• 응용통계연구
• /
• 제15권2호
• /
• pp.337-354
• /
• 2002

본 논문에서는 오차성분이 계열상관을 갖는 불균형 랜덤모형에서 분산성분의 추정방법에 대하여 연구하였다. 분산성분에 대한 추정량으로 조건부 ANOVA(cANOVA), ML및 REML추정량등을 유도하였으며, 계열상관값과 불균형의 정도에 따른 추정량의 변동성을 추정량의 분위수를 이용하는 EQDGs플롯을 이용하여 비교하였다. 모의실험결과 cANOVA추정방법은 불균형의 정도에는 추정량값이 크게 영향을 받지 않는 것으로 나타났으나 계열상관값의 증가에 따라서는 변동성을 보이고 있다. 불균형의 정도와 계열상관값을 동시에 고려하는 경우에는 ML추정방법이 cANOVA, REML추정방법보다 변동성이 안정적으로 나타났다.

• Communications for Statistical Applications and Methods
• /
• 제10권3호
• /
• pp.895-907
• /
• 2003

Recent demands for representing the location of multivariate data produce various multivariate medians such as Tukey median, Oja median and spatial median. They are considered as multivariate versions of the median which is widely recognized as a robust alternative to the arithmetic mean. Many studies show that those multivariate median preserve the robustness. However, the effectiveness of those medians is not fully identified. In this note the relative efficiencies of the multivariate medians are investigated in various configurations under the bivariate t-distribution. It is shown that Tukey median outperforms the others in most configurations.

• Journal of the Korean Data and Information Science Society
• /
• 제14권3호
• /
• pp.673-685
• /
• 2003

In this paper we consider the problem of identifying and testing outliers in linear regression. First we consider the problem for testing the null hypothesis of no outliers. A test based on the ratio of two residual scale estimates is proposed. We show the asymptotic distribution of the test statistics by Monte Carlo simulation and investigate its properties. Next we consider the problem of identifying the outliers. A forward sequential procedure using the suggested test is proposed and shown to perform fairly well. Unlike other forward procedures, the present one is unaffected by masking and swamping effects because the test statistic is based on robust scale estimate.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2002년도 ICCAS
• /
• pp.96.3-96
• /
• 2002

A scheme of clarifying fault feature in the residual is given for model-based fault detection and diagnosis of control systems. It is based on the residual generation using a robust filter and the noise suppresion in test statistics of the residual by multi-scale discrete wavelet transform. By clarifying the fault feature in the residual, the difficulties of existing model based approaches via adopting a threshold can be overcomed and it has advantage of taking the false alarm and missed detection into acount at the same time, which can make the fault detection and diagnosis easy and correct. To show the effectiveness of our approach, the simulation results are illustrated for a linear syste...

• 대한음성학회:학술대회논문집
• /
• 대한음성학회 2005년도 춘계 학술대회 발표논문집
• /
• pp.91-94
• /
• 2005

The performance of speech recognition is degraded by the mismatch between training and test environments. Many methods have been presented to compensate for additive noise and channel effect in the cepstral domain, and Cepstral Mean Subtraction (CMS) is the representative method among them. Recently, high order cepstral moment normalization method has introduced to improve recognition accuracy. In this paper, we apply high order moment normalization method and smoothing filter for real-time processing. In experiments using Aurora2 DB, we obtained error rate reduction of 49.7% with the proposed algorithm in comparison with baseline system.

• Communications for Statistical Applications and Methods
• /
• 제20권4호
• /
• pp.311-319
• /
• 2013

Cooray and Ananda (2005) proposed a composite lognormal-Pareto model to analyze loss payment data in the actuarial and insurance industries. Their model is based on a lognormal density up to an unknown threshold value and a two-parameter Pareto density. In this paper, we implement the minimum density power divergence estimation for the composite lognormal-Pareto density. We compare the performances of the minimum density power divergence estimator (MDPDE) and the maximum likelihood estimator (MLE) by simulations and an example. The minimum density power divergence estimator performs reasonably well against various violations in the distribution. The minimum density power divergence estimator better fits small observations and better resists against extraordinary large observations than the maximum likelihood estimator.

• 천문학논총
• /
• 제20권1호
• /
• pp.49-53
• /
• 2005

We are often faced with the task of having to estimate the amplitude of a source signal in the presence of a background. In the simplest case, the background can be taken as being flat, and of unknown magnitude B, and the source signal of interest assumed to be the amplitude A of a peak of known shape and position. We present a robust method to find the most probable values of A and B by applying the one-dimensional Newton-Raphson method. In the derivation of the formula, we adopted the Bayesian statistics and assmumed Poisson distribution so that the results could be applied to the analysis of very weak signals, as observed in FIMS (Far-ultraviolet IMaging Spectrogaph).

• Journal of the Korean Data and Information Science Society
• /
• 제16권4호
• /
• pp.1053-1066
• /
• 2005

In this paper we consider the problem of identifying and testing outliers in linear regression. First we consider the use of the so-called scale ratio tests for testing the null hypothesis of no outliers. This test is based on the ratio of two residual scale estimates. We show the asymptotic distribution of the test statistics and investigate its properties. Next we consider the problem of identifying the outliers. A forward sequential procedure using the suggested test is proposed. The new method is compared with classical procedure in the real data example. Unlike other forward procedures, the present one is unaffected by masking and swamping effects because the test statistic is based on robust scale estimate.

• Computers and Concrete
• /
• 제20권1호
• /
• pp.23-29
• /
• 2017

In this study engineered steel fibres used as reinforcement for concrete were characterized by number of key mechanical and spatial parameters, which are easy to measure and quantify. Such commonly used parameters as length, diameter, fibre intrinsic efficiency ratio (FIER), hook geometry, tensile strength and ductility were considered. Effective classification of various fibres was demonstrated using simple multivariate computations involving principal component analysis (PCA). Contrary to univariate data mining approach, the proposed analysis can be efficiently adapted for fast, robust and direct classification of engineered steel fibres. The results have revealed that in case of particular spatial/geometrical conditions of steel fibres investigated the FIER parameter can be efficiently replaced by a simple aspect ratio. There is also a need of finding new parameters describing properties of steel fibre more precisely.

• 음성과학
• /
• 제15권4호
• /
• pp.85-96
• /
• 2008

Automatic speech recognition has severe performance degradation under noisy environments. To cope with the noise problem, many methods have been proposed. Most of them focused on noise-robust features or model adaptation. However, researchers have overlooked utterance verification (UV) under noisy environments. In this paper we discuss UV problems based on the normalized confidence measure. First, we show that UV performance is also degraded in noisy environments with the experiments of an isolated word recognition. Then we observe how the degradation of UV performances is suffered. Based on the UV experiments we propose a modeling method of the statistics of phone confidences using sigmoid functions. For obtaining the parameters of the sigmoidal models, the particle swarm optimization (PSO) is adopted. The proposed method improves 20% rejection performance. Our experimental results show that the PSO-NCM can apply noise speech recognition successfully.