Kernel methods such as the support vector machines map input vectors into a high-dimensional feature space and linearly separate them there. The dimensionality of the feature space depends on a kernel function and is sometimes of an infinite dimension. The Gauss kernel is such an example. We discuss the effective dimension of the feature space with the Gauss kernel and show that it can be approximated to a sum of polynomial kernels and that its dimensionality is determined by the boundedness of the input space by considering the Taylor expansion of the kernel Gram matrix.

We propose a new method for Depth from Defocus (DFD) using wavelet transform. Most of the existing DFD methods use inverse filtering in a transform domain to determine the measure of defocus. These methods suffer from inaccuracies in finding the frequency domain representation due to windowing and border effects. The proposed method uses wavelets that allow performing both the local analysis and windowing with variable-sized regions for images with varying textural properties. Experimental results show that the proposed method gives more accurate depth maps than the previous methods.

In a content-based image retrieval (CBIR) system, both the retrieval relevance and the response time are very important. This letter presents the condensed two-stage search method as a new fast image retrieval approach by making use of the property of Cauchy-Schwarz inequality. The method successfully reduces the overall processing time for similarity computation, while maintaining the same retrieval relevance as the conventional exhaustive search method. By the extensive computer simulations, we observe that the condensed two-stage search method is more effective as the number of images and dimensions of the feature space increase.

This report describes face image recognition by 2-dimensional discrete Walsh transform and multi-layer neural networks. Neural network (NN) is one of the powerful tools for pattern recognition. In the previous researches of face image recognition by NN, the gray levels on each pixel of the face image have been used for input data to NN. However, because the face image has usually too many pixels, a variety of approaches have been required to reduce the number of the input data. In this research, 2-dimensional discrete Walsh transform is used for reduction of input data and the recognition is done by multi-layer neural networks. Finally, the validity of our method is varified.

Given the importance of the traffic on modern communication networks, advanced error correction methods are needed to overcome the changes expected in channel quality. Conventional countermeasures that use high dimensionality parity codes often fail to provide sufficient error correction capability. We propose a parity code with high dimensionality that is iteratively decoded. It provides better error correcting capability than conventional decoding methods. The proposal uses the steepest descent method to increase code bit reliability and the coherency between parities and code bits gradually. Furthermore, the quantization of the decoding algorithm is discussed. It is found that decoding with quantization can keep the error correcting capability high.

In light of the increasing number of computer applications that rely heavily on multimedia data, the database community has focused on the management and retrieval of multidimensional data. Nearest Neighbor queries (NN queries) have been widely used to perform content-based retrieval (e.g., similarity search) in multimedia applications. Incremental NN (INN) query is a kind of NN queries and can also be used when the number of the NN objects to be retrieved is not known in advance. This paper points out the weaknesses of the existing INN search algorithms and proposes a new one, called Batch-Incremental Nearest Neighbor search algorithm (denoted B-INN search algorithm), which can be used to process the INN query efficiently. The B-INN search algorithm is different from the existing INN search algorithms in that it does not employ the priority queue that is used in the existing INN search algorithms and is very CPU and memory intensive for large databases in high-dimensional spaces. And it incrementally reports b(b > 1) objects simultaneously (Batch-Incremental), whereas the existing INN search algorithms report the neighbors one by one. In order to implement the B-INN search, a new search (called k-d-NN search) with a new pruning strategy is proposed. Performance tests indicate that the B-INN search algorithm clearly outperforms the existing INN search algorithms in high-dimensional spaces.

The present paper deals with the frequency-dependent finite difference time domain ((FD)2TD) method analysis of the light-beam diffraction from a land/groove recording phase-change (PC) disk model with a metal (Al or Au) reflective layer in order to improve the conventional analysis for PC optical disk models with a perfectly conducting reflective layer. The diffracted fields are numerically calculated for both recorded and non-recorded states of the recording layer, and the comparison of the detected signal characteristics between two states is discussed. The crosstalk between the recording marks on lands and grooves are evaluated and the optimum groove depth is examined for Al,Au and perfectly conducting layer models.

