• Proceedings of the Acoustical Society of Korea Conference
• /
• spring
• /
• pp.40-43
• /
• 1999

본 연구에서는 고립단어 인식시스템에 사용되고 있는 DTW(DynamicTimeWarping) 알고리즘의 계산량을 줄일 수 있는 방법을 제안한다. 일반적으로 고립단어 인식시 가장 인식률이 좋은 알고리즘은 DW라고 알려져 있으나, 인식대상어휘가 늘어나면 계산량이 비례해서 늘어나고 인식률이 저하되는 단점이 있으므로 일반적으로 200단어 이하의 어휘에만 사용되고 있다. 따라서 대상어휘를 감소시켜 계산량을 줄이기 위해 본 논문에서는 유성/무성/묵음 (V/U/S) 정보를 이용하여 코드워드를 구성하고 같은 코드워드에 해당되는 단어들을 추출해이들 만을 비교대상 어휘로 제한하므로서 DW 알고리즘을 적용할 대상 어휘수를 줄이는 방법을 사용하여 계산 속도를 향상시켰다 또한 입력 단어와 대상 단어와의 누적거리 계산 시 끝점 정보 뿐 만 아니라 유성/무성/묵음 경계 정보를 이용하여 piecewise DTW를 구현함으로서 탐색 영역을 축소함으로써 추가적인 계산량 감소가 가능하다. 따라서 상기 기법들을 이용하면 PC상에서도 DTW를 이용한 대어휘 고립단어 음성 인식기의 구현이 가능할 것이다.

• Communications for Statistical Applications and Methods
• /
• v.20 no.4
• /
• pp.283-290
• /
• 2013

Subjects can experience two types of recurrent events in a longitudinal study. In addition, there may exist intermittent dropouts that results in repeated observation gaps during which no recurrent events are observed. Therefore, theses periods are regarded as non-risk status. In this paper, we consider a special case where information on the observation gap is incomplete, that is, the termination time of observation gap is not available while the starting time is known. For a statistical inference, incomplete termination time is incorporated in terms of interval-censored data and estimated with two approaches. A shared frailty effect is also employed for the association between two recurrent events. An EM algorithm is applied to recover unknown termination times as well as frailty effect. We apply the suggested method to young drivers' convictions data with several suspensions.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.20 no.1
• /
• pp.7-14
• /
• 1983

Linear feature such as lines and curves are one of important features in image processing. In this paper, new method of linear feature detection is suggested. Also, we have studied approximation technique which transforms detected linear feature into data structure for the practical. This method is based on graph theory and principle of this method is based on minimal spanning tree concept which is widely used in edge linking process. By postprocessing, Hairs and inconsistent line segments are removed. To approximate and describe traced linear feature, piecewise linear approximation is adapted. The algorithm is demonstrated through computer simulations.

• Proceedings of the Korean Geotechical Society Conference
• /
• 2001.03a
• /
• pp.363-370
• /
• 2001

A computationally efficient algorithm to analyze a group pile behavior is proposed by consideration of both soil-pile and pile-cap interactions. Using toad transfer method the nonlinear characteristics of the soil-pile interaction for a single pile is modeled by piecewise linear soil springs (p-y, t-z, and q-z curves). Beam-column method, one of the most practical approaches, is used for numerical modeling of the soil-pile system. In addition to the group effect resulting from the soil-pile-soil interaction, for a more realistic analysis it is essential to consider the effect of pile-cap interaction including geometric configuration of the piles in a group and conectivity conditions between piles and the cap. This paper mainly focuses on the pile-cap interaction and the development of a rational numerical procedure of its incorporation with the beam-column method.

• Journal of the Korean Institute of Gas
• /
• v.4 no.4 s.12
• /
• pp.58-64
• /
• 2000

Data compression and retrieval method are investigated for the effective utilization of measured process data. In this paper, a new data compression method, Clustering Compression(CC), which is based on the k-means clustering algorithm and piecewise linear approximation method is suggested. Case studies on industrial data set showed the superior performance of clustering based techniques compared to other conventional methods and showed that CC could handle the compression of multi-dimensional data.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.2 no.3
• /
• pp.227-230
• /
• 2002

A new iris segmentation and recognition method is described. Combining a statistical classification and elastic boundary fitting, the iris is first segmented robustly and accurately. Once the iris is segmented, one-dimensional signals are computed in the iris and decomposed into multiple frequency bands. Each decomposed signal is approximated by a piecewise linear curve connecting a small set of node points. The node points represent features of each signal. The similarity measture between two iris images is the normalized cross-correlation coefficients between simplified signals.

• Speech Sciences
• /
• v.10 no.3
• /
• pp.263-277
• /
• 2003

One of major problems in speech recognition is performance degradation due to the mismatch between the training and test environments. Recently, Stereo-based Piecewise LInear Compensation for Environments (SPLICE), which is frame-based bias removal algorithm for cepstral enhancement using stereo training data and noisy speech model as a mixture of Gaussians, was proposed and showed good performance in noisy environments. In this paper, we propose several methods to improve the conventional SPLICE. First we apply Cepstral Mean Subtraction (CMS) as a preprocessor to SPLICE, instead of applying it as a postprocessor. Secondly, to compensate residual distortion after SPLICE processing, two-stage SPLICE is proposed. Thirdly we employ phonetic information for training SPLICE model. According to experiments on the Aurora 2 database, proposed method outperformed the conventional SPLICE and we achieved a 50% decrease in word error rate over the Aurora baseline system.

• Journal of Power Electronics
• /
• v.9 no.1
• /
• pp.124-131
• /
• 2009

This paper presents a PWM dc-ac converter regulation using a Single Input Fuzzy PI Controller (SIFPIC). The SIFPIC is derived from the signed distanced method, which is a simplification of a conventional fuzzy controller. The simplification results in a one-dimensional rule table, that allows its control surface to be approximated by a piecewise linear relationship. The controller multi-loop structure is comprised of an outer voltage and an inner current feedback loop. To verify the performance of the SIFPIC, a low power PWM dc-ac converter prototype is constructed and the proposed control algorithm is implemented. The experimental results show that the SIFPIC performance is comparable to a conventional Fuzzy PI controller, but with a much reduced computation time.

• Proceedings of the KIEE Conference
• /
• 2005.07c
• /
• pp.2315-2317
• /
• 2005

In this paper, an iterative method to model the electromagnetic heating of electrically large lossy dielectrics is presented. Frequency domain finite element (FEM) solutions of the wave equation are determined for the lossy inhomogeneous dielectric as the material properties are change with temperature and time. The power absorbed from microwave losses is applied to a finite element time domain (FETD) calculation of the heat diffusion equation. Time steps appropriate for updating the piecewise material properties in the wave equation and the time stepping of the heat equation are presented. The effects of preheating and source frequency are investigated.

• Proceedings of the KIEE Conference
• /
• 2003.11c
• /
• pp.946-949
• /
• 2003

In this paper, a DDFS(Direct Digital Frequency Synthesis)chip has been designed focusing on the reduction of ROM size and implemented using FPGA. When calculating the sine value for the input phase value, we used the Taylor series expansion approximation method to reduce the number of addresses of ROM. We also used the piecewise straight line approximation method, ie, the stored value int the ROM is the difference of the sine value and the straight line approximation. Using this method, we could reduce four bits for each ROM data.