• Communications for Statistical Applications and Methods
• /
• v.19 no.4
• /
• pp.597-605
• /
• 2012

Evidence framework enables us to determine the tuning constant in a penalized likelihood formula. We apply the framework to the estimating parameters of normal mixtures. Evidence, which is a solely data-dependent measure, can be evaluated by Laplace approximation. According to a synthetic data simulation, we found that the proper values of the tuning constant can be systematically obtained.

• Communications for Statistical Applications and Methods
• /
• v.8 no.2
• /
• pp.347-355
• /
• 2001

We propose a method for detecting influential observations that have a large influence on the likelihood ratio test statistic for the two sets of variables are uncorrelated with one another. For this purpose we derive a local influence measure for the likelihood ratio test statistic under certain perturbation scheme. An illustrative example is given to show the effectiveness of the proposed method on the identification of influential observations.

• MALSORI
• /
• no.67
• /
• pp.79-102
• /
• 2008

In this paper, we investigate the relation between Bayes factor and likelihood ratio test (LRT) approaches and apply the neighborhood information of Bayes factor to building an alternate hypothesis model of the LRT system. To consider the neighborhood approaches, we contemplate a distance measure between models and algorithms to be applied. We also evaluate several methods to improve performance of utterance verification using neighborhood information. Among these methods, the system which adopts anti-models built by collecting mixtures of neighborhood models obtains maximum error rate reduction of 17% compared to the baseline, linear and weighted combination of neighborhood models.

• Proceedings of the KSPS conference
• /
• 2005.04a
• /
• pp.95-98
• /
• 2005

Many utterance verification (UV) algorithms have been studied to reject out-of-vocabulary (OOV) in speech recognition systems. Most of conventional confidence measures for UV algorithms are mainly based on log likelihood ratio test, but these measures take much time to evaluate the alternative hypothesis or anti-model likelihood. We propose a novel confidence measure which makes use of a momentary best scored state sequence during Viterbi search. Our approach is more efficient than conventional LRT-based algorithms because it does not need to build anti-model or to calculate the alternative hypothesis. The proposed confidence measure shows better performance in additive noise-corrupted speech as well as clean speech.

• Journal of Information Technology Applications and Management
• /
• v.24 no.1
• /
• pp.25-32
• /
• 2017

For repairable software systems, the Mean Time Between Failure (MTBF) is used as a measure of software system stability. Therefore, the evaluation of software reliability requirements or reliability characteristics can be applied MTBF. In this paper, we want to compare MTBF in terms of parameter estimation using Makeham life distribution. The parameter estimates used the least square method which is regression analyzer method and the maximum likelihood method. As a result, the MTBF using the least square method shows a non-decreased pattern and case of the maximum likelihood method shows a non-increased form as the failure time increases. In comparison with the observed MTBF, MTBF using the maximum likelihood estimation is smallerd about difference of interval than the least square estimation which is regression analyzer method. Thus, In terms of MTBF, the maximum likelihood estimation has efficient than the regression analyzer method. In terms of coefficient of determination, the mean square error and mean error of prediction, the maximum likelihood method can be judged as an efficient method.

• The Korean Journal of Applied Statistics
• /
• v.4 no.2
• /
• pp.117-127
• /
• 1991

Coefficient of determination $R^2$ is most frequently used descriptive measure in practical use of linear regression analysis. But there have been controversies on defining this measure in the cases of linear regression without the intercept, weighted linear regression and robust linear regression. Several authors such as Kvalseth(1985) and Willet and Singer(1988) proposed many variations of $R^2$ to meet the situations. However, theire measures are not satisfactory due to the lack of a universal principle. In this study, we propose a unfied approach to defining the coefficient of determination $R^2$ using the concept of likelihood distance. This new measure is in good accordance with typical $R^2$ in linear regression and, moreover, can be applied to nonlinear regression models and generalized linear models such as logit and log-linear models.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.45 no.4
• /
• pp.100-106
• /
• 2008

In this paper we apply a discriminative weight training employing power spectral flatness measure (PSFM) to a statistical model-based voice activity detection (VAD) in various noise environments. In our approach, the VAD decision rule is expressed as the geometric mean of optimally weighted likelihood ratio test (LRT) based on a minimum classification error (MCE) method which is different from the previous works in th at different weights are assigned to each frequency bin and noise environments depending on PSFM. According to the experimental results, the proposed approach is found to be effective for the statistical model-based VAD using the LRT.

• Journal of information and communication convergence engineering
• /
• v.13 no.3
• /
• pp.152-158
• /
• 2015

We consider the problem of designing independently operating local quantizers at nodes in distributed estimation systems, where many spatially distributed sensor nodes measure a parameter of interest, quantize these measurements, and send the quantized data to a fusion node, which conducts the parameter estimation. Motivated by the discussion that the estimation accuracy can be improved by using the quantized data with a high probability of occurrence, we propose an iterative algorithm with a simple design rule that produces quantizers by searching boundary values with an increased likelihood. We prove that this design rule generates a considerably reduced interval for finding the next boundary values, yielding a low design complexity. We demonstrate through extensive simulations that the proposed algorithm achieves a significant performance gain with respect to traditional quantizer designs. A comparison with the recently published novel algorithms further illustrates the benefit of the proposed technique in terms of performance and design complexity.

• Phonetics and Speech Sciences
• /
• v.3 no.3
• /
• pp.91-98
• /
• 2011

In this paper, a multi-channel speech enhancement method using the linearly constrained minimum variance (LCMV) algorithm and a variable learning rate control is proposed. To control the learning rate for adaptive filters of the LCMV algorithm, the direction of arrival (DOA) is measured for each short-time input signal and the likelihood function of the target speech presence is estimated to control the filter learning rate. Using the likelihood measure, the learning rate is increased during the pure noise interval and decreased during the target speech interval. To optimize the parameter of the mapping function between the likelihood value and the corresponding learning rate, an exhaustive search is performed using the Bark's scale distortion (BSD) as the performance index. Experimental results show that the proposed algorithm outperforms the conventional LCMV with fixed learning rate in the BSD by around 1.5 dB.

• Nuclear Engineering and Technology
• /
• v.48 no.1
• /
• pp.114-128
• /
• 2016

The development of operational performance indicators is of utmost importance for nuclear power plants, since they measure, track, and trend plant operation. Leading indicators are ideal for reducing the likelihood of consequential events. This paper describes the operational data analysis of the information contained in the Corrective Action Program. The methodology considers human error and organizational factors because of their large contribution to consequential events. The results include a tool developed from the data to be used for the identification, prediction, and reduction of the likelihood of significant consequential events. This tool is based on the resilience curve that was built from the plant's operational data. The stress is described by the number of unresolved condition reports. The strain is represented by the number of preventive maintenance tasks and other periodic work activities (i.e., baseline activities), as well as, closing open corrective actions assigned to different departments to resolve the condition reports (i.e., corrective action workload). Beyond the identified resilience threshold, the stress exceeds the station's ability to operate successfully and there is an increased likelihood that a consequential event will occur. A performance indicator is proposed to reduce the likelihood of consequential events at nuclear power plants.