• The KIPS Transactions:PartB
• /
• v.17B no.4
• /
• pp.303-308
• /
• 2010

In this paper, we propose a medical image registration technique combining the gradient vector flow and modified conditional entropy. The registration is conducted by the use of a measure based on the entropy of conditional probabilities. To achieve the registration, we first define a modified conditional entropy (MCE) computed from the joint histograms for the area intensities of two given images. In order to combine the spatial information into a traditional registration measure, we use the gradient vector flow field. Then the MCE is computed from the gradient vector flow intensity (GVFI) combining the gradient information and their intensity values of original images. To evaluate the performance of the proposed registration method, we conduct experiments with our method as well as existing method based on the mutual information (MI) criteria. We evaluate the precision of MI- and MCE-based measurements by comparing the registration obtained from MR images and transformed CT images. The experimental results show that the proposed method is faster and more accurate than other optimization methods.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.14 no.2
• /
• pp.23-33
• /
• 2004

Because of a rapid growth of internet environment, it is also fast increasing to exchange message using e-mail. But, despite the convenience of e-mail, it is rising a currently bi9 issue to waste their time and cost due to the spam mail in an individual or enterprise. Many kinds of solutions have been studied to solve harmful effects of spam mail. Such typical methods are as follows; pattern matching using the keyword with representative method and method using the probability like Naive Bayesian. In this paper, we propose a classification method of spam mails from normal mails using Support Vector Machine, which has excellent performance in pattern classification problems, to compensate for the problems of existing research. Especially, the proposed method practices efficiently a teaming procedure with a word dictionary including a generated index by the n-Gram. In the conclusion, we verified the proposed method through the accuracy comparison of spm mail separation between an existing research and proposed scheme.

• The Journal of the Acoustical Society of Korea
• /
• v.26 no.8
• /
• pp.397-402
• /
• 2007

In this paper, we apply a support vector machine(SVM) that incorporates an optimized nonlinear decision rule over different sets of feature vectors to improve the performance of statistical model-based voice activity detection(VAD). Conventional method performs VAD through setting up statistical models for each case of speech absence and presence assumption and comparing the geometric mean of the likelihood ratio (LR) for the individual frequency band extracted from input signal with the given threshold. We propose a novel VAD technique based on SVM by treating the LRs computed in each frequency bin as the elements of feature vector to minimize classification error probability instead of the conventional decision rule using geometric mean. As a result of experiments, the performance of SVM-based VAD using the proposed feature has shown better results compared with those of reported VADs in various noise environments.

• Journal of the Korean Statistical Society
• /
• v.29 no.1
• /
• pp.95-108
• /
• 2000

Consider the oblique factor model X=Af＋$\varepsilon$, with defining relation $\Sigma$=Λ$\Phi$Λ'＋Ψ. This paper is concerned with suggesting an optimal Bayes criterion for determining the number of factors in the model, i.e. dimension of the vector f. The use of marginal likelihood as a method for calculating posterior probability of each model with given dimension is developed under a generalized conjugate prior. Then based on an appropriate loss function, a Bayes rule is developed by use of the posterior probabilities. It is shown that the approach is straightforward to specify distributionally and to imploement computationally, with output readily adopted for constructing required cirterion.

• 제어로봇시스템학회:학술대회논문집
• /
• 2002.10a
• /
• pp.32.3-32
• /
• 2002

We analyze the effect of main inertial sensor errors such as, misalignment, scale factor error and bias on the performance of modified sequential probability ratio test (SPRT) for sequential fault detection and isolation (FDI). The inertial sensor errors cause the modified SPRT method to give false alarm. We use a two-stage KF to obtain a modified parity vector with which the inertial sensor errors can be removed and thus modified SPRT method can be used regardless of inertial sensor errors.

• Journal of the Korea Society of Computer and Information
• /
• v.15 no.12
• /
• pp.75-84
• /
• 2010

As broadcasting in Mobile Ad hoc NETwork (MANET) is the process that a node sends a packet to all other nodes in the network. it is used for routing protocols such as Ad hoc On demand Distance Vector (AODV) to disseminate control information for establishing the routes. In this paper, we propose Probabilistic Broadcasting mechanism based on Distance Ratio between sender and receive node in MANETs. The proposed approach is based on the combination of probability and distance based approach. A mobile node receiving broadcast packets determines the probability of rebroadcasting considering distance ratio from sender. The distance ratio of a node is calculated by the distance from sender and the length of radio field strength. As a node with high distance ratio is located far away from sender, rebroadcast probability is set to high value. On contrary, the low rebroadcast probability is set for a node with low distance ratio which is close to sender. So it reduces packets transmission caused by the early die-out of rebroadcast packets. Compared with the simple flooding and fixed probabilistic flooding by simulation, our approach shows better performances results. Proposed algorithm can reduce the rebroadcast packet delivery more than 30% without scanting reachability, where as it shows up to 96% reachability compared with flooding.

