• 한국정보통신학회논문지
• /
• 제7권3호
• /
• pp.437-447
• /
• 2003

Circulant Matrix Factorization (CMF)는 covariance 행렬의 spectral factorization된 결과를 얻을 수 있다. 우리는 얻어진 결과를 가지고 일반적으로 잘 알려진 방법인 Schur algorithm을 이용하여 finite impulse response(FIR)와 infinite impulse response (IIR) lattice 필터를 설계하는 방법을 제안하였다. CMF는 기존에 많이 사용되는 root finding을 사용하지 않고 covariance polynomial로부터 minimum phase 특성을 가지는 polynomial을 얻는데 유용한 방법이다. 그리고 Schur algorithm은 toeplitz matrix를 빠르게 Cholesky factorization하기 위한 방법으로 이 방법을 이용하면 FIR/IIR lattice 필터의 계수를 쉽게 찾아낼 수 있다. 본 논문에서는 이러한 방법들을 이용하여 FIR과 IIR lattice 필터의 설계의 계산적인 예제를 제시했으며, 제안된 방법과 다른 기존에 제시되었던 방법 (polynomial root finding과 cepstral deconvolution)들과 성능을 비교 평가하였다.

• International Journal of Control, Automation, and Systems
• /
• 제6권3호
• /
• pp.378-385
• /
• 2008

This paper aims to present a polynomial approach to the steady-state optimal filtering for delayed systems. The design of the steady-state filter involves solving one polynomial equation and one spectral factorization. The key problem in this paper is the derivation of spectral factorization for systems with delayed measurement, which is more difficult than the standard systems without delays. To get the spectral factorization, we apply the reorganized innovation approach. The calculation of spectral factorization comes down to two Riccati equations with the same dimension as the original systems.

• Kyungpook Mathematical Journal
• /
• 제52권4호
• /
• pp.359-373
• /
• 2012

The block LU factorization is used to solve the coupled Stokes equations arisen from an optimal control problem subject to Stokes equations. The convergence of the spectral element solution is proved. Some numerical evidences are provided for the model coupled Stokes equations. Moreover, as an application, this algorithm is performed for an optimal control problem.

• 대한전자공학회논문지TC
• /
• 제41권1호
• /
• pp.35-44
• /
• 2004

Circulant Matrix Factorization (CMF)는 covariance 행렬의 spectral factorization된 결과를 얻을 수 있다. 우리는 얻어진 결과를 가지고 일반적으로 잘 알려진 방법인 Schur algorithm을 이용하여 finite impulse response (FIR)차 infinite impulse response (IIR) lattice 필터를 설계하는 방법을 제안하였다. CMF는 기존에 많이 사용되는 root finding을 사용하지 않고 covariance Polynomial로부터 minimum phase 특성을 가지는 polynomial을 얻는데 유용한 방법이다. 그리고 Schur algorithm은 toeplitz matrix를 빠르게 Cholesky factorization하기 위한 방법으로 이 방법을 이용하면 FIR/IIR lattice 필터의 계수를 쉽게 찾아낼 수 있다. 본 논문에서는 이러한 방법들을 이용하여 FIR과 IIR lattice 필터의 설계의 계산적인 예제를 제시했으며, 제안된 방법과 다른 기존에 제시되었던 방법 (polynomial root finding과 cepstral deconvolution)들과 성능을 비교 평가하였다.

• 대한수학회보
• /
• 제39권3호
• /
• pp.495-509
• /
• 2002

We propose variants of the modified incomplete Cho1esky factorization preconditioner for a symmetric positive definite (SPD) matrix. Spectral properties of these preconditioners are discussed, and then numerical results of the preconditioned CG (PCG) method using these preconditioners are provided to see the effectiveness of the preconditioners.

• International Journal of Control, Automation, and Systems
• /
• 제1권2호
• /
• pp.206-209
• /
• 2003

An $H_2$/$H_{\infty}$ -controller realization is carried out by considering an entropy integral. Using J-spectral factorization, the parametrizations of all $H_{\infty}$ stabilizing controllers are derived. By the relation of a mixed $H_2$/$H_{\infty}$ control problem and a minimum entropy/$H_{\infty}$ control problem, the mixed $H_2$/$H_{\infty}$-controller state-space realization is presented.

• ETRI Journal
• /
• 제39권1호
• /
• pp.21-29
• /
• 2017

This paper addresses the problem of unsupervised speech separation based on robust non-negative matrix factorization (RNMF) with ${\beta}$-divergence, when neither speech nor noise training data is available beforehand. We propose a robust version of non-negative matrix factorization, inspired by the recently developed sparse and low-rank decomposition, in which the data matrix is decomposed into the sum of a low-rank matrix and a sparse matrix. Efficient multiplicative update rules to minimize the ${\beta}$-divergence-based cost function are derived. A convolutional extension of the proposed algorithm is also proposed, which considers the time dependency of the non-negative noise bases. Experimental speech separation results show that the proposed convolutional RNMF successfully separates the repeating time-varying spectral structures from the magnitude spectrum of the mixture, and does so without any prior training.

• 한국정밀공학회지
• /
• 제18권11호
• /
• pp.58-66
• /
• 2001

GPC has been reported as a useful self-tuning control algorithm for systems with unknown time-delay and parameters. GPC is easy to understand and implement, and thus has won popularity among many practicing engineers. Despite its success, GPC does not guarantee is nominal stability. So, in this paper, GPC is rederived in frequency domain instead of in the time domain to guarantee its nominal stability. Derivation of GPC in frequency domain involves spectral factorization and Diophantine equation. Frequency domain GPC control law is stable because the zeros of characteristic polynomial are strictly Schur. Recursive least square algorithm is used to identify unknown parameters. To see the effectiveness of the proposed controller, the controller is simulated for a numerical problem that changes in dead-time, in order and in parameters.

• 한국언어정보학회:학술대회논문집
• /
• 한국언어정보학회 2007년도 정기학술대회
• /
• pp.430-439
• /
• 2007

In this paper, we use non-negative matrix factorization (NMF) to refine the document clustering results. NMF is a dimensional reduction method and effective for document clustering, because a term-document matrix is high-dimensional and sparse. The initial matrix of the NMF algorithm is regarded as a clustering result, therefore we can use NMF as a refinement method. First we perform min-max cut (Mcut), which is a powerful spectral clustering method, and then refine the result via NMF. Finally we should obtain an accurate clustering result. However, NMF often fails to improve the given clustering result. To overcome this problem, we use the Mcut object function to stop the iteration of NMF.

• 한국정보과학회논문지:컴퓨팅의 실제 및 레터
• /
• 제14권5호
• /
• pp.497-501
• /
• 2008

본 논문에서는 연속적인 뇌파 분류를 위해 비음수 텐서 분해를 이용한 특징 추출과 비터비 알고리즘을 이용한 연속적인 데이타의 클래스 분류를 결합한 새로운 알고리즘을 제시한다. 비음수 텐서 분해는 이미 스펙트럼 데이타에 대해 뇌파의 주요한 특징을 잘 추출한다고 알려진 비음수 행렬 분해의 확장으로써 행렬이라는 제한된 틀에서 벗어나 데이타가 가지는 다양한 차원으로의 확대가 가능하다. 뇌-컴퓨터 인터페이스 컴피티션을 통해 공개된 데이터를 이용한 실험을 통해 제안된 방법의 유용함을 증명하도록 하겠다.