• 말소리와 음성과학
• /
• 제1권1호
• /
• pp.69-73
• /
• 2009

This paper proposes an efficient global covariance-based principal component analysis (GCPCA) for speaker identification. Principal component analysis (PCA) is a feature extraction method which reduces the dimension of the feature vectors and the correlation among the feature vectors by projecting the original feature space into a small subspace through a transformation. However, it requires a larger amount of training data when performing PCA to find the eigenvalue and eigenvector matrix using the full covariance matrix by each speaker. The proposed method first calculates the global covariance matrix using training data of all speakers. It then finds the eigenvalue matrix and the corresponding eigenvector matrix from the global covariance matrix. Compared to conventional PCA and Gaussian mixture model (GMM) methods, the proposed method shows better performance while requiring less storage space and complexity in speaker identification.

• 한국정보과학회:학술대회논문집
• /
• 한국정보과학회 2001년도 봄 학술발표논문집 Vol.28 No.1 (B)
• /
• pp.529-531
• /
• 2001

We present an expectation-maximization algorithm for principal component analysis via orthogonalization. The algorithm finds actual principal components, whereas previously proposed EM algorithms can only find principal subspace. New algorithm is simple and more efficient thant probabilistic PCA specially in noiseless cases. Conventional PCA needs computation of inverse of the covariance matrices, which makes the algorithm prohibitively expensive when the dimensions of data space is large. This EM algorithm is very powerful for high dimensional data when only a few principal components are needed.

• Journal of the Korean Data and Information Science Society
• /
• 제20권2호
• /
• pp.321-328
• /
• 2009

회귀분석에서 설명변수들 사이에 상관이 높으면 최소제곱추정법에서 구한 회귀계수들의 정도가 떨어진다. 다중공선성이라 불리는 이 현상은 실제 자료분석에서 심각한 문제를 야기시킨다. 이 다중공선성의 문제를 극복하기 위한 여러 가지 방법이 제안되었다. 능형회귀, 축소추정량 그리고 주성분분석에 기초한 주성분회귀와 고유값회귀등이 있다. 지난 수십 년간 많은 통계학자들은 일반적인 중 회귀에서 감도분석에 관해 연구하였으며, 주성분회귀, 고유값회귀와 로지스틱 주성분회귀에 대해서도 같은 주제로 연구하였다. 이 모든 방법에서 주성분분석은 중요한 역할을 하였다. 또한, 많은 통계학자들이 주성분분석과 관련된 다변량 방법에서 감도분석에 대해 연구를 하였다. 본 연구논문에서는 주성분회귀와 고유값회귀를 소개하고, 또한 주성분회귀와 고유값회귀에서 감도분석의 방법을 소개하고, 마지막으로 이들두방법에 대한 감도분석의 성질에 대해 논의하였다.

• 한국정밀공학회:학술대회논문집
• /
• 한국정밀공학회 2010년도 춘계학술대회 논문집
• /
• pp.1035-1036
• /
• 2010

• Genomics & Informatics
• /
• 제18권1호
• /
• pp.11.1-11.3
• /
• 2020

In genome-wide association studies, pathway-based analysis has been widely performed to enhance interpretation of single-nucleotide polymorphism association results. We proposed a novel method of hierarchical structural component model (HisCoM) for pathway analysis of common variants (HisCoM for pathway analysis of common variants [HisCoM-PCA]) which was used to identify pathways associated with traits. HisCoM-PCA is based on principal component analysis (PCA) for dimensional reduction of single nucleotide polymorphisms in each gene, and the HisCoM for pathway analysis. In this study, we developed a HisCoM-PCA software for the hierarchical pathway analysis of common variants. HisCoM-PCA software has several features. Various principle component scores selection criteria in PCA step can be specified by users who want to summarize common variants at each gene-level by different threshold values. In addition, multiple public pathway databases and customized pathway information can be used to perform pathway analysis. We expect that HisCoM-PCA software will be useful for users to perform powerful pathway analysis.

