A new impulse noise detection algorithm is presented, which can successfully remove impulse noise from corrupted images while preserving image details. The impulse detection algorithm is combined with median filtering to achieve noise removal. The main advantage of the proposed algorithm is that it can detect the impulse noise with high accuracy while reducing the probability of detecting image details as impulses. Also, it can be applied iteratively to improve the quality of restored images. It is efficient and low in complexity. Furthermore, it requires no previous training. Extensive experimental results show that the proposed approach significantly outperforms many well-known techniques.

Finding DC operating points of nonlinear circuits is an important problem in circuit simulation. The Newton-Raphson method employed in SPICE-like simulators often fails to converge to a solution. To overcome this convergence problem, homotopy methods have been studied from various viewpoints. There are several types of homotopy methods, one of which succeeded in solving bipolar analog circuits with more than 20000 elements with the theoretical guarantee of global convergence. In this paper, we propose an improved version of the homotopy method that can find DC operating points of practical nonlinear circuits smoothly and efficiently. Numerical examples show the effectiveness of the proposed method.

We propose an accurate and efficient model of having an unbalanced differential line structure, where mode-conversion and frequency dependent loss effects are considered in above the GHz frequency range. To extract model parameters of the proposed unbalanced differential line model, we measured s-parameters of test patterns using a 2-port VNA and defined a new type of mixed-mode s-parameter. The model parameters were obtained and are described for various types of the unbalanced differential line structures. Finally, the validity of the proposed model and the model parameters were successfully confirmed by a series of time-domain measurements and a lattice diagram analysis.

To develop human interfaces such as home information equipment, highly capable word learning ability is required. In particular, in order to realize user-customized and situation-dependent interaction using language, a function is needed that can build new categories online in response to presented objects for an advanced human interface. However, at present, there are few basic studies focusing on the purpose of language acquisition with category formation. In this study, taking hints from an analogy between machine learning and infant developmental word acquisition, we propose a taxonomy-based word-learning model using a neural network. Through computer simulations, we show that our model can build categories and find the name of an object based on categorization.

This paper presents a novel method of structure selection and identification for Hammerstein type nonlinear systems. An unknown nonlinear static part to be estimated is approximately represented by an automatic choosing function (ACF) model. The connection coefficients of the ACF and the system parameters of the linear dynamic part are estimated by the linear least-squares method. The adjusting parameters for the ACF model structure, i.e. the number and widths of the subdomains and the shape of the ACF are properly selected by using a genetic algorithm, in which the Akaike information criterion is utilized as the fitness value function. The effectiveness of the proposed method is confirmed through numerical experiments.

In deep submicron designs, the interconnect wires play a major role in the timing behavior of logic gates. The effective capacitance Ceff concept is usually used to calculate the delay of gate with interconnect loads. In this paper, we present a new method of Integration Approximation to calculate Ceff. In this new method, the complicated nonlinear gate output is assumed as a piecewise linear (PWL) waveform. A new model is then derived to compute the value of Ceff. The introduction of Integration Approximation results in Ceff being insensitive to output waveform shape. Therefore, the new method can be applied to various output waveforms of CMOS gates with RC-π loads. Experimental results show a significant improvement in accuracy.

Principal Component Analysis (PCA) has been applied in various areas such as pattern recognition and data compression. In some cases, however, PCA does not extract the characteristics of the data-distribution efficiently. In order to overcome this problem, we have proposed a novel method of Nonlinear PCA which preserves the order of the principal components. In this paper, we reduce the dimensionality of image data using the proposed method, and examine its effectiveness in the compression and recognition of images.

This paper presents Dynamic Attention Map by Ising model for face detection. In general, a face detector can not know where faces there are and how many faces there are in advance. Therefore, the face detector must search the whole regions on the image and requires much computational time. To speed up the search, the information obtained at previous search points should be used effectively. In order to use the likelihood of face obtained at previous search points effectively, Ising model is adopted to face detection. Ising model has the two-state spins; "up" and "down". The state of a spin is updated by depending on the neighboring spins and an external magnetic field. Ising spins are assigned to "face" and "non-face" states of face detection. In addition, the measured likelihood of face is integrated into the energy function of Ising model as the external magnetic field. It is confirmed that face candidates would be reduced effectively by spin flip dynamics. To improve the search performance further, the single level Ising search method is extended to the multilevel Ising search. The interactions between two layers which are characterized by the renormalization group method is used to reduce the face candidates. The effectiveness of the multilevel Ising search method is also confirmed by the comparison with the single level Ising search method.

The problem of aggregating different stochastic process into a unique one that must be characterized based on the statistical knowledge of its components is a key point in the modeling of many complex phenomena such as the merging of traffic flows at network nodes. Depending on the physical intuition on the interaction between the processes, many different aggregation policies can be devised, from averaging to taking the maximum in each time slot. We here address flows averaging and maximum since they are very common modeling options. Then we give a set of axioms defining a general aggregation operator and, based on some advanced results of functional analysis, we investigate how the decay of correlation of the original processes affect the decay of correlation (and thus the self-similar features) of the aggregated process.

