• 대한음성학회지:말소리
• /
• 제56호
• /
• pp.207-223
• /
• 2005

The goal of language model adaptation is to improve the background language model with a relatively small adaptation corpus. This study presents a language model adaptation technique where additional text data for the adaptation do not exist. We propose the information retrieval (IR) technique with N-gram language modeling to collect the adaptation corpus from baseline text data. We also propose to use a dynamic language model interpolation coefficient to combine the background language model and the adapted language model. The interpolation coefficient is estimated from the word hypotheses obtained by segmenting the input speech data reserved for held-out validation data. This allows the final adapted model to improve the performance of the background model consistently The proposed approach reduces the word error rate by $13.6\%$ relative to baseline 4-gram for two-hour broadcast news speech recognition.

• 대한음성학회지:말소리
• /
• 제51호
• /
• pp.99-115
• /
• 2004

In this parer, we propose LM adaptation for broadcast news recognition. We collect information of recent articles from the internet on real time, make a recent small size LM, and then interpolate recent LM with a existing LM composed of existing large broadcast news corpus. We performed interpolation experiments to get the best type of articles from recent corpus because collected recent corpus is composed of articles which are related with test set, and which are unrelated. When we made an adapted LM using recent LM with similar articles to test set through Tf-Idf method and existing LM, we got the best result that ERR of pseudo-morpheme based recognition performance has 17.2 % improvement and the number of OOV has reduction from 70 to 27.

• 정보과학회 컴퓨팅의 실제 논문지
• /
• 제23권7호
• /
• pp.446-451
• /
• 2017

본 논문에서의 문맥의존 철자오류(Context-Sensitive Spelling Error) 교정 기법은 샤논(Shannon)의 노이지 채널 모형(noisy channel model)을 기반으로 한다. 논문에서 제안하는 교정 기법의 향상에는 보간(interpolation)을 사용하며, 일반적인 보간 방법은 확률의 중간 값을 채우는 방식으로 N-gram에 존재하지 않는 빈도를 (N-1)-gram과 (N-2)-gram 등에서 얻는다. 이와 같은 방식은 동일 통계 말뭉치를 기반으로 계산하는데 제안하는 방식에서는 통계 말뭉치와 교정 문서간의 빈도 정보를 이용하여 보간 한다. 교정 문서의 빈도를 이용하였을 때 이점은 다음과 같다. 첫째 통계 말뭉치에 존재하지 않고 교정 문서에서만 나타나는 신조어의 확률을 얻을 수 있다. 둘째 확률 값이 모호한 두 교정 후보가 있더라도 교정 문서를 참고로 교정하게 되어 모호성을 해소한다. 제안한 방법은 기존 교정 모형보다 정밀도와 재현율의 성능향상을 보였다.

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2022년도 학술발표회
• /
• pp.368-368
• /
• 2022

The Soil and Water Assessment Tool (SWAT) has been widely used to simulate the long-term hydrological conditions of a catchment. Two output variables, outflow and sediment yield have been widely investigated in the field of water resources management, especially in determining the conditions of ungauged subbasins. The presence of missing data in weather input data can cause poor representation of the climate conditions in a catchment especially for large or mountainous catchments. Therefore, in this study, a custom module was developed and evaluated to determine the efficiency of utilizing basic spatial interpolation methods in the estimation of weather input data. The module has been written in Python language and can be considered as a pre-processing module prior to using the SWAT model. The results of this study suggests that the utilization of the proposed pre-processing module can improve the simulation results for both outflow and sediment yield in a catchment, even in the presence of missing data.

• 전자공학회논문지S
• /
• 제36S권7호
• /
• pp.71-80
• /
• 1999

수화 통신은 이종 언어간의 통신 수단으로 사용될 수 있다. 이 논문에서는 3차원 모델을 이용하여 한-일간 수화 통신 시스템을 구현하여 그 가능성을 실험하였다. 실시간 통신을 위하여 통신 시스템을 클라이언트/서버 구조로 하였으며, 지적 통신방식을 도입하였다. 각 클라이언트에 3차원 모델을 준비하여 놓고, 실제의 수화영상 대신에 애니메이션 생성을 위한 파라미터 만을 전송하였다. 클라이언트에서 입력된 문장은 서버로 전송되어 한국 또는 일본 수화 파라미터로 변환한 다음 다시 클라이언트로 전송되어 수화 애니메이션으로 재생된다. 또한 자연스러운 수화 애니메이션을 위하여 감정 표현과 가변 프레임 방식 및 3차 스플라인 보간식을 이용하였다. 실험을 위한 통신 시스템은 윈도우 플랫폼에서 Visual $C^{++}$ 와 Open Inventor 라이브러리를 이용하여 구현하였다. 실험 결과 제안 시스템이 언어의 장벽을 넘을 수 있는 비언어 통신수단으로 이용될 수 있는 가능성을 보였다.

• 말소리와 음성과학
• /
• 제2권1호
• /
• pp.87-95
• /
• 2010

In this paper, we address issues associated with mixed-lingual text-to-speech based on context-dependent HMMs, where there are multiple sets of HMMs corresponding to each individual language. In particular, we propose smoothing techniques of synthesis parameters at the boundaries between different languages to obtain more natural quality of speech. In other words, mel-frequency cepstral coefficients (MFCCs) at the language boundaries are smoothed by applying several linear and nonlinear approximation techniques. It is shown from an informal listening test that synthesized speech smoothed by a modified version of linear least square approximation (MLLSA) and a quadratic interpolation (QI) method is preferred than that without using any smoothing technique.

