• The Journal of the Acoustical Society of Korea
• /
• v.41 no.2
• /
• pp.122-129
• /
• 2022

In this paper, we propose a method to improve the performance of Transformer-based speech recognizers using data augmentation that locally adjusts the frame rate. First, the start time and length of the part to be augmented in the original voice data are randomly selected. Then, the frame rate of the selected part is changed to a new frame rate by using linear interpolation. Experimental results using the Wall Street Journal and LibriSpeech speech databases showed that the convergence time took longer than the baseline, but the recognition accuracy was improved in most cases. In order to further improve the performance, various parameters such as the length and the speed of the selected parts were optimized. The proposed method was shown to achieve relative performance improvement of 11.8 % and 14.9 % compared with the baseline in the Wall Street Journal and LibriSpeech speech databases, respectively.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2022.05a
• /
• pp.312-312
• /
• 2022

우리나라는 하천법 및 소하천정비법에 의거하여 하천기본계획 수립 또는 하천의 지정 및 변경고시 시 하천구역을 지정하고 있다. 하천구역은 일반적으로 제방부지 및 제방하심측 토지경계를 기준으로 지정하나, 제방이 존재하지 않는 무제부 구간의 경우 명확한 제방경계가 없는관계로 하천법 제 10조 3항에서 5항까지 별도의 기준을 통해 하천구역 지정을 권장하고 있는 실정이다. 하천구역 설정 시 기준으로 삼는 횡단측점 자료의 경우 그 특성상 하천의 종단방향으로 불연속적인 특징을 갖고 있어 평면상 정확한 경계의 파악이 어려우며, 도로·철도 등 선형시설경계를 하천구역으로 설정 시 편입용지의 보상기준이 모호하여 다량의 민원이 발생하고 있다. 본 연구에서는 하천정비 시 기본계획이 수립되어 각 횡단측점별 결정된 계획홍수위를 기반으로 인접 지형의 홍수위 영향범위를 자동으로 추출하여 하천구역을 정밀하게 결정할 수 있는 방법론을 정립하고자 한다. 첫째로, 하천중심선의 각 측점의 위치정보와 하천의 지형을 위상정보체계로 구성하여 DB를 구축하였다. 둘째로, 측선과 측선사이 절점에 계획홍수위를 선형보간하여 부여하고 이를 지형도의 최단거리에 위치한 지형의 격자표고와 비교해 침수여부를 판단한다. 셋째, 최단거리 지형격자가 침수로 판단될 시 인접한 8개 지형격자의 지형표고와 홍수위를 비교하여 반복적으로 위 과정 수행을 통해 계획홍수위 기반 침수범위를 추출한다. 마지막으로, 이를 수치지형도에 중첩시켜 최종 정밀 하천구역을 결정할 수 있다. 이를 이용하여 산정된 정밀한 하천구역 경계설정을 통해 하천 내사유지 편입을 최소화함과 동시에 명확한 하천구역 구획기준을 정립하여 토지소유주와 담당부처 간 논의 시 기반자료로 활용될 수 있고, 하천구역의 신속하고 정확한 구역설정을 통해 하천인근의 토지이용 고도화 및 효율적인 이용이 가능할 것으로 기대된다.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.45 no.5
• /
• pp.40-45
• /
• 2008

In this paper, a microstrip linear tapered slot antenna is designed. A tapered slot antenna(TSA) has many advantages such as low profile, low weight, easy fabrication, and compatibility with monolithic microwave integrated circuits(MMIC). In addition, it has demonstrated multi octave bandwidth, moderately high gain, and symmetrical E- and H-plane beam patterns. A feed network is implemented with transition between a microstrip and a slot line for the microstrip linear tapered slot antenna. The transition is consist of two sides. One side has a microstrip line, the other side has a slot line. The dimensions of the microstrip and slot line are ${\lambda}_m/4$ and ${\lambda}_s/4$ at the center of the cross section of the microstrip and slot line. In order to get broad bandwidth antenna characteristics, the tapered length is chosen as $4{\lambda}_o$ and termination width is chosen as $1.75{\lambda}_o$. Experimental results show that the microstrip tapered slot antenna has symmetrical E- and H-plane beam patterns with around 5GHz of bandwidth at center frequency of 5.0GHz.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.18 no.7
• /
• pp.723-730
• /
• 2007