Due to the explosive growth of the network technologies, 3D models and animations have led to a great interest in various media. Especially 3D mesh models (3D meshes), which approximate surfaces by polygonal meshes are widely used to model 3D objects. In 1D and 2D signals such as speech, audio, images, video, etc., the signal values are located on "grids", for example the signals of images are defined on pixels. Thus, the errors of such signals can be explicitly defined by differences of the values on the "grids". However since in the 3D meshes, vertices are located on arbitrary positions in a 3D space and are triangulated in arbitrary ways, the grids cannot be defined. This makes it difficult to measure error on the 3D meshes. In this paper, we propose a new numerical method to measure the errors between two different 3D meshes.

An exact and an approximated reliabilities of a 2-dimensional consecutive k-out-of-n:F system are discussed. Although analysis to obtain exact reliability requires many calculation resources for a system with a large number of components, the proposed method obtains the reliability lower bound by using a combinatorial equation that does not depend on the system size. The method has an assumption on the maximum number of failed components in an operable system. The reliability is exact when the total number of failed components is less than the assumed maximum number. The accuracy of the method is confirmed by numerical examples.

In this paper, we present a novel approach for efficient search of high-dimensional databases, such as video shots. The idea is to map feature vectors from the high-dimensional feature space into a point in a low-dimensional distance space. Then, a spatial access method, such as an R-tree, is used to cluster these points based on their distances in the low-dimensional space. Our mapping method, called topological mapping, guarantees no false dismissals in the result of a query. However, the result of a query might contain some false alarms. Hence, two refinement steps are performed to remove these false alarms. Comparative experiments on a database of video shots show the superior efficiency of the topological mapping method over other known methods.

A new power and bandwidth efficient modulation technique - Superposed Quadrature Quadrature Amplitude Modulation (SQ2AM) - for use in nonlinear satellite channel is presented. SQ2AM technique expands 2-dimensional SQAM signals into 4-dimensional quadrature modulated signals by using orthogonal baseband waveforms and carriers. The power spectrum and BER performance of SQ2AM are analyzed and compared with those of QPSK, SQAM and Q2PSK in a nonlinearly amplified multi-channel environment.

A new analysis method for random dopant induced threshold voltage fluctuations by using Monte Carlo ion implantation were presented. The method was applied to investigate Vt fluctuations due to statistical variation of pocket dopant profile in 0.1µm MOSFET's by 3D process-device simulation system. This method is very useful to analyze a statistical fluctuation in sub-100 nm MOSFET's efficiently.

It is still an open problem to elucidate the scaling merits of an embedded SRAM with Low Operating Power (LOP) MOSFETs fabricated in 50, 70 and 100 nm CMOS technology nodes. Taking into account a realistic SRAM cell layout, we evaluated the parasitic capacitance of the bit line (BL) as well as the word line (WL) in each generation. By means of a 3-Dimensional (3D) interconnect simulator (Raphael), we focused on the scaling merit through a comparison of the simulated SRAM BL delay for each CMOS technology node. In this paper, we propose two kinds of original interconnect structure which modify ITRS (International Technology Roadmap for Semiconductors), and make it clear that the original interconnect structures with reduced gate overlap capacitance guarantee the scaling merits of SRAM cells fabricated with LOP MOSFETs in 50 and 70 nm CMOS technology nodes.

This paper concerns recognizing 3-dimensional object using proposed multi-layer block model. In particular, we aim to achieve desirable recognition performance while restricting the computational load to a low level using 3-step feature extraction procedure. An input image is first precisely partitioned into hierarchical layers of blocks in the form of base blocks and overlapping blocks. The hierarchical blocks are merged into a matrix, with which abundant local feature information can be obtained. The local features extracted are then employed by the kernel based support vector machines in tournament for enhanced system recognition performance while keeping it to low dimensional feature space. The simulation results show that the proposed feature extraction method reduces the computational load by over 80% and preserves the stable recognition rate from varying illumination and noise conditions.

