• 대한음성학회지:말소리
• /
• 제67호
• /
• pp.195-216
• /
• 2008

This paper addresses the problem of single-channel speech separation to extract the speech signal uttered by the speaker of interest from a mixture of speech signals. We propose to apply time-frequency smoothing to the existing statistical single-channel speech separation algorithms: The soft mask and the minimum-mean-square-error (MMSE) algorithms. In the proposed method, we use the two smoothing later. One is the uniform mask filter whose filter length is uniform at the time-Sequency domain, and the other is the met-scale filter whose filter length is met-scaled at the time domain. In our speech separation experiments, the uniform mask filter improves speaker-to-interference ratio (SIR) by 2.1dB and 1dB for the soft mask algorithm and the MMSE algorithm, respectively, whereas the mel-scale filter achieves 1.1dB and 0.8dB for the same algorithms.

• 한국환경과학회지
• /
• 제4권5호
• /
• pp.481-488
• /
• 1995

This study was carried out to evaluate the variations of headloss rate and of specific deposit to depths with effective size of media and configuration of filter layer during algae blooming period. 0.51mm size media was disqualified because most of headloss occurred rapidly below 5cm from surface layer however 0.91mm size media acted deep filtration more than 20cm from top, as result 0.91mm sixte media filter had 2~3 times longer filtration time than 0.51mm sixte media filter, but 0.91mm size media have break-through potentiality. multi-layer filter with 1.02mm anthracite and 0.51mm sand had large deposit volume in upper layer that could longer filtration time, moreover smaller media in lower layer that could protect break-through.

• 한국환경과학회지
• /
• 제4권5호
• /
• pp.95-95
• /
• 1995

This study was carried out to evaluate the variations of headloss rate and of specific deposit to depths with effective size of media and configuration of filter layer during algae blooming period. 0.51mm size media was disqualified because most of headloss occurred rapidly below 5cm from surface layer however 0.91mm size media acted deep filtration more than 20cm from top, as result 0.91mm sixte media filter had 2∼3 times longer filtration time than 0.51mm sixte media filter, but 0.91mm size media have break-through potentiality. multi-layer filter with 1.02mm anthracite and 0.51mm sand had large deposit volume in upper layer that could longer filtration time, moreover smaller media in lower layer that could protect break-through.

• 한국음향학회지
• /
• 제3권2호
• /
• pp.13-22
• /
• 1984

A surface acoustic wave signal processing device using the silicon surface state is presented and shown capable of storing a reference signal and later correlating another signal with the stored reference. The device memory consists of the storage of the spatial 2k pattern of an acoustic wave as stored charges in the surface state of silicon surface. Results of experiments are presented which characterize the operation of device. Simpliied models for charging process and nonlinear acoustoelectric interactions based on consideration of single surface state at the surface of silicon The validity of simplified model has been qualitatibely confirmed with experimental results and the application of this device to aprogrammable matched filter of communication is considered.

• 한국정보통신학회논문지
• /
• 제11권12호
• /
• pp.2374-2381
• /
• 2007

표면 탄성파 필터 적절한 임피던스 정합 조건도 얻고자 단상 단방향성 및 양방향성 변환기형 필터를 제작하였다. 그리고 경사진 빗살무의 변환기 형태인 표면 탄성파 필터 제작하기 위하여 Langasite 기판위에 빗살무의 전극들을 형성시켜 모의실험을 수행하였으며, 전극재료로는 Al-Cu를 사용하였다. Langasite 기판위에 형성시킨 단상 단방향성 필터의 전극 수는 50쌍, 두께는 $5000{\AA}$으로 하였으며, 반사기 폭은 $3.6{\mu}m$이고 전극 폭과 전극과 반사기 사이의 간격은 각각 $1.2{\mu}m$로 하였다. 양방향성 필터에서는 전극 폭과 간격은 각각 $1.8{\mu}m$로 하였다. 제작한 필터의 임피던스 정합 후 주파수 특성에서 중심 주파수는 190MHz, 대역폭은 5.3MHz정도로 측정되었으며, 리플 특성은 0.3dB 이하이고, -20dB 정도의 삽입 손실이 측정되었다.

• Journal of Electrical Engineering and Technology
• /
• 제11권6호
• /
• pp.1714-1728
• /
• 2016

This paper presents improved harmonics extraction based on Adaptive Linear Neuron (ADALINE) algorithm for single phase photovoltaic (PV) shunt active power filter (SAPF). The proposed algorithm, named later as Improved ADALINE, contributes to better performance by removing cosine factor and sum of element that are considered as unnecessary features inside the existing algorithm, known as Modified Widrow-Hoff (W-H) ADALINE. A new updating technique, named as Fundamental Active Current, is introduced to replace the role of the weight factor inside the previous updating technique. For evaluation and comparison purposes, both proposed and existing algorithms have been developed. The PV SAPF with both algorithms was simulated in MATLAB-Simulink respectively, with and without operation or connection of PV. For hardware implementation, laboratory prototype has been developed and the proposed algorithm was programmed in TMS320F28335 DSP board. Steady state operation and three critical dynamic operations, which involve change of nonlinear loads, off-on operation between PV and SAPF, and change of irradiances, were carried out for performance evaluation. From the results and analysis, the Improved ADALINE algorithm shows the best performances with low total harmonic distortion, fast response time and high source power reduction. It performs well in both steady state and dynamic operations as compared to the Modified W-H ADALINE algorithm.