This paper shows how the ACLR of power amplifier can be reduced by using Anti-phase IMD linearization technique which generate anti-phase IMD in the driver stage compare to output stage's IMD. And design process proposed. From the experimental result of W-CDMA 4FA input signal, this amplifier has ACLR -55 dBc@5 MHz offset at 30 watts average power. Compare to optimum matching technique to get maximum power gain, this technique has been improved ACLR by 12 dBc. Also this amplifier meets 50 watts average output power amplifier specification in domestic market.

• Proceedings of the KIEE Conference
• /
• 2015.07a
• /
• pp.357-358
• /
• 2015

본 논문에서는 고온 초전도 케이블 고장점 탐지를 위해 시간-주파수 반사파 계측법(Time-Frequency Domain Reflectometry)을 적용하여 EMTP 시뮬레이션을 수행하고, 고온 초전도 케이블에 적합한 가우시안 첩 포락선 선형 기준 신호를 제시하였다. 고온 초전도 케이블을 EMTP로 모델링하여 TFDR 기법을 통해 케이블의 종단점을 추정하고 기준 신호의 전파 속도를 계산하였다. EMTP를 활용하여 국부적 결함이 발생한 고온 초전도 케이블을 모델링하였고, 결함 발생 케이블의 고장점 탐지 시뮬레이션 결과를 확인하였다.

• Journal of Korea Water Resources Association
• /
• v.30 no.1
• /
• pp.35-44
• /
• 1997

Petrov-Galerkin finite element model for analyzing dynamic wave equation is applied to gradually and rapidly varied unsteady flow. The model in verified by applying to hydraulic jump, nonlinear disturbance propagation in frictionless horizontal channel and dam-break analysis. It shows stable and accurate results compared with analytical solutions for various cases. The model in applied to a surge propagation in a frictionless horizontal channel. Three-dimensional water surface profiles show that the computed result converges to the analytical one with sharp discontinuity. The model is also applied to the Taehaw River to analyze unsteady floodwave propagation. The computed results have good agreements with those of DWOPER model in terms of discharge and stage hydrographs.

• Journal of Korean Society for Geospatial Information Science
• /
• v.4 no.2 s.8
• /
• pp.173-180
• /
• 1996

In a grid based data transformation, the different gridding methods provide different interpretations of scattered data because each method calculate grid node values using a different weighted mathematical algorithms. Therefore, it is necessary to review the interpolated characteristics of some gridding methods according to search distance, search area and search options before determing the best method with a data set. For this, in this paper, six different gridding methods with the same search conditions are applied to a scattered data obtained from sterro-plotter. The interpolated characteristics of the scattered geographic data considered through comparison of coincidence between the data point and the grid node being interpolated. And also, shows the real application of gridding methods through calculating volumes and creating cross sections.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.8 no.5
• /
• pp.1076-1081
• /
• 2007

In this paper, we propose a new switch structure for time division duplex(TDD) system. The existing TDD structure utilizes a circulator fur isolation characteristic between ports. However, the circulator produces intermodulation distortion signals which are undesired signal because of its nonlinear properties. The proposed circuit is composed of a modified branch-line hybrid coupler which controls the signal flow while the isolated port is open-/short- terminated. In order to prove the validity of the presented structure, the switch circuit is fabricated and measured at 2.3GHz, the center frequency of Wibro service system.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.12
• /
• pp.2674-2685
• /
• 1997

