• Phonetics and Speech Sciences
• /
• v.7 no.3
• /
• pp.131-138
• /
• 2015

In this paper, we propose L1-norm regularization for state vector adaptation of subspace Gaussian mixture model (SGMM). When you design a speaker adaptation system with GMM-HMM acoustic model, MAP is the most typical technique to be considered. However, in MAP adaptation procedure, large number of parameters should be updated simultaneously. We can adopt sparse adaptation such as L1-norm regularization or sparse MAP to cope with that, but the performance of sparse adaptation is not good as MAP adaptation. However, SGMM does not suffer a lot from sparse adaptation as GMM-HMM because each Gaussian mean vector in SGMM is defined as a weighted sum of basis vectors, which is much robust to the fluctuation of parameters. Since there are only a few adaptation techniques appropriate for SGMM, our proposed method could be powerful especially when the number of adaptation data is limited. Experimental results show that error reduction rate of the proposed method is better than the result of MAP adaptation of SGMM, even with small adaptation data.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.36 no.5A
• /
• pp.485-494
• /
• 2011

In this paper, novel sub-space interference alignment algorithms are proposed to boost the capacity in multi-cell environment. In the case of conventional sub-space alignment, arbitrary reference vectors have been adopted as transmitting vectors at the transmitter side, and the inter-cell interference among users are eliminated by using orthogonal vectors of the chosen reference vectors at the receiver side. However, in this case, sum-rate varies using different reference vectors even though the channel values keep constant, and vice versa. Therefore, the relationship between reference vectors and channel values are analyzed in this paper, and novel interference alignment algorithms are proposed to increase multi-cell capacity. Reference vectors with similar magnitude are adopted in the proposed algorithm. Simulation results show that the proposed algorithms provide about 50 % higher sum-rate than conventional algorithm.

• Journal of KIISE:Computing Practices and Letters
• /
• v.10 no.3
• /
• pp.273-281
• /
• 2004

There is the temporal correlation of the video sequence between the motion vector of current block and the motion vector of previous block. In this paper, we propose the prediction search algorithm for block matching using the temporal correlation of the video sequence and the center-biased property of motion vectors. The proposed algorithm determines the location of a better starting point for the search of an exact motion vector using the point of the smallest SAD(sum of absolute difference) value by the predicted motion vector from the same block of the previous frame and the predictor candidate point on each search region. Simulation results show that PSNR(Peak-to-Signal Noise Ratio) values are improved up to the 1.06㏈ as depend on the video sequences and improved about 0.19∼0.46㏈ on an average except the full search(FS) algorithm.

• Journal of the Korea Computer Industry Society
• /
• v.9 no.4
• /
• pp.149-156
• /
• 2008

There is the temporal correlation of a animation sequence between the motion vector of current block and the motion vector of previous block. In this paper, we propose the scene change detection algorithm for block matching using the temporal correlation of the animation sequence and the center-biased property of motion vectors. The proposed algorithm determines the location of a better starting point for the search of an exact motion vector using the point of the smallest SAD(sum of absolute difference) value by the predicted motion vector from the same block of the previous frame and the predictor candidate point on each search region. Simulation results show that the proposed algorithm has better detection performance, such as recall rate, then the existing method. The algorithm has the advantage of speed, simplicity and accuracy. In addition, it requires less amount of storage.

• Bulletin of the Korean Mathematical Society
• /
• v.55 no.6
• /
• pp.1691-1701
• /
• 2018

New inclusion sets are obtained for the eigenvalues of real matrices for which the all 1's vector is an eigenvector, i.e., the constant row-sum real matrices. A number of examples are provided. This paper builds upon the work of the authors in [7]. The results of this paper are in terms of $Ger{\check{s}}gorin$ discs of the second type. An application of the main theorem to bounding the algebraic connectivity of connected simple graphs is obtained.

• Journal of Korean Institute of Industrial Engineers
• /
• v.26 no.3
• /
• pp.220-226
• /
• 2000

The problem (P) of optimizing a linear function $d^Tx$ over the set of efficient points for a multiple objective linear program (M) is difficult because the efficient set is nonconvex. There are some interesting properties between the objective linear vector d and the matrix of multiple objectives C and those properties lead us to establish criteria to solve (P) with a linear program. In this paper we investigate a system of the linear equations $C^T{\alpha}$ = d and construct two linearly independent positive vectors u, v such that ${\alpha}$ = u - v. From those vectors u, v, solving an weighted sum linear program for finding an efficient extreme point for the (M) is a way of getting an optimal solution of the problem (P). Therefore the theorems presented in this paper provided us an easy way of solving nonconvex program (P) with a weighted sum linear program.

• Communications for Statistical Applications and Methods
• /
• v.9 no.3
• /
• pp.765-774
• /
• 2002

Suppose one is given a vector X of a finite set of quantities $X_i$ which are independent Poisson random variables. A null hypothesis $H_0$ about E(X) is to be tested against an alternative hypothesis $H_1$. A quantity $\sum\limits_{i}w_ix_i$ is to be computed and used for the test. The optimal values of $W_i$ are calculated for three cases: (1) signal to noise ratio is used in the test, (2) normal approximations with unequal variances to the Poisson distributions are used in the test, and (3) the Poisson distribution itself is used. The above three cases are considered to the situations that are without background noise and with background noise. A comparison is made of the optimal values of $W_i$ in the three cases for both situations.

• Journal of the Korean Statistical Society
• /
• v.24 no.1
• /
• pp.149-160
• /
• 1995

We consider the problem of optimal adaptive estiamtion of the euclidean parameter vector $\theta$ of the univariate non-linerar autogressive time series model ${X_t}$ which is defined by the following system of stochastic difference equations ; $X_t = \sum^p_{i=1} \theta_i \cdot T_i(X_{t-1})+e_t, t=1, \cdots, n$, where $\theta$ is the unknown parameter vector which descrives the deterministic dynamics of the stochastic process ${X_t}$ and ${e_t}$ is the sequence of white noises with unknown density $f(\cdot)$. Under some general growth conditions on $T_i(\cdot)$ which guarantee ergodicity of the process, we construct a sequence of adaptive estimatros which is locally asymptotic minimax (LAM) efficient and also attains the least possible covariance matrix among all regular estimators for arbitrary symmetric density.

• Proceedings of the IEEK Conference
• /
• 2008.06a
• /
• pp.803-804
• /
• 2008

In this paper, we propose an edge dependent interpolation (EDI) method based on adaptive search range. The proposed EDI method uses the vector matching to determine the edge direction, and the vector matching process is terminated when the previous sum of absolute difference (SAD) is smaller than the next one. The adaptive search range method enables the EDI algorithm to estimate edge direction more accurately and to reduce the computational complexity. The experimental results show that the proposed method produces better performance than conventional algorithms.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.8
• /
• pp.1091-1098
• /
• 1995

This paper proposes an efficient noiseless encoding method of vector quantization(VQ) index using a dynamic index mapping. Using high interblock correlation, the proposed index mapper transforms an index into a new one with lower entropy. In order to achieve good performance with low computational complexity, we adopt 'the sum of differences in pixel values on the block boundaries' as the cost function for index mapping. Simulation results show that the proposed scheme reduces the average bitrate by 40 - 50 % in ordinary VQ system for image compression. In addition, it is shown that the proposed index mapping method can be also applied to mean-residual VQ system, which allows the reduction of bitrate for VQ index by 20 - 30 %(10 - 20 % reduction in total bitrate). Since the proposed scheme is one for noiseless encoding of VQ index, it provides the same quality of the reconstructed image as the conventional VQ system.