• MALSORI
• /
• no.60
• /
• pp.109-123
• /
• 2006

Since computing the log-likelihood of Gaussian mixture density is a major computational burden for the speech recognizer based on the continuous HMM, several techniques have been proposed to reduce the number of mixtures to be used for recognition. In this paper, we propose a modified Bucket Box Intersection (BBI) algorithm, in which two relative thresholds are employed: one is the relative threshold in the conventional BBI algorithm and the other is used to reduce the number of the Gaussian boxes which are intersected by the hyperplanes at the boxes' edges. The experimental results show that the proposed algorithm reduces the number of Gaussian mixtures by 12.92% during the recognition phase with negligible performance degradation compared to the conventional BBI algorithm.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.13 no.5
• /
• pp.512-519
• /
• 2003

In this paper, we propose a neuro-fuzzy modeling to improve the performance using the hierarchical clustering and Gaussian Mixture Model(GMM). The hierarchical clustering algorithm has a property of producing unique parameters for the given data because it does not use the object function to perform the clustering. After optimizing the obtained parameters using the GMM, we apply them as initial parameters for Adaptive Network-based Fuzzy Inference System. Here, the number of fuzzy rules becomes to the cluster numbers. From this, we can improve the performance index and reduce the number of rules simultaneously. The proposed method is verified by applying to a neuro-fuzzy modeling for Box-Jenkins s gas furnace data and Sugeno's nonlinear system, which yields better results than previous oiles.

• The Journal of the Acoustical Society of Korea
• /
• v.19 no.2E
• /
• pp.27-31
• /
• 2000

Evaluating log-likelihood of Gaussian mixture density is major computational burden for the keyword spotting system using continuous HMM. In this paper, we employ the bucket box intersection (BBI) algorithm to reduce the computational complexity of keyword spotting. We make some modification in implementing BBI algorithm in order to increase the discrimination ability among the keyword models. According to our keyword spotting experiments, the modified BBI algorithm reduces 50% of log-likelihood computations without performance degradation, while the original BBI algorithm under the same condition reduces only 30% of log-likelihood computations.

• Bulletin of the Korean Mathematical Society
• /
• v.50 no.3
• /
• pp.935-949
• /
• 2013

In this paper, we study surfaces in $\mathb{E}^3$ whose Gauss map G satisfies the equation ${\Box}G=f(G+C)$ for a smooth function $f$ and a constant vector C, where ${\Box}$ stands for the Cheng-Yau operator. We focus on surfaces with constant Gaussian curvature, constant mean curvature and constant principal curvature with such a property. We obtain some classification and characterization theorems for these kinds of surfaces. Finally, we give a characterization of surfaces whose Gauss map G satisfies the equation ${\Box}G={\lambda}(G+C)$ for a constant ${\lambda}$ and a constant vector C.

• 제어로봇시스템학회:학술대회논문집
• /
• 1997.10a
• /
• pp.442-445
• /
• 1997

We present a hybrid self-tuning method of fuzzy inference systems with hyper elliptic Gaussian membership functions using genetic algorithm(GA) and back-propagation algorithm. The proposed self-tuning method has two phases : one is the coarse tuning process based on GA and the other is the fine tuning process based on back-propagation. But the parameters which is obtained by a GA are near optimal solutions. In order to solve the problem in GA applications, it uses a back-propagation algorithm, which is one of learning algorithms in neural networks, to finely tune the parameters obtained by a GA. We provide Box-Jenkins time series to evaluate the advantage and effectiveness of the proposed approach and compare with the conventional method.

• Journal of applied mathematics & informatics
• /
• v.20 no.1_2
• /
• pp.445-454
• /
• 2006

We describe a solution to the Ornstein-Uhlenbeck equation $\frac{dI}{dt}-\frac{1}{\tau}$I(t)=cV(t) where V(t) is a constant multiple of a Gaussian white noise. Our solution is based on a discrete set of Gaussian white noise obtained by taking sample points from a sum of single frequency harmonics that have random amplitudes, random frequencies, and random phases. Hence, it is different from the solution by the standard random walk using random numbers generated by the Box-Mueller algorithm. We prove that the power of the signal has the additive property, from which we derive that the Lyapunov characteristic exponent for our solution is positive. This compares with the solution by other methods where the noise is kept to be in an error range so that its Lyapunov exponent is negative.

• Journal of Korean Society for Atmospheric Environment
• /
• v.19 no.5
• /
• pp.569-581
• /
• 2003

A method of synthetic time series generation was developed and applied to the simulation of homogeneous turbulence in a periodic 3 - D box and the hourly wind data simulation. The method can simulate almost exact sample auto and cross correlations of multiple time series and control non-Gaussian distribution. Using the turbulence simulation, influence of correlations, non-Gaussian distribution, and one-direction anisotropy on homogeneous structure were studied by investigating the spatial distribution of turbulence kinetic energy and enstrophy. An hourly wind data of Typhoon Robin was used to illustrate a capability of the method to simulate sample cross correlations of multiple time series. The simulated typhoon data shows a similar shape of fluctuations and almost exactly the same sample auto and cross correlations of the Robin.

• Journal of Korea Water Resources Association
• /
• v.36 no.3 s.134
• /
• pp.399-411
• /
• 2003

An approach determining a confidence interval of Nash's observed mean instantaneous unit hydrograph is developed. In the approach, both two parameters are treated as correlated gaussian random variables based on the theory of Box-Cox transformation and the regional similarity relation, so that linear statistical parameter estimation is possible. A parametric bootstrap method is adopted to give the confidence interval of the mean observed hydrograph. The proposed methodology is also applicable to estimate the parameters of Nash's model for un-gauged basins. An application to a watershed has shown that the proposed approach is adequate to assess the uncertainty of the Nash's hydrograph and to evaluate parameters for un-gauged basins.

• Smart Structures and Systems
• /
• v.25 no.5
• /
• pp.593-604
• /
• 2020

Temperature may have more significant influences on structural responses than operational loads or structural damage. Therefore, a comprehensive understanding of temperature distributions has great significance for proper design and maintenance of bridges. In this study, the temperature distribution of the steel box girder is systematically investigated based on the structural health monitoring system (SHMS) of the Sutong Cable-stayed Bridge. Specifically, the characteristics of the temperature and temperature difference between different measurement points are studied based on field temperature measurements. Accordingly, the probability density distributions of the temperature and temperature difference are calculated statistically, which are further described by the general formulas. The results indicate that: (1) the temperature and temperature difference exhibit distinct seasonal characteristics and strong periodicity, and the temperature and temperature difference among different measurement points are strongly correlated, respectively; (2) the probability density of the temperature difference distribution presents strong non-Gaussian characteristics; (3) the probability density function of temperature can be described by the weighted sum of four Normal distributions. Meanwhile, the temperature difference can be described by the weighted sum of Weibull distribution and Normal distribution.

• Journal of Satellite, Information and Communications
• /
• v.8 no.1
• /
• pp.77-82
• /
• 2013

The object tracking method using particle filtering has been proved successful since it is based on the Monte Carlo simulation to estimate the posterior distribution of the state vector that is nonlinear and non-Gaussian in the real-world situation. In this paper, we present two nobel methods that can improve the performance of the object tracking algorithm based on the particle filtering. First one is the feedback method that replace the low-weighted tracking sample by the estimated state vector in the previous frame. The second one is an tracking box correction method to find an confidence interval of back projection probability on the estimated candidate object area. An sample propagation equation is also presented, which is obtained by experiments. We designed well-organized test data set which reflects various challenging circumstances, and, by using it, experimental results proved that the proposed methods improves the traditional particle filter based object tracking method.