• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2015.05a
• /
• pp.47-50
• /
• 2015

Speech Signals consist of signals of consonants and vowels, but the lasting time of vowels is much longer than that of consonants. It can be assumed that the correlations between signal blocks in speech signal is very high. Each speech signal is divided into blocks which have 128 speech data. FFT is applied to each block. Low frequency areas of the results of FFT is taken and Covariance matrix between blocks in a speech signal is extracted and finally eigenvalues of those matrix are obtained. It is studied that what the distribution of eigenvalues of various speech files is. The differences between speech signals and noise signals from cars are also studied.

• Journal of IKEEE
• /
• v.27 no.1
• /
• pp.38-44
• /
• 2023

Performance of a STAP(space-time adaptive processing) algorithm highly depends on how closely the estimated covariance matrix(CM) resembles the actual CM by the interference in CUT(cell under test). A STAP has 2 dimensional data structure determined by the number of array elements and the number of transmitting pulses and both numbers are generally not small. Thus, to meet the degree of freedom(DOF) of the CM, a huge amount of training data is required. This paper presents an algorithm to generate virtual training data from small received data, via converting them into the data in spatial frequency-Doppler plane. We theoretically derive where the clutter exist in the plane and present the procedure to implement the proposed algorithm. Finally, with the simulated scenario of small received data, we show the proposed algorithm can improve STAP performance.

• The Journal of the Acoustical Society of Korea
• /
• v.20 no.7
• /
• pp.94-100
• /
• 2001

When using a sample covariance matrix data in paucity of snapshots, adaptive matched field processing will have problem in inverting covariance matrix due to the rank deficiency. The general solutions are diagonal loading and eigenanalysis methods, but there is a significant bias in the power output. This paper presents a quantitative study of bias of power output and the performance of source localization through the simulation and the measured data analysis in fixed source case using the diagonal loading method for the minimum variance processor. Results show that the bias in power output is reduced and the performance of source localization is improved when the number of snapshots is greater than the number of array sensors.

• The Korean Journal of Applied Statistics
• /
• v.4 no.1
• /
• pp.1-11
• /
• 1991

It is of great interest to consider the homogeniety of covariance matrices in MANOVA of discriminant analysis. If we lock at the problem of testing hypothesis, H : $\Sigma_1 = \Sigma_2$ from an invariance point of view where $\Sigma_i$ are the covariance matrix of two independent p-variate distribution, the testing problem is invariant under the group of nonsingular transformations and the hypothesis becomes H : $\delta_1 = \delta_2 = \cdots = \delta_p = 1$ where $\delta = (\delta_1, \delta_2, \cdots, \delta_p)$ is a vector of latent roots of $\Sigma$. Bias-corrected estimators of eigenvalues and sampling distribution of the test statistics proposed are obtained. Pooled-bootstrap method also considered for Bartlett's modified likelihood ratio statistics.

• The Korean Journal of Applied Statistics
• /
• v.37 no.4
• /
• pp.455 -465
• /
• 2024

SMOTE (synthetic minority over-sampling technique) has been used the most as a solution to the problem of imbalanced data. SMOTE selects the nearest neighbor based on Euclidean distance. However, Euclidean distance has the disadvantage of not considering the correlation between variables. In particular, the Mahalanobis distance has the advantage of considering the covariance of variables. But if there are outliers, they usually influence calculating the Mahalanobis distance. To solve this problem, we use the Mahalanobis distance by estimating the covariance matrix using MCD (minimum covariance determinant). Then apply Mahalanobis distance based on MCD to SMOTE to create new data. Therefore, we showed that in most cases this method provided high performance indicators for classifying imbalanced data.

• The Korean Journal of Applied Statistics
• /
• v.28 no.2
• /
• pp.151-158
• /
• 2015

This paper discusses how to use projections for the analysis of data from balanced incomplete block designs. A model is suggested as a matrix form for the interblock analysis. A second set of treatment effects can be found by projections from the suggested interblock model. The variance and covariance matrix of two estimated vectors of treatment effects is derived. The uncorrelation of two estimated vectors can be verified from their covaraince structure. The fitting constants method is employed for the calculation of block sum of squares adjusted for treatment effects.

• Proceedings of the Korean Statistical Society Conference
• /
• 2005.11a
• /
• pp.147-152
• /
• 2005

본 논문에서는 k개의 포아송 확률변수가 서로 종속 되어 있는 다변량 포아송 분포를 따를 때, 주어진 분산-공분산 행렬 구조를 유지하는 다변량 포아송 확률난수 생성방법에 대해 다루었다. 특히, 확률난수를 생성하기 위해 선형방정식을 푸는 두 가지 수치해석 알고리즘을 제안하였으며, Park 등 (1996)의 다변량 베르누이 확률난수 생성에 활용된 알고리즘과의 연관성을 다루었다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.5 no.3
• /
• pp.459-464
• /
• 2001

We propose a transmit beamforming algerian for array antenna in FDD (Frequency Division Duplex) environments. The proposed method estimates the directions and spectra of the users, and constructs the spatial covariance matrix of the interferences at the downlink frequency. The weights are computed by that covariance matrix and desired user's direction vector Simulations are performed under Rayleigh fading environments. The proposed method don't need the data feedback, has the enhanced performance in BER (Bit Error Rate).

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.26 no.11B
• /
• pp.1510-1521
• /
• 2001

In this paper a subspace-based blind interference cancellation is proposed and its performance is analyzed. Then the blind adaptive implementation is devolped using the improved natural power method which is the signal subspace tracking algorithm. The theoretical analysis shows that when the exact covariance matrix is kown the performance of the proposed detector is the same as that of the decorrelating detector. And when the covariance matrix is estimated the asymptotic results are examined. The results of computer simulation demonstrate that the proposed detector outperforms the previous blind adaptive RLS MOE detector.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2011.01a
• /
• pp.315-316
• /
• 2011

일반적으로 영상에서 얼굴 영상을 검출하고 인식하는 알고리즘은 패턴 인식 연구에 있어서 인간과 컴퓨터의 상호작용의 연구라는 면에서 아주 중요한 문제로 연구되어 왔다. 본 논문에서는 고유얼굴을 이용하여 유클리디언 거리법과 퍼지기법의 인식률을 비교해보고자 한다. PCA(Principal Component Analysis) 방식은 우수한 인식 결과를 보장하는 얼굴인식 기법중의 하나이며, 얼굴 영상을 이용하여 공분산 행렬을 계산하고, 공분산 행렬을 통해 생성된 저차원의 벡터, 즉 고유얼굴(Eigenface)을 이용하여 가중치를 계산하고, 이 가중치를 기준으로 인식을 수행하는 기법이다. 이를 기반으로 하여, 본 논문에서는 전처리 과정, 고유얼굴 과정, 유클리디언 거리법 및 퍼지 소속도 함수 설계 과정, 신경망 학습과정, 인식과정으로 구성된 5단계의 얼굴 인식 알고리즘을 제안한다.