Phonetic transcription of text is an indispensable component of text-to-speech (TTS) systems and is used in acoustic modeling for speech recognition and other natural language processing applications. One approach to the transcription of written text into phonetic entities or sounds is to use a set of well-defined context and language-dependent rules. The process of transcribing text into sounds starts by preprocessing the text and representing it by lexical items to which the rules are applicable. The rules can be segregated into phonemic and phonetic rules. Phonemic rules operate on graphemes to convert them into phonemes. Phonetic rules operate on phonemes and convert them into context-dependent phonetic entities with actual sounds. Converting from written text into actual sounds, developing a comprehensive set of rules, and transforming the rules into implementable algorithms for any language cause several problems that have their origins in the relative lack of correspondence between the spelling of the lexical items and their sound contents. For Standard Malay (SM) these problems are not as severe as those for languages of complex spelling systems, such as English and French, but they do exist. In this paper, developing a comprehensive computerized system for processing SM text and transcribing it into phonetic entities and evaluating the performance of this system, irrespective of the application, is discussed. In particular, the following issues are dealt with in this paper: (1) the spelling and other problems of SM writing and their impact on converting graphemes into phonemes, (2) the development of a comprehensive set of grapheme-to-phoneme rules for SM, (3) a description of the phonetic variations of SM or how the phonemes of SM vary in context and the development of a set of phoneme-to-phonetic transcription rules, (4) the formulation of the phonemic and phonetic rules into algorithms that are applicable to the computer-based processing of input SM text, and (5) the evaluation of the performance of the process of converting SM text into actual sounds by the above mentioned methods.

This investigation proposes a fast wavelet-based color segmentation (FWCS) technique and a modified directional region-growing (DRG) technique for semantic image segmentation. The FWCS is a subsequent combination of progressive color truncation and histogram-based color extraction processes for segmenting color regions in images. By exploring specialized centroids of segmented fragments as initial growing seeds, the proposed DRG operates a directional 1-D region growing on pairs of color segmented regions based on those centroids. When the two examined regions are positively confirmed by DRG, the proposed framework subsequently computes the texture features extracted from these two regions to further check their relation using texture similarity testing (TST). If any pair of regions passes double checking with both DRG and TST, they are identified as associated regions. If two associated regions/areas are connective, they are unified to a union area enclosed by a single contour. On the contrary, the proposed framework merely acknowledges a linking relation between those associated regions/areas highlighted with any linking mark. Particularly, by the systematic integration of all proposed processes, the critical issue to decide the ending level of wavelet decomposition in various images can be efficiently solved in FWCS by a quasi-linear high-frequency analysis model newly proposed. The simulations conducted here demonstrate that the proposed segmentation framework can achieve a quasi-semantic segmentation without priori a high-level knowledge.

The optical network is a promising approach for realizing a scalable backbone network. In backbone networks, survivability is very important because great volumes of traffic incur damage from faulty equipment. To address this issue, various recovery schemes have been proposed for optical backbone networks. Among those schemes, shared mesh restoration utilizes link bandwidth efficiently because the backup lightpaths share link bandwidth if they protect against different failures and are never utilized simultaneously. However, a route computation method for the backup lightpaths that promotes such bandwidth sharing is necessary to achieve efficient bandwidth utilization. This paper proposes a distributed route computation method for the backup lightpaths in shared mesh restoration. In this method, the link weight is estimated to be smaller if a backup lightpath newly established can share the link bandwidth with the backup lightpaths already accommodated in that link. The link weight can be calculated using the Markov Decision Theory. The bandwidth sharing between the backup lightpaths can be promoted by selecting the shortest route based on such modified link weights. The proposed method effectively realizes efficient utilization of the link bandwidth and achieves low loss rate of reliable lightpath establishment requests under the same traffic load. The proposed method restricts the amount of link state information advertised by the routing protocol and achieves a sufficiently small amount of route calculation.

This investigation proposes a virtual-FIFO (VFIFO) back-off algorithm for wireless networks. The proposed scheme takes advantage of the central unit (CU) in a wireless network to broadcast a common back-off window size to all the users, significantly alleviating the unfairness of bandwidth utilization in conventional binary exponential back-off (BEB) algorithms. The proposed scheme exploits the CU's capability for collision detection to estimate the number of simultaneously competing users. Additionally, packets generated in a given cycle are split into groups according to their times of arrivals and are guaranteed to be serviced one after another within the next cycle. Although the proposed algorithm is not strictly first come fist served, the FIFO principle is virtually accomplished. Simulation results demonstrate that the standard deviation of delay can be improved by more than two orders and the throughput can be maintained at 0.42 when the number of users approaches infinity. The capture effect even further improves system performance.

The optimal antenna weighting scheme that minimizes the average bit error rate in a closed-loop transmit antenna diversity system is investigated under the assumption that channel state information is provided at both the transmitter and the receiver. A closed-form expression for the optimal transmitter weights is derived with a fixed average transmit power constraint. Also, the effect of limited peak transmit power on the performance of the optimal weighting method is analyzed. Base on this analysis, it is shown that the proposed transmitter weights yield significant performance improvements over the conventional weights on the wide range of practical system parameters.

