• 음성과학
• /
• 제13권4호
• /
• pp.177-186
• /
• 2006

In an effort to enhance the quality of feature vector classification and thereby reduce the recognition error rate of the speaker-independent speech recognition, we employ the Mahalanobis distance in the calculation of the similarity measure between feature vectors. It is assumed that the metric matrix of the Mahalanobis distance be diagonal for the sake of cost reduction in memory and time of calculation. We propose that the diagonal elements be given in terms of the variations of the feature vector components. Geometrically, this prescription tends to redistribute the set of data in the shape of a hypersphere in the feature vector space. The idea is applied to the speech recognition by hidden Markov model with fuzzy vector quantization. The result shows that the recognition is improved by an appropriate choice of the relevant adjustable parameter. The Viterbi score difference of the two winners in the recognition test shows that the general behavior is in accord with that of the recognition error rate.

• ETRI Journal
• /
• 제39권6호
• /
• pp.841-850
• /
• 2017

We investigate neural network image reconstruction for magnetic particle imaging. The network performance strongly depends on the convolution effects of the spectrum input data. The larger convolution effect appearing at a relatively smaller nanoparticle size obstructs the network training. The trained single-layer network reveals the weighting matrix consisting of a basis vector in the form of Chebyshev polynomials of the second kind. The weighting matrix corresponds to an inverse system matrix, where an incoherency of basis vectors due to low convolution effects, as well as a nonlinear activation function, plays a key role in retrieving the matrix elements. Test images are well reconstructed through trained networks having an inverse kernel matrix. We also confirm that a multi-layer network with one hidden layer improves the performance. Based on the results, a neural network architecture overcoming the low incoherence of the inverse kernel through the classification property is expected to become a better tool for image reconstruction.

• Journal of Power Electronics
• /
• 제18권2호
• /
• pp.522-532
• /
• 2018

The cascaded H Bridge (CHB) multilevel inverter (MLI) is popular among the classical MLI topologies due to its modularity and reliability. Although space vector modulation (SVM) is the most suitable modulation scheme for MLIs, it has not been used widely in industry due to the higher complexity involved in its implementation. In this paper, a simple and novel generalized SVM algorithm is proposed, which has both reduced time and space complexity. The proposed SVM involves the generalization of both the duty cycle calculation and switching sequence generation for any n-level inverter. In order to generate the gate pulses for an inverter, a generalized switch matrix (SM) for the CHB inverter is also introduced, which further simplifies the algorithm. The algorithm is tested and verified for three-phase, three-level and five-level CHB inverters in simulations and hardware implementation. A comparison of the proposed method with existing SVM schemes shows the superiority of the proposed scheme.

• 한국항공우주학회지
• /
• 제34권1호
• /
• pp.81-88
• /
• 2006

본 논문에서는 RLG를 사용하는 관성항법시스템에서 자이로 출력의 적분 과정에서 유발되는 비교환 오차를 정의하고 이에 대한 해석을 수행한다. 이를 위하여 RLG를 사용하는 관성항법시스템에 나타나는 진동성 운동, 원추운동, AV 마운트에 의하여 유발되는 ISA 운동, 항체의 실제 운동 등을 살펴보고, 각각의 운동에 의하여 유발되는 비교환 오차를 해석한다. 비교환 오차 해석은 회전벡터와 자이로 출력 사이의 관계식과 좌표변환행렬과 각속도 벡터 사이의 관계식을 이용하여 유도된다.

• 전력전자학회:학술대회논문집
• /
• 전력전자학회 2005년도 전력전자학술대회 논문집
• /
• pp.638-642
• /
• 2005

This paper proposes a modified space-vector pulse width modulation (PWM) strategy which can restrict the common-mode voltage for three-phase to three-phase matrix converter and still keep sinusoidal input and output waveforms and unity power factor at the input side. The proposed control method has been developed based on contributing the appropriate space vectors instead of using zero space vectors. The advantages of this proposed method is to reduce the peak value of common-mode voltage to 42% beside the lower high harmonic components as compared to the conventional SVM method. Hence, the new table is also presented with the new space vector rearrangement. Furthermore, the voltage transfer ratio is unaffected by the proposed method. A simulation of the overall system has been carried out to validate the advantages of the proposed method.

