• The Journal of the Acoustical Society of Korea
• /
• v.21 no.8
• /
• pp.735-743
• /
• 2002

In this paper, we propose three new text-independent speaker identification methods. At first, to exclude the frames not having enough features of speaker's vocal from calculation of the maximum likelihood, we propose the FS(Frame Selection) method. This approach selects the important frames by evaluating the difference between the biggest likelihood and the second in each frame, and uses only the frames in calculating the score of likelihood. Our secondly proposed, called the Hybrid, is a combined version of the FS and WMR(Weighting Model Rank). This method determines the claimed speaker using exponential function weights, instead of likelihood itself, only on the selected frames obtained from the FS method. The last proposed, called MWMR (Modified WMR), considers both original likelihood itself and its relative position, when the claimed speaker is determined. It is different from the WMR that take into account only the relative position of likelihood. Through the experiments of the speaker identification, we show that the all the proposed have higher identification rates than the ML. In addition, the Hybrid and MWMR have higher identification rate about 2% and about 3% than WMR, respectively.

• Journal of the Korean Statistical Society
• /
• v.26 no.1
• /
• pp.23-30
• /
• 1997

We show that the Maximum Penalized Likelihood Estimate uniquely exits in a Sobolve spece which consists of bivariate density functions. The Maximum Penalized Likehood Estimate is represented as the square of the sum of the solutions of the Modified Helmholtz's equation on the compact subset of R$^{2}$.

• Journal of the Korean Data and Information Science Society
• /
• v.23 no.1
• /
• pp.89-98
• /
• 2012

In this paper, the ratio of parameters in two independent Maxwell distributions is parameter of interest. We proposed test statistics, which converge to standard normal distribution, based on likelihood function. The exact distribution for testing the ratio is hard to obtain. We proposed the signed log-likelihood ratio statistic and the modified signed log-likelihood ratio statistic for testing the ratio. Through simulation, we show that the modified signed log-likelihood ratio statistic converges faster than signed log-likelihood ratio statistic to standard normal distribution. We compare two statistics in terms of type I error and power. We give an example using real data.

• Journal of the Korean Statistical Society
• /
• v.9 no.1
• /
• pp.13-29
• /
• 1980

The statistical theory of factor analysis is briefly reviewed with emphasis on the maximum-likelihood method. A modified version of Joreskog(1975) is used for the implementation of the maximum-likelihood method. For the minimization of the conditional minimum function, an adaptive Newton-Raphson method is applied.

• Journal of the Korean Data and Information Science Society
• /
• v.25 no.1
• /
• pp.195-209
• /
• 2014

In this paper, we derive the estimators of the location parameter and the scale parameter in a logistic distribution based on multiply type-II censored samples by the approximate maximum likelihood estimation method. We use four modified empirical distribution function (EDF) types test for the logistic distribution based on multiply type-II censored samples using proposed approximate maximum likelihood estimators. We also propose the modified normalized sample Lorenz curve plot for the logistic distribution based on multiply type-II censored samples. For each test, Monte Carlo techniques are used to generate the critical values. The powers of these tests are also investigated under several alternative distributions.

• Journal of the Korean Data and Information Science Society
• /
• v.9 no.2
• /
• pp.289-298
• /
• 1998

We investigate the distribution of likelihood ratio test(LRT) of null hypothesis a sample is from single gamma with unknown shape and scale against the alternative hypothesis a sample is from a mixture of two gammas, each with unknown scale and unknown (but equal) scale. To obtain stable maximum likelihood estimates(MLE) of a mixture of two gamma distributions, the EM(Dempster, Laird, and Robin(1977))and Modified Newton(Jensen and Johansen(1991)) algorithms were implemented. Based on EM, we made a simple structure likelihood equation for each parameter and could obtain stable solution by Modified Newton Algorithms. Simulation study was conducted to investigate the distribution of LRT for sample size n = 25, 50, 75, 100, 50, 200, 300, 400, 500 with 2500 replications. To determine the small sample distribution of LRT, I considered the model of a gamma distribution with shape parameter equal to 1 + f(n) and scale parameter equal to 2. The simulation results indicate that the null distribution is essentially invariant to the value of the shape parameter. Modeling of the null distribution indicates that it is well approximated by a gamma distribution with shape parameter equal to the quantity $0.927+1.18/\sqrt{n}$ and scale parameter equal to 2.16.

• Journal of Applied Reliability
• /
• v.1 no.2
• /
• pp.183-192
• /
• 2001

Jelinski-Moranda model and modified Jelinski-Moranda model in software reliability are studied and we consider maximum likelihood estimator and Bayes estimates of the number of faults and the fault-detection rate per fault. A gibbs sampling approach is employed to compute the Bayes estimates, future survival function is examined. Model selection based on prequential likelihood of the conditional predictive ordinates. A numerical example with simulated data set is given.

• Journal of Korea Society of Industrial Information Systems
• /
• v.4 no.2
• /
• pp.39-43
• /
• 1999

We consider generalized exponential regression model with randomly censored data and propose a modified Fisher scoring method which estimates the model parameters. For this, the likelihood equations are derived and then the estimating algorithm is developed. We illustrate the proposed method using a real data.

• MALSORI
• /
• no.63
• /
• pp.125-137
• /
• 2007

Recently, many techniques have been proposed to improve the noise robustness for speaker verification. In this paper, we consider the feature recombination technique in multi-band approach. In the conventional feature recombination for speaker verification, to compute the likelihoods of speaker models or universal background model, whole feature components are used. This computation method is not effective in a view point of multi-band approach. To deal with non-effectiveness of the conventional feature recombination technique, we introduce a subband likelihood computation, and propose a modified feature recombination using subband likelihoods. In decision step of speaker verification system in noise environments, a few very low likelihood scores of a speaker model or universal background model cause speaker verification system to make wrong decision. To overcome this problem, a reliable feature selection method is proposed. The low likelihood scores of unreliable feature are substituted by likelihood scores of the adaptive noise model. In here, this adaptive noise model is estimated by maximum a posteriori adaptation technique using noise features directly obtained from noisy test speech. The proposed method using subband-based reliable feature selection obtains better performance than conventional feature recombination system. The error reduction rate is more than 31 % compared with the feature recombination-based speaker verification system.

• Water Engineering Research
• /
• v.4 no.3
• /
• pp.155-173
• /
• 2003

Parameters and quantiles of the 2-parameter generalized Pareto distribution were estimated using the methods of regular moments, modified moments, probability weighted moments, linear moments, maximum likelihood, and entropy for Monte Carlo-generated samples. The performance of these seven estimators was statistically compared, with the objective of identifying the most robust estimator. It was found that in general the methods of probability-weighted moments and L-moments performed better than the methods of maximum likelihood estimation, moments and entropy, especially for smaller values of the coefficient of variation and probability of exceedance.