• 한국컴퓨터정보학회논문지
• /
• 제27권2호
• /
• pp.15-23
• /
• 2022

특허 분야에서 자연어처리(Natural Language Processing) 태스크는 특허문헌의 언어적 특이성으로 문제 해결의 난이도가 높은 과제임에 따라 한국 특허문헌에 최적화된 언어모델의 연구가 시급한 실정이다. 최근 자연어처리 분야에서는 특정 도메인에 특화되게 사전 학습(Pre-trained)한 언어모델을 구축하여 관련 분야의 다양한 태스크에서 성능을 향상시키려는 시도가 지속적으로 이루어지고 있다. 그 중, ELECTRA는 Google이 BERT 이후에 RTD(Replaced Token Detection)라는 새로운 방식을 제안하며 학습 효율성을 높인 사전학습 언어모델이다. 본 연구에서는 대량의 한국 특허문헌 데이터를 사전 학습한 KorPatELECTRA를 제안한다. 또한, 특허 문헌의 특성에 맞게 학습 코퍼스를 정제하고 특허 사용자 사전 및 전용 토크나이저를 적용하여 최적화된 사전 학습을 진행하였다. KorPatELECTRA의 성능 확인을 위해 실제 특허데이터를 활용한 NER(Named Entity Recognition), MRC(Machine Reading Comprehension), 특허문서 분류 태스크를 실험하였고 비교 대상인 범용 모델에 비해 3가지 태스크 모두에서 가장 우수한 성능을 확인하였다.

• 대한전기학회논문지:시스템및제어부문D
• /
• 제50권12호
• /
• pp.590-596
• /
• 2001

The pre-emphasis is an essential process for speech signal processing. Widely used two methods are the typical method using a fixed value near unity and te optimal method using the autocorrelation ratio of the signal. This study proposes a new pre-emphasis method using the short-term energy difference of speech signal, which can effectively compensate the glottal source characteristics and lip radiation characteristics. Using the proposed pre-emphasis, speech analysis, such as spectrum estimation, formant detection, is performed and the results are compared with those of the conventional two pre-emphasis methods. The speech analysis with 5 single vowels showed that the proposed method enhanced the spectral shapes and gave nearly constant formant frequencies and could escape the overlapping of adjacent two formants. comparison with FFT spectra had verified the above results and showed the accuracy of the proposed method. The computational complexity of the proposed method reduced to about 50% of the optimal method.

• 반도체디스플레이기술학회지
• /
• 제9권1호
• /
• pp.11-16
• /
• 2010

Wafer Pre-Alignment is to find the center and the orientation of a wafer and to move the wafer to the desired position and orientation. In this paper, an area camera based pre-aligning method is presented that captures 8 wafer images regularly during 360 degrees rotation. From the images, wafer edge positions are extracted and used to estimate the wafer's center and orientation using least squares circle fitting. These data are utilized for the proper alignment of the wafer. For accurate alignments, camera calibration methods using high order polynomials are used for converting pixel coordinates into real-world coordinates. A complete pre-alignment system was constructed using mechanical and optical components and tested. Experimental results show that alignment of wafer center and orientation can be done with the standard deviation of 0.002 mm and 0.028 degree, respectively.

• 전자공학회논문지B
• /
• 제30B권12호
• /
• pp.12-20
• /
• 1993

In a conventional adaptive echo canceller, an ADF(Adaptive Digital Filter) with TDL(Tapped-Delay Line) structure modelling the echo path uses the LMS(Least Mean Square) algorithm to compute the coefficients, and NET detector using energy comparison method prevents the ADF to update the coefficients during the periods of the NET signal presence. The convergence speed of the LMS algorithm depends on the eigenvalue spread ratio of the reference signal and NET detector using the energy comparison method yields poor detection performance if the magnitude of the NET signal is small. This paper presents a new adaptive echo canceller which uses the pre-whitening filter to improve the convergence speed of the LMS algorithm. The pre-whitening filter is realized by using a low-order lattice predictor. Also, a new NET signal detection algorithm is presented, where the start point of the NET signal is detected by computing the cross-correlation coefficient between the primary input and the ADF output while the end point is detected by using the energy comparison method. The simulation results show that the convergence speed of the proposed adaptive echo canceller is faster than that of the conventional echo canceller and the cross-correlation coefficient yields more accurate detection of the start point of the NET signal.

