression rate at low bit rate by these properties. The LAR region is encoded by LAR coding method which is proposed in this paper, the HAR by a technique similar to bitplane coding in hierarchical tree. Simulation results show that th,$\boxUl$ proposed coding method has better performance than EZW and SPIHT schemes in terms of image quality and transmitted bit rates, can be successfully applied to the application areas that require of progressive transmission.

• Database Research
• /
• v.34 no.3
• /
• pp.137-147
• /
• 2018

A key to a well-performing human activity recognition (HAR) system through machine learning technique is the availability of a substantial amount of labeled data. Collecting sufficient labeled data is an expensive and time-consuming task. To build a HAR system in a new environment (i.e., the target domain) with very limited labeled data, it is unfavorable to naively exploit the data or trained classifier model from the existing environment (i.e., the source domain) as it is due to the domain difference. While traditional machine learning approaches are unable to address such distribution mismatch, transfer learning approach leverages the utilization of knowledge from existing well-established source domains that help to build an accurate classifier in the target domain. In this work, we propose a transfer learning approach to create an accurate HAR classifier with very limited data through the multitask neural network. The classifier loss function minimization for source and target domain are treated as two different tasks. The knowledge transfer is performed by simultaneously minimizing the loss function of both tasks using a single neural network model. Furthermore, we utilize the unlabeled data in an unsupervised manner to help the model training. The experiment result shows that the proposed work consistently outperforms existing approaches.