• Journal of KIEE
• /
• 제11권1호
• /
• pp.1-7
• /
• 2001

Various methods for blind source separation (BSS) are based on independent component analysis (ICA) which can be viewed as a nonlinear extension of principal component analysis (PCA). Most existing ICA methods require certain nonlinear functions (which leads to higher-order statistics) depending on the probability distributions of sources, whereas PCA is a linear learning method based on second-order statistics. In this paper we show that the PCA can be applied to the task of BBS, provided that source are spatially uncorrelated but temporally correlated. Since the resulting method is based on only second-order statistics, it avoids the nonlinear function and is able to separate mixtures of several colored Gaussian sources, in contrast to the conventional ICA methods.

• Food Science and Biotechnology
• /
• 제14권4호
• /
• pp.554-557
• /
• 2005

As a device for taste sensation, an 8-channel taste evaluation system was prepared and applied for discriminant analysis of marketed liquor. The biomimetic polymer membranes for the system were prepared through a casting procedure by employing polyvinyl chloride, bis (2-ethylhexyl)sebacate as plasticizer and electroactive materials such as valinomycin in the ratio of 33:66:1, and were separately attached over the sensitive area of ion-selective electrodes to construct the corresponding taste sensor array. The sensor array in conjunction with a double junction reference electrode was connected to a high-input impedance amplifier and the amplified sensor signals were interfaced to a personal computer via an A/D converter. When the signal data from the sensor array for 3 groups of marketed liquor like Maesilju, Soju and beer were analyzed by principal component analysis after normalization, it was observed that the 1st, 2nd and 3rd principal component were responsible for most of the total data variance, and the analyzed liquor samples were discriminated well in 2 dimensional principal component planes composed of the 1st-2nd and the 1st-3rd principal component.

• Communications for Statistical Applications and Methods
• /
• 제27권1호
• /
• pp.149-161
• /
• 2020

In this paper we extend the functional principal component analysis for real-valued random functions to the case of Hilbert-space-valued functional random objects. For this, we introduce an autocovariance operator acting on the space of real-valued functions. We establish an eigendecomposition of the autocovariance operator and a Karuhnen-Loève expansion. We propose the estimators of the eigenfunctions and the functional principal component scores, and investigate the rates of convergence of the estimators to their targets. We detail the implementation of the methodology for the cases of compositional vectors and density functions, and illustrate the method by analyzing time-varying population composition data. We also discuss an extension of the methodology to multivariate cases and develop the corresponding theory.

• 한국진공학회:학술대회논문집
• /
• 한국진공학회 2013년도 제44회 동계 정기학술대회 초록집
• /
• pp.320-320
• /
• 2013

Film thickness monitoring with plasma impedance monitoring (PIM) is demonstrated for small area $SiO_2$ RF plasma etching processes in this work. The chamber conditions were monitored by the impedance signal variation from the I-V monitoring system. Moreover, modified principal component analysis (mPCA) was applied to estimate the $SiO_2$ film thickness. For verification, the PIM was compared with optical emission spectroscopy (OES) signals which are widely used in the semiconductor industry. The results indicated that film thickness can be estimated by 1st principal component (PC) and 2nd PC. Film thickness monitoring of small area $SiO_2$ etching was successfully demonstrated with RF plasma harmonic impedance monitoring and mPCA. We believe that this technique can be potentially applied to plasma etching processes as a sensitive process monitoring tool.

• 로봇학회논문지
• /
• 제19권3호
• /
• pp.254-265
• /
• 2024

Pattern recognition for surface electromyogram (sEMG) suffers from its nonstationary and stochastic property. Although it can be relieved by acquiring new training data, it is not only time-consuming and burdensome process but also hard to set the standard when the data acquisition should be held. Therefore, we propose an adaptive sEMG pattern recognition algorithm using principal component analysis. The proposed algorithm finds the relationship between sEMG channels and extracts the optimal principal component. Based on the relative distance, the proposed algorithm determines whether to update the existing patterns or to register the new pattern. From the experimental result, it is shown that multiple patterns are generated from the sEMG data stream and they are highly related to the motion. Furthermore, the proposed algorithm has shown higher classification accuracy than k-nearest neighbor (k-NN) and support vector machine (SVM). We expect that the proposed algorithm is utilized for adaptive and long-lasting pattern recognition.