• 한국정보과학회:학술대회논문집
• /
• 한국정보과학회 2005년도 한국컴퓨터종합학술대회 논문집 Vol.32 No.1 (B)
• /
• pp.715-717
• /
• 2005

Support vector machine(SVM)은 최근 각광받는 기계학습 방법 중 하나로서, kernel function 이라는 사상(mapping)을 이용하여 입력 공간의 벡터를 classification이 용이한 특징 (feature) 공간의 벡터로 변환하는 것을 근간으로 한다. SVM은 이러한 특징 공간에서 두 클래스를 구분 짓는 hyperplane을 일련의 최적화 방법론을 사용하여 찾아내며, 주어진 문제가 convex problem 인 경우 항상 global optimal solution 을 보장하는 등의 장점을 지닌다. 한편 bioinformatics 연구에서 주로 사용되는 데이터는 측정 오류 등 일련의 오류를 포함하고 있으며, 이러한 오류는 기계학습 방법론이 어떤 decision boundary를 찾아내는가에 영향을 끼치게 된다. 특히 SVM의 경우 이러한 오류는 특징 공간 벡터간의 관계를 나타내는 Gram matrix를 변화로 나타나게 된다. 본 연구에서는 입력 공간에 오류가 발생할 때 그것이 SVM 의 decision boundary를 어떻게 변화시키는가를 대표적인 두 가지 kernel function, 즉 linear kernel과 Gaussian kernel에 대해 분석하였다. Wisconsin대학의 유방암(breast cancer) 데이터에 대해 실험한 결과, 데이터의 오류에 따른 SVM 의 classification 성능 변화 양상을 관찰하여 커널의 종류에 따라 SVM이 어떠한 특성을 보이는가를 밝혀낼 수 있었다. 또 흥미롭게도 어떤 조건 하에서는 오류가 크더라도 오히려 SVM 의 성능이 향상되는 것을 발견했는데, 이것은 바꾸어 생각하면 Gram matrix 의 일부를 변경하여 SVM 의 성능 향상을 꾀할 수 있음을 나타낸다.

• 소음진동
• /
• 제4권1호
• /
• pp.71-82
• /
• 1994

We propose a systematic method to select the master states, which are retained in the reduced model after the order reduction process. The proposed method is based on the fact that the range space of right eigenvector matrix is spanned by orthogonal base vectors, and tries to keep the orthogonality of the submatrix of the base vector matrix as much as possible during the reduction process. To quentify the skewness of that submatrix, we define "Absolute Singularity Factor(ASF)" based on its singular values. While the degree of observability is concerned with estimation error of state vector and up to n'th order derivatives, ASF is related only to the minimum state estimation error. We can use ASF to evaluate the estimation performance of specific partial measurements compared with the best case in which all the state variables are identified based on the full measurements. A heuristic procedure to find suboptimal master states with reduced computational burden is also proposed. proposed.

• 한국통신학회논문지
• /
• 제34권7C호
• /
• pp.635-639
• /
• 2009

본 논문에서는 행벡터의 집합이 이진 벡터합 연산에 관해 닫혀있는 모든 하다마드 (Hadmard) 행렬들은 서로 동치(equivalent) 임융 증명한다. 이를 이용하면, 최대길이 수열로부터 생성된 순회 (cyclic) 하다마드 행렬과 크로네커 (Kronecker) 곱에 의해 생성된 월쉬-하다마드 (Walsh-Hadamard) 행렬이 동치임을 간단히 보일 수 있다.

• Journal of applied mathematics & informatics
• /
• 제28권1_2호
• /
• pp.351-361
• /
• 2010

The solution of large sparse linear systems is one of the most important problems in large scale scientific computing. Among the many methods developed, the preconditioned Krylov subspace methods are considered the preferred methods. Selecting a suitable preconditioner with appropriate parameters for a specific sparse linear system presents a challenging task for many application scientists and engineers who have little knowledge of preconditioned iterative methods. The prediction of ILU type preconditioners was considered in [27] where support vector machine(SVM), as a data mining technique, is used to classify large sparse linear systems and predict best preconditioners. In this paper, we apply the data mining approach to the sparse approximate inverse(SAI) type preconditioners to find some parameters with which the preconditioned Krylov subspace method on the linear systems shows best performance.

• Journal of Astronomy and Space Sciences
• /
• 제25권3호
• /
• pp.255-266
• /
• 2008

The current paper presents application of a new analytic solution in general relative motion to spacecraft formation flying in an elliptic orbit. The calculus of variations is used to analytically find optimal trajectories and controls for the given problem. The inverse of the fundamental matrix associated with the dynamic equations is not required for the solution in the current study. It is verified that the optimal thrust vector is a function of the fundamental matrix of the given state equations. The cost function and the state vector during the reconfiguration can be analytically obtained as well. The results predict the form of optimal solutions in advance without having to solve the problem. Numerical simulation shows the brevity and the accuracy of the general analytic solutions developed in the current paper.