• 한국음향학회지
• /
• 제13권1호
• /
• pp.14-23
• /
• 1994

본 논문에서는 한국어 음성인식에 적합한 음성 인식 단위에 대해서 연구하였다. 좋은 음성 인식 시스템을 구현하기 위해서는 발음된 음성내의 조음화현상을 처리할 수 있는 인식단위를 선택해야만 한다. 따라서 음소보다 개념적으로 확대된 인식단위가 필요하게 되는데, diphone은 음소간의 전이영역을 modeling하기때문에 좋은 인식 단위가 될 수 있다. Diphone을 인식 단위로 할 경우에 안정적인 음소영역을 diphone사이에 삽입할 수도 있다. 7명의 남성화자가 발음한 74단어로 구성된 고립단어 인식 실험결과 diphone을 2-state HMM으로, 터짐소리 `ㅂ',`ㄷ','ㄱ'와 묵음을 제외한 음소에 대해서 1-state HMM으로 나타냈을 때 가장 높은 인식률을 보였다. 이때 드물게 발생하는 diphone들을 하나의 단위로 merging했을 때 인식률이 $93.98\%$에서 $96.29\%$로 향상되었다. 또한 merging된 diphone과 제안한 국소보간법 (local interpolation technique)을 사용함으로써 $97.22\%$까지 인식률이 향상되었다.

• 한국농업기계학회:학술대회논문집
• /
• 한국농업기계학회 2000년도 THE THIRD INTERNATIONAL CONFERENCE ON AGRICULTURAL MACHINERY ENGINEERING. V.II
• /
• pp.124-131
• /
• 2000

In this study, the NEURAL NETWORK (hereinafter referred to as NN) was applied to control of the nonlinear factors for turning movement of the crawler vehicle and experiment was carried out using a small model of crawler vehicle in order to inspect an application of NN. Furthermore, CHAOS NEURAL NETWORK (hereinafter referred to as CNN) was also applied to this control so as to compare with conventional NN. CNN is especially effective for plane in many variables with local minimum which conventional NN is apt to fall into, and it is relatively useful to nonlinear factors. Experiment of turning on the slope of crawler vehicle was performed in order to estimate an adaptability of nonlinear problems by NN and CNN. The inclination angles of the road surface which the vehicles travel on, were respectively 4deg, 8deg, 12deg. These field conditions were selected by the object for changing nonlinear magnitude in turning phenomenon of vehicle. Learning of NN and CNN was carried out by referring to positioning data obtained from measurement at every 15deg in turning. After learning, the sampling data at every 15deg were interpolated based on the constructed learning system of NN and CNN. Learning and simulation programs of NN and CNN were made by C language ("Association of research for algorithm of calculating machine (1992)"). As a result, conventional NN and CNN were available for interpolation of sampling data. Moreover, when nonlinear intensity is not so large under the field condition of small slope, interpolation performance of CNN was a little not so better than NN. However, when nonlinear intensity is large under the field condition of large slope, interpolation performance of CNN was relatively better than NN.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1988년도 전기.전자공학 학술대회 논문집
• /
• pp.971-974
• /
• 1988

In this study, we designed a model of NC controller with IBM-PC as a host CPU and intelligent servo controller for 2 axes that can be expaned to 4 axes. OS software was developed with C language at the base of mode selection technic for 9 NC operating modes. Servo controller design was based on the application of interpolator IC(3701) and position controller IC(3702) that permits low cost and high performance. For connection of two systems, parallel I/O communication was implemented. Finally, auto interpolation program test was executed for linear and circular paths resulting 1 LSB accuracy.

• Advances in aircraft and spacecraft science
• /
• 제6권2호
• /
• pp.169-187
• /
• 2019

This work presents a practical detailed finite element (FE) approach for the three-dimensional (3D) free-vibration analysis of actual aircraft and spacecraft-type lightweight and thin honeycomb sandwich panels. It consists of calling successively in $MATLAB^{(R)}$, via a developed user-friendly GUI, a detailed 3D meshing tool, a macrocommands language translator and a commercial FE solver($ABAQUS^{(R)}$ or $ANSYS^{(R)}$). In contrary to the common practice of meshing finely the faces and core cells, the proposed meshing tool represents each wall of the actual hexagonal core cells as a single two-dimensional (2D) 4 nodes quadrangularshell element or two 3 nodes triangular ones, while the faces meshes are obtained simply using the nodes at the core-faces interfaces. Moreover, as the same 2D FE interpolation type is used for meshing the core and faces, this leads to an automatic handling of their required FE compatibility relations. This proposed approach is applied to a sample made of very thin glass fiber reinforced polymer woven composite faces and a thin aluminum alloy hexagonal honeycomb core. The unknown or incomplete geometric and materials properties are first collected through direct measurements, reverse engineering techniques and experimental-FE modal analysis-based inverse identification. Then, the free-vibrations of the actual honeycomb sandwich panel are analyzed experimentally under different boundary conditions and numerically using different mesh basic cell shapes. It is found that this approach is accurate for the first few modes used for pre-design purpose.