• Mass Spectrometry Letters
• /
• 제8권4호
• /
• pp.90-97
• /
• 2017

Androgenic anabolic steroids (AASs) are synthetic derivatives of testosterone with a common structure containing cyclopentanoperhydrophenanthrene nucleus. Their use enhances the muscle building capacity and is beneficial during performance. The AASs are one of the most abused group of substances in horse doping. Liquid chromatography tandem mass spectrometry ($LC/MS^n$) has been successfully applied to the detection of anabolic steroids in biological samples. However, the saturated hydroxysteroids viz: nandrolone, $5{\alpha}-estrane-3{\beta}$, $17{\alpha}-diol$ exhibit lower detection responses in electrospray ionisation (ESI) because of their poor ionisation efficiency. To overcome this limitation pre-column chemical derivatization has been introduced to enhance their detection responses in $LC-ESI-MS^n$ analysis. The aim of present study was to develop a sensitive method for identification and confirmation of nandrolone and its metabolite in horse urine incorporating pre-column derivatization using picolinic acid. The method consists of extraction of targeted steroid conjugates by solid phase extraction (SPE). The eluted steroid conjugates were hydrolysed by methanolysis and free steroids were recovered with liquid-liquid extraction. The resulting steroids were derivatized to form picolinoyl esters and identification was done using LC-ESI-MS/MS in positive ionization mode. The picolinated steroid adduct enhanced the detection levels in comparison to underivatized steroids.

• Journal of Information Processing Systems
• /
• 제12권3호
• /
• pp.502-510
• /
• 2016

Considering video copy transform diversity, a multi-feature video copy detection algorithm based on a Speeded-Up Robust Features (SURF) local descriptor is proposed in this paper. Video copy coarse detection is done by an ordinal measure (OM) algorithm after the video is preprocessed. If the matching result is greater than the specified threshold, the video copy fine detection is done based on a SURF descriptor and a box filter is used to extract integral video. In order to improve video copy detection speed, the Hessian matrix trace of the SURF descriptor is used to pre-match, and dimension reduction is done to the traditional SURF feature vector for video matching. Our experimental results indicate that video copy detection precision and recall are greatly improved compared with traditional algorithms, and that our proposed multiple features algorithm has good robustness and discrimination accuracy, as it demonstrated that video detection speed was also improved.

• Journal of Electrochemical Science and Technology
• /
• 제10권2호
• /
• pp.206-213
• /
• 2019

The development of simple methods for As detection has received great attention because As is a toxic chemical element causing environmental and health-related issues. In this work, the effect of nanostructures of Au electrodes on their electroanalytical performance during As detection was investigated. Different Au nanostructures with various surface morphologies such as nanoplate Au, nanospike Au, and dendritic Au structures were prepared, and their electrochemical behaviors toward square-wave anodic stripping voltammetric As detection were examined. The difference in intrinsic efficiency for As detection between nanostructured and flat Au electrodes was explained based on the crystallographic orientations of Au surfaces, as examined by the underpotential deposition of Pb. The most efficient As detection performance was obtained with nanoplate Au electrodes, and the effects of the pre-deposition time and interference on As detection of the nanoplate Au electrodes were also investigated.

• 대한전자공학회논문지TC
• /
• 제45권12호
• /
• pp.152-158
• /
• 2008

천해 환경에서 정합필터와 LFM(linear frequency modulation) 펄스를 사용하는 능동소나의 탐지성능은 비백색잡음인 잔향에 의해 크게 저하될 수 있다. 이 경우 잔향의 영향을 줄이기 위해 일반적으로 정합필터에 선행하여 백색화 필터를 사용한다. 기존에는 탐지 블록과 선행 블록의 잔향이 통계적으로 정상성을 유지한다고 가정하고 선행 블록의 잔향 특성을 이용하여 탐지 블록의 잔향을 추정하고 백색화 하였다. 잔향의 정상성은 인접 블록 사이뿐만 아니라 좀더 넓은 범위에서 유지될 수 있다. 이 경우 더 많은 인접 블록을 이용하면 탐지 블록의 잔향을 더 정밀하게 추정할 수 있을 것이다. 본 논문에서는 실제 해안에서 획득한 잔향 신호를 분석하여 적절한 추정 블록의 범위를 구한다. 그리고 추정 블록의 범위에 따른 선-백색화 정합필터의 성능을 비교, 분석한다.

• 대한산업공학회지
• /
• 제38권4호
• /
• pp.258-261
• /
• 2012

In this paper, I propose a new asset allocation framework to cope with the dynamic nature of the financial market. The investment performance can be much improved by protecting the capital from the market crashes, and such crashes can be pre-identified with high probabilities by regime detection analysis via a specialized unsupervised machine learning technique.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2022년도 추계학술대회
• /
• pp.126-128
• /
• 2022

경량화 기법 중 하나인 Knowledge distillation 은 최근 object detection task 에 적용되고 있다. Knowledge distillation 은 3 가지 범주로 나뉘는데 그들 중에서 Self-Knowledge distillation 은 기존의 Knowledge distillation 에서의 pre-trained teacher 에 대한 의존성 문제를 완화시켜준다. Self-Knowledge distillation 또한 object detection task 에 적용되어 training cost 를 줄이고 고전적인 teacher-based methods 보다 좋은 성능을 성취했다.