• 상하수도학회지
• /
• 제18권5호
• /
• pp.666-672
• /
• 2004

Granular media filtration is used almost universally as the last particle removal process in conventional water treatment plants. Therefore, superb particle removal efficiency is needed during this process to ensure a high quality of drinking water. However, every particle can not be removed during granular media filtration. Besides the pattern of particle attachment is different depending on physicochemical aspects of particles and suspension. Filtration experiments were performed in a laboratory-scale filter using spherical glass beads with a diameter of 0.55 mm as collectors. A single type of particle suspension (Min-U-Sil 5) and pH control was used to destabilize particles. The operating conditions were similar to those of standard media filtration practice: a filtration velocity of 5 m/h. More favorable particles, i.e., particles with smaller surface charge, were well attached to the collectors especially during the early stage of filtration. This selective attachment of the lower charged particles caused the zeta potential distribution (ZPD) of the effluent to move to a more negative range. On the other hand, the ZPD of the effluent did not keep moving from less negative to more negative during the later stages of filtration, and this result was thought to be caused by two reasons: ripening effects and the detachment of flocs.

• 천문학회보
• /
• 제35권1호
• /
• pp.32.2-32.2
• /
• 2010

The Korean Solar Radio Burst Locator (KSRBL) is a single dish radio spectrograph, which is designed to record the spectra of microwave (0.5 - 18 GHz) bursts with 1 MHz spectral resolution and 1 s time cadence, and locate their positions on the solar disk within 2 arcmin. It was installed at KASI in 2009 August, and operational thereafter. The antenna pointing coefficients were initially determined during the installation and refined later using a series of antenna pointing calibrations. The filter to prevent the radio frequency interference around 2 GHz was designed and is to be installed. After the installation, the full frequency coverage will be recovered from the temporarily restricted frequency coverage (5 - 14 GHz). Also an effort to solve a couple of minor problems for the full performance of the system is in progress.

• 방송공학회논문지
• /
• 제17권5호
• /
• pp.742-755
• /
• 2012

본 논문에서는 high efficiency video coding (HEVC) 복호화기의 디블록킹 필터를 병렬화할 때 발생하는 작업량 불균형 문제를 해결하는 병렬화 방법을 제안한다. HEVC의 디블록킹 필터는 인-루프 필터로써 먼저 수직 에지에서 필터링을 수행한 후, 수평 에지에서 필터링을 수행한다. 수직 및 수평 에지에 대해 필터링을 수행하는 경우 주변 에지와 의존성이 없기 때문에 데이터 레벨의 병렬화를 통하여 복호화를 고속화 할 수 있다. 그러나 데이터 레벨 병렬화 방법을 통해 데이터가 균등하게 분할된 경우에도 영역 간의 작업량은 불균등 할 수 있으며, 이는 복호화기의 병렬화 성능을 저하시킨다. 본 논문에서는 coding tree block (CTB)에서 coding unit (CU)의 깊이 정보를 사용하여, 현재 프레임에 대한 디블록킹 필터링 과정의 연산량을 예측하고, 이를 통해 각 코어에 동등한 작업량이 분배되게 함으로써 작업량 불균형 문제를 해결하였다. 실험 결과, 제안하는 작업량 예측 기반의 데이터 레벨 병렬화 방법은 단일 코어를 사용하여 디블록킹 필터를 수행하는 것에 비하여 64.3%의 평균 시간 감소 (average time saving; ATS)를 얻었고, 기존의 균등 분할 데이터 레벨 병렬화 방법보다 평균 6.7%, 최대 13.5% 감소를 얻었다.

• 전자공학회논문지
• /
• 제54권4호
• /
• pp.68-74
• /
• 2017

본 논문은 컬러 필터 어레이 (CFA) 영상에 대한 공동의 컬러보간(Demosaicking)과 임의 배율의 다운샘플링(arbitrary-ratio down sampling)알고리즘을 제시한다. 컬러보간은 단일 센서를 사용하는 다양한 영상 저장 시스템에서 영상 신호 처리 파이프 라인의 필수적인 부분이다. 또한, 스마트폰과 같은 소구경 카메라 시스템에서는 영상 센서에서 획득되는 고해상도 영상이 보다 작은 해상도의 영상으로 스크린에 다운샘플링되어 디스플레이 된다. 기존 방법에서는 이러한 과정들을 거치기 위해 "컬러보간 후 다운샘플링" 모듈의 순서대로 영상을 처리하게 된다. 하지만 이러한 독립적이고 순차적인 방법은 많은 메모리 소모와 계산량을 필요로 하게 되고, 또한 영상 처리 과정에서 아티팩트(artifact)가 발생하여 영상 디테일의 손상을 가져오게 된다. 본 논문에서는 이러한 문제점을 해결하기 위해 컬러보간과 다운샘플링을 동시에 동작하도록 하는 알고리즘을 제안한다. 제안하는 방법에서는 입력되는 컬러 필터 영상으로부터 고주파 및 저주파 성분으로 신호를 분해하는 방법에 기초하여 임의의 배율에서 컬러 필터 영상으로 역 사상(inverse mapping)을 하고 컬러보간 및 다운샘플링을 동시에 수행하게 된다. 제안하는 알고리즘은 종래의 기술보다 더 나은 영상 품질 성능을 나타내면서 동시에 보다 적은 계산량이 필요로 하는 것을 보여준다.