• Journal of Institute of Control, Robotics and Systems
• /
• v.10 no.12
• /
• pp.1295-1304
• /
• 2004

In this paper, the Bayesian recurrent neural network is proposed to predict time series data. A neural network predictor requests proper learning strategy to adjust the network weights, and one needs to prepare for non-linear and non-stationary evolution of network weights. The Bayesian neural network in this paper estimates not the single set of weights but the probability distributions of weights. In other words, the weights vector is set as a state vector of state space method, and its probability distributions are estimated in accordance with the particle filtering process. This approach makes it possible to obtain more exact estimation of the weights. In the aspect of network architecture, it is known that the recurrent feedback structure is superior to the feedforward structure for the problem of time series prediction. Therefore, the recurrent neural network with Bayesian inference, what we call Bayesian recurrent neural network (BRNN), is expected to show higher performance than the normal neural network. To verify the proposed method, the time series data are numerically generated and various kinds of neural network predictor are applied on it in order to be compared. As a result, feedback structure and Bayesian learning are better than feedforward structure and backpropagation learning, respectively. Consequently, it is verified that the Bayesian reccurent neural network shows better a prediction result than the common Bayesian neural network.

• Bulletin of the Korean Chemical Society
• /
• v.33 no.9
• /
• pp.2873-2877
• /
• 2012

Quasi-classical trajectory (QCT) calculations of H+HLi reaction have been carried out on a new potential energy surface of the ground state reported by Prudente et al. [Chem. Phys. Lett. 2009, 474, 18]. The four polarization-dependent differential cross sections have been carried out in the center of mass (CM) frame at various collision energies. The reaction probability for the depletion channel has been studied over a wide collision energy range. It has been found that the collision energy decreases remarkably reaction probability, which shows the expected behavior of the title reaction belonging to an exothermic barrierless reaction. The results are in good agreement with previous RMP results. The P(${\theta}_r$), P(${\phi}_r$) and P(${\theta}_r,\;{\phi}_r$) distributions, the k-k'-j' correlation and the angular distribution of product rotational vectors are presented in the form of polar plots. The average rotational alignment factor <$P_2(j{\prime}{\cdot}k)$> as a function of collision energy is also calculated. The results indicate that the collision energy has a great influence on the polarization of the product rotational angular momentum vector j'.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.44 no.3
• /
• pp.61-69
• /
• 2007

This paper proposes new preprocessing algorithm to extract minutiae in the process of fingerprint recognition. Fingerprint images quality enhancement is a topic phase to ensure good performance in a topic phase to ensure good performance in a Automatic Fingerprint Identification System(AFIS) based on minutiae matching. This paper proposes an algorithm to improve fingerprint image preprocessing to extract minutiae accurately based on directional filter. We improved the suitability of low quality fingerprint images to better suit fingerprint recognition by using valid ridge vector and ridge probability of fingerprint images. With the proposed fingerprint improvement algorithm, noise is removed and presumed ridges are more clearly ascertained. The algorithm is based on five step: computation of effective ridge vector, computation of ridge probability, noise reduction, ridge emphasis, and orientation compensation and frequency estimation. The performance of the proposed approach has been evaluated on two set of images: the first one is self collected using a capacitive semiconductor sensor and second one is DB3 database from Fingerprint Verification Competition (FVC).

• The Journal of the Acoustical Society of Korea
• /
• v.15 no.6
• /
• pp.5-11
• /
• 1996

This paper is the study for recognizing single vowels of speaker-independent, and we suppose a method of speech recognition using VQ(Vector Quantization)/NN(Neural Network). This method makes a VQ codebook, which is used for obtaining the observation sequence, and then claculates the probability value by comparing each codeword with the data, finally uses these probability values for the input value of the neural network. Korean signle vowels are selected for our recognition experiment, and ten male speakers pronounced eight single vowels ten times. We compare the performance of our method with those of fuzzy VQ/HMM and conventional VQ/NN According to the experiment result, the recognition rate by VQ/NN is 92.3%, by VQ/HMM using fuzzy is 93.8% and by VQ/NN using fuzzy is 95.7%. Therefore, it is shown that recognition rate of speech recognition by fuzzy VQ/NN is better than those of fuzzy VQ/HMM and conventional VQ/HMM because of its excellent learning ability.