We propose herein a motion detection artificial vision model which uses analog electronic circuits. The proposed model is comprised of four layers. The first layer is a differentiation circuit of the large CR coefficient, and the second layer is a differentiation circuit of the small CR coefficient. Thus, the speed of the movement object is detected. The third layer is a difference circuit for detecting the movement direction, and the fourth layer is a multiple circuit for detecting pure motion output. When the object moves from left to right the model outputs a positive signal, and when the object moves from right to left the model outputs a negative signal. We first designed a one-dimensional model, which we later enhanced to obtain a two-dimensional model. The model was shown to be capable of detecting a movement object in the image. Using analog electronic circuits, the number of connections decrease and real-time processing becomes feasible. In addition, the proposed model offers excellent fault tolerance. Moreover, the proposed model can be used to detect two or more objects, which is advantageous for detection in an environment in which several objects are moving in multiple directions simultaneously. Thus, the proposed model allows practical, cheap movement sensors to be realized for applications such as the measurement of road traffic volume or counting the number of pedestrians in an area. From a technological viewpoint, the proposed model facilitates clarification of the mechanism of the biomedical vision system, which should enable design and simulation by an analog electric circuit for detecting the movement and speed of objects.

This paper investigates properties of space-bounded "two-dimensional Turing machines (2-tm's)," whose input tapes are restricted to square ones, with bounded input head reversals in vertical direction. We first investigate a relationship between determinism and nondeterminism for space-bounded and input head reversal-bounded 2-tm's. We then investigate how the number of input head reversals affects the accepting power of sublinearly space-bounded 2-tm's. Finally, we investigate necessary and sufficient spaces for three-way 2-tm's to simulate four-way two-dimensional finite automata with constant input head reversals.

Compatibility of conventional lossless discrete cosine transforms (LDCTs) with the discrete cosine transform (DCT) is not high due to rounding operations. In this paper, we design an LDCT which has high compatibility with the DCT. We first design an 8-point DCT (DCT3) by changing the order of row of the transform matrix and also the way of decomposing the DCT in order to obtain an 8-point LDCT which has high compatibility with the DCT. Next we design an 88-point nonseparable 2D LDCT based on a 4-point lossless Walsh-Hadamard Transform (LWHT) which is multiplier-free. The DCT3 is used, when the nonseparable 2D LDCT is designed. Simulation results show that compatibility of the nonseparable 2D LDCT with the separable 2D DCT is high. We also design an 88-point nonseparable 2D LWHT which is multiplier-free and indicate that its compatibility with the separable 2D Walsh-Hadamard Transform is high.

A number of results on the estimation of direction of arrival have been obtained based on the assumption that the signal sources are point sources. Recently, it has been shown that signal source localization can be accomplished more adequately with distributed source models in some real surroundings. In this paper, we consider modeling of three-dimensional distributed signal sources, in which a source location is represented by the center angles and degrees of dispersion. We address estimation of the elevation and azimuth angles of distributed sources based on the proposed distributed source modeling in the three-dimensional space using two linear arrays. Some examples are included to more explicitly show the estimation procedures under the model: numerical results obtained by a MUSIC-based method with two uniform linear arrays are discussed.

Optical near-field distributions of planar dielectric and metallic objects placed on a large dielectric substrate plate have been calculated by the volume integral equation using an iterative method called generalized minimal residual method with the fast Fourier transform technique. The basic characteristics of the near-field have been investigated in detail for large and small objects, dielectric and metallic objects and incident p-polarized and s-polarized evanescent fields.

This paper investigates the behavior of one-dimensional discrete-time binary cellular neural networks with both the A- and B-templates and gives the necessary and sufficient conditions for the above network to be stable for unspecified fixed boundaries.