In this paper, the design and imprlementation of dual-band power amplifier which is used as a critical part for mobile phone to be simultaneously working at a dual band, 800MHz CDAM and PCS frequency band is described. DC operating point of power FET is limited to Class-B to enable long talk time considering that the tyupical power range of CDMA phones in working is around 10 to Class-B to enable long talk time considering that the typical power range of CDMA phones in working is around 10 to 15dBm, i.e., liner range. The power amplifier which employs two GaAs FETs with good linerity at a low operating point has duplexer cuplexer circuit to separate two frequency bands at input and output stage. Electromagnetic analysis for via holes and coupling between narrow transmission lines is included to design a circuit. Moduld size of 0.96CC($22{\times}14.5{\times}3mm^3$) and maximum module current of 130mA at output power range, 10 to 15dBm are attained. The power amplifer module has achieved ACPR performance with 2 to 3dB marging from IS-95 requirement at output powers, 23.5dBm for PCS and 28dBm for 800MHz CDMA respectively.

• Journal of KIISE:Computer Systems and Theory
• /
• v.26 no.11
• /
• pp.1373-1381
• /
• 1999

이 논문에서는 통신망 설계 응용분야의 문제를 그래프 이론 문제로써 고려해 보았다. 개별 기업체가 서로 떨어진 두 곳을 연결하고자 할 때 공용통신망의 회선을 빌려 통신망을 구축하게 되는데 많은 경우 여러 종류의 회선들이 공급됨으로 어떤 회선을 선택하느냐의 문제가 생긴다. 일반적으로 빠른 회선(low delay)은 느린 회선(high delay)에 비해 비싸다. 그러나 서비스의 질(Quality of Service)이라는 요구사항이 종종 종단지연(end-to-end delay)시간에 의해 결정되므로, 무조건 낮은 가격의 회선만을 사용할 수는 없다. 결국 개별 기업체의 통신망을 위한 통로를 공용 통신망 위에 덮어씌워(overlaying) 구축하는 것의 여부는 두 개의 상반된 인자인 가격과 속도의 조절에 달려 있다. 따라서 일반적인 최소경로 찾기의 변형이라 할 수 있는 다음의 문제가 본 논문의 관심사이다. 두 개의 지점을 연결하는데 종단지연시간의 한계를 만족하면서 최소경비를 갖는 경로에 대한 해결을 위하여, 그래프 채색(coloring) 문제와 최단경로문제를 함께 포함하는 그래프 이론의 문제로 정형화시켜 살펴본다. 배낭문제로의 변환을 통해 이 문제는 {{{{NP-complete임을 증명하였고 {{{{O($\mid$E$\mid$D_0 )시간에 최적값을 주는 의사선형 알고리즘과O($\mid$E$\mid$)시간의 근사 알고리즘을 보였다. 특별한 경우에 대한 {{{{O($\mid$V$\mid$ + $\mid$E$\mid$)시간과 {{{{O($\mid$E$\mid$^2 + $\mid$E$\mid$$\mid$V$\mid$log$\mid$V$\mid$)시간 알고리즘을 보였으며 배낭 문제의 해결책과 유사한 그리디 휴리스틱(greedy heuristic) 알고리즘이 그물 구조(mesh) 그래프 상에서 좋은 결과를 보여주고 있음을 실험을 통해 확인해 보았다.Abstract This paper considers a graph-theoretic problem motivated by a telecommunication network optimization. When a private organization wishes to connect two sites by leasing physical lines from a public telecommunications network, it is often the cases that several categories of lines are available, at different costs. Typically a faster (low delay) lines costs more than a slower (high delay) line. However, low cost lines cannot be used exclusively because the Quality of Service (QoS) requirements often impose a bound on the end-to-end delay. Therefore, overlaying a path on the public network involves two diametrically opposing factors: cost and delay. The following variation of the standard shortest path problem is thus of interest: the shortest route between the two sites that meets a given bound on the end-to-end delay. For this problem we formulate a graph-theoretical problem that has both a shortest path component as well as coloring component. Interestingly, the problem could be formulated as a knapsack problem. We have shown that the general problem is NP-complete. The optimal polynomial-time algorithms for some special cases and one heuristic algorithm for the general problem are described.