Recent advances in 77-GHz MMIC module design techniques for automotive radar applications are reviewed in this paper. The target of R&D activities is moving from high performance to low cost, mass production, high-yield manufacturing and testing. To meet the stringent requirements, millimeter-wave module design techniques have made significant progress especially in packaging, bonding, and making interface with other modules. In addition, millimeter-wave semiconductor devices and MMICs have made remarkable improvements for low cost and mass production. In this paper, the topics focusing on millimeter-wave semiconductor devices and 77-GHz MMICs are reviewed first. Then the recent R&D results on 77-GHz MMIC module design techniques are introduced, showing the technical trend of packaging, bonding, and making interface with other modules for millimeter-wave, highly-integrated, low-cost MMIC modules. Finally, the existing and future module design issues for automotive radar applications are discussed.

The millimeter-wave (MMW) broadband mixers that are useful for measurement instruments to analyze MMW high data rate signals have been investigated. At first, we propose the specialized RF front-end for analyses of MMW high data rate signals. Next, the required specifications for the 1st mixers of the front-end are estimated, and the design, fabrication, and testing results of Q, V, and W-band monolithic broadband resistive mixers are described. The testing results are compared with performances of the diode mixer designed for V-band. It was found that the resistive mixers have very attractive performances of low conversion loss, good frequency flatness and high third order intercept point (IP3) with low Local (LO) oscillators power. The developed resistive mixers are suitable for the proposed MMW band measurement instruments.

In this paper, we present a learning approach, positive correlation learning (PCL), that creates a multilayer neural network with good generalization ability. A correlation function is added to the standard error function of back propagation learning, and the error function is minimized by a steepest-descent method. During training, all the unnecessary units in the hidden layer are correlated with necessary ones in a positive sense. PCL can therefore create positively correlated activities of hidden units in response to input patterns. We show that PCL can reduce the information on the input patterns and decay the weights, which lead to improved generalization ability. Here, the information is defined with respect to hidden unit activity since the hidden unit plays a crucial role in storing the information on the input patterns. That is, as previously proposed, the information is defined by the difference between the uncertainty of the hidden unit at the initial stage of learning and the uncertainty of the hidden unit at the final stage of learning. After deriving new weight update rules for the PCL, we applied this method to several standard benchmark classification problems such as breast cancer, diabetes and glass identification problems. Experimental results confirmed that the PCL produces positively correlated hidden units and reduces significantly the amount of information, resulting improved generalization ability.

As a representative of the linear discriminant analysis, the Fisher method is most widely used in practice and it is very effective in two-class classification. However, when it is expanded to a multi-class classification problem, the precision of its discrimination may become worse. A main reason is an occurrence of overlapped distributions on the discriminant space built by Fisher criterion. In order to take such overlaps among classes into consideration, our approach builds a new discriminant space by hierarchically classifying the overlapped classes. In this paper, we propose a new hierarchical discriminant analysis for texture classification. We divide the discriminant space into subspaces by recursively grouping the overlapped classes. In the experiment, texture images from many classes are classified based on the proposed method. We show the outstanding result compared with the conventional Fisher method.

Recently, multiple-input multiple-output (MIMO) systems that realize high bit rate data transmission with multiple antennas at both transmitter and receiver have drawn much attention for their high spectral efficiency. In MIMO systems, space division multiplexing (SDM) has been researched widely. In SDM, the input data symbols are transmitted from multiple transmit antennas at the transmitter, and the output data symbols are extracted by the signal processing at the receiver. In recent wireless communications, the environments that the number of transmit antennas is larger than that of receive antennas often exist. Under such environments, the MIMO system that transmits independent data streams from each transmit antenna simultaneously cannot separate the received signals, and the signal quality deteriorates largely. Therefore, we need the scheme that attains high quality and high throughput data transmission under such environments. In this paper, we propose a throughput maximization transmission control scheme for MIMO systems. The proposed transmission control scheme selects a transmission scheme (a set of transmit antennas, modulation schemes, and coding rates) with maximum throughput based on output signal to interference and noise ratio (SINR) and output signal to noise ratio (SNR). We show that the proposed transmission control scheme attains high throughput by our computer simulation.

In this paper, a semi-dynamic protection path configuration method is proposed for WDM optical networks. In the method, the protection path is established by connecting several sub-lightpaths from the source node to the destination node of the original working lightpath, as opposed to conventional path restoration method where a single protection lightpath between the source-destination pair performs restoration. The proposed method provides enhanced flexibility in protection path configuration and relieves the cost of spare capacity reservation. This paper also studies the effects of wavelength conversion capability of intermediate optical cross-connect nodes on protection path routing and spare capacity utilization. In terms of spare capacity utilization, the proposed method shows substantial reduction of spare capacity overhead compared with dedicated path restoration in all optical networks without wavelength conversion, and shows similar capacity efficiency compared with shared path restoration in opaque networks with full wavelength conversion capability. In terms of robustness, the proposed method shows nearly the same restoration ratio for double-link failure as that of dynamic restoration method.