• ETRI Journal
• /
• v.24 no.6
• /
• pp.469-472
• /
• 2002

Speaker adaptation techniques are generally used to reduce speaker differences in speech recognition. In this work, we focus on the features fitted to a linear regression-based speaker adaptation. These are obtained by feature transformation based on independent component analysis (ICA), and the feature transformation matrices are estimated from the training data and adaptation data. Since the adaptation data is not sufficient to reliably estimate the ICA-based feature transformation matrix, it is necessary to adjust the ICA-based feature transformation matrix estimated from a new speaker utterance. To cope with this problem, we propose a smoothing method through a linear interpolation between the speaker-independent (SI) feature transformation matrix and the speaker-dependent (SD) feature transformation matrix. From our experiments, we observed that the proposed method is more effective in the mismatched case. In the mismatched case, the adaptation performance is improved because the smoothed feature transformation matrix makes speaker adaptation using noisy speech more robust.

• Journal of the Korean Statistical Society
• /
• v.22 no.2
• /
• pp.339-345
• /
• 1993

It is well known that distribution of functions of eigen values and vectors of a certain matrix plays an important role in multivariate analysis. This paper deals with the transformation of a correlation-type random matrix to its eigen values and vectors. Properties of the transformation are also considered. The results obtained are applied to express the joint distribution of eigen values and vectors of the correlation matrix when sample is taken from a m-variate spherical distribution.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.54 no.12
• /
• pp.723-731
• /
• 2005

The $4{\times}4$ homogeneous transformation matrix is a compact representation of orientation and position of an object in robotics and computer graphics. A coordinate transformation is accomplished through the successive multiplications of homogeneous matrices, each of which represents the orientation and position of each corresponding link. Thus, for real time control applications in robotics or animation in computer graphics, the fast multiplication of homogeneous matrices is quite demanding. In this paper, a parallel-architecture vector processor is designed for this purpose. The processor has several key features. For the accuracy of computation for real application, the operands of the processors are floating point numbers based on the IEEE Standard 754. For the parallelism and reduction of hardware redundancy, the processor takes column vectors of homogeneous matrices as multiplication unit. To further improve the throughput, the processor structure and its control is based on a pipe-lined structure. Since the designed processor can be used as a special purpose coprocessor in robotics and computer graphics, additionally to special matrix/matrix or matrix/vector multiplication, several other useful instructions for various transformation algorithms are included for wide application of the new design. The suggested instruction set will serve as standard in future processor design for Robotics and Computer Graphics. The design is verified using FPGA implementation. Also a comparative performance improvement of the proposed design is studied compared to a uni-processor approach for possibilities of its real time application.

• MALSORI
• /
• no.43
• /
• pp.127-136
• /
• 2002

The speaker adaptation technique is generally used to reduce the speaker difference in speech recognition. In this work, we focus on the features fitted to a linear regression-based speaker adaptation. These are obtained by feature transformation based on independent component analysis (ICA), and the transformation matrix is learned from a speaker independent training data. When the amount of data is small, however, it is necessary to adjust the ICA-based transformation matrix estimated from a new speaker utterance. To cope with this problem, we propose a smoothing method: through a linear interpolation between the speaker-independent (SI) feature transformation matrix and the speaker-dependent (SD) feature transformation matrix. We observed that the proposed technique is effective to adaptation performance.

• Journal of Institute of Control, Robotics and Systems
• /
• v.15 no.2
• /
• pp.128-131
• /
• 2009

Recently it has been shown in the literature that a biquadratic transformation can be used to deal with the pathological cases arising in the bilinear transformation. This paper concerns the computational aspect of the biquadratic transformation. Two efficient algorithms are presented to compute the transformation matrix.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 1996.04a
• /
• pp.355-360
• /
• 1996

When the operational strains of a structure can not be directly measured in order to predict the life of the structure due to the problem of the attachment, those must be obtained indirectly. Since the displacement and the strain are interrelated, the strain can be predicted from the measured displacement and displacement-strain transformation matrix. The transformation matrix is dependent on the boundary condition, unfortunately, and it is also difficult to know exactly that of the operational system. In this study, for the structure with arbitrary boundary condition under the operation, the approximate method is proposed in order to predict the operational strains using the transformation matrix obtained by using free boundary conditions. And the method is applied to predict the strains of leads of surface mount component under the vibration of the printed circuit board.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2002.11a
• /
• pp.226-230
• /
• 2002

This paper proposes a method of improvement of accuracy of projective transformation matrix for Image Mosaicing. Using shift theorem, we extracted global translation components between images and using translation components, we found matching points between images so we solve general matching point problem we extracted highly trusted matching point using RANSAC algorithm. we normalized matching point coordinates and improved accuracy of projective transformation matrix.

• Korean Journal of Plant Tissue Culture
• /
• v.24 no.4
• /
• pp.225-231
• /
• 1997

For genetic engineering to be commercially viable, an efficient transformation system is needed to produce transgenic plane from diverse genotypes ("generalized protocol"). Development of such a system requires optimization of a number of components such as gene transfer agent, plant tissues competent for both regeneration and transformation, and control of transgene expression. Although several novel gene transfer methods have been developed for plane, a majority of stably transformed plane express the introduced genes at low levels. Moreover, silencing of selectable marker genes shortly after their incorporation into plant chromosomes may result in low recovery of transgenic tissues from selection. Matrix attachment regions (MARs) are DNA sequences that bind to the cell's proteinaceous nuclear matrix to form DNA loop domains. MARs have been shown to increase transgene expression in tobacco cells, and reduce position in mature transgenic plants. Flanking an antibiotic resistance transgene with MARs should therefore lead to improved rates of transformation in a diversity of species, and may permit recalcitrant species and genotypes to be successfully transformed. Literature review and recent data from my laboratory suggest that MARs can serve as a transformation booster in recalcitrant plant species.

• International Journal of Control, Automation, and Systems
• /
• v.5 no.6
• /
• pp.652-662
• /
• 2007

This paper presents a new and simple method for sensorless operation of matrix converter drives using a constant air-gap flux and the imaginary power flowing to the motor. To improve low-speed sensorless performance, the non-linearities of a matrix converter drive such as commutation delays, turn-on and turn-off times of switching devices, and on-state switching device voltage drop are modeled using PQR transformation and compensated using a reference current control scheme. The proposed compensation method is applied for high performance induction motor drives using a 3 kW matrix converter system. Experimental results are shown to illustrate the feasibility of the proposed strategy.

• Journal of Korean Institute of Industrial Engineers
• /
• v.23 no.4
• /
• pp.741-754
• /
• 1997

Code matrix is the matrix of which on element and its neighbors are arranged to have a code value. The code matrix was originally designed by the author for developing a vision system but has not been theoretically studied. In this paper some theoretical properties of the code matrix are investigated. The studied characteristics of the code matrix are useful for not only understanding the matrix itself but efficiently restructuring the matrix. A number of transformation functions, which enable the matrix to have different shape, are thus developed based on the investigated properties. The transformation functions are then applied to build a perfectly and uniformly structured square code matrix, which is proven useful in on image processing example. The study in this paper is expected to serve a theoretical background for the application of the code matrix in many areas.