This article provides an overview of an energy-based approach to the study of the steady-state and transient performances of adaptive filters. The analysis employs energy-conservation arguments and is based on studying the energy flow through each iteration of an adaptive filter. Among other results, the approach characterizes the transient behavior of adaptive filters in terms of a linear time-invariant state-space model. The stability and steady-state behavior of the model then translate into stability and mean-square performance results for the adaptive filter. In addition to deriving earlier results in a unified manner, the approach does not restrict the regression data to being Gaussian or white.

抄録全体を表示

In this paper, a clustering method is proposed to segment all circumferential range image , which is employed as preprocessing of 3-D shapes modeling for the object that has arbitrary topology. The proposed method performs the image segmentation that the range data points are classified into domains with similar quadric surfaces. In order to determine the desirable clusters, this method consists of two stages. First, New clusters are created on demand with learning the range data in recursive form (Stage 1). However, a number of wrong clusters are created. In next stage, after dispensable clusters are vanished by the competition of clusters, remaining clusters are gradually agglomerated (Stage 2). Consequently, since the only appropriate clusters are left, the image segmentation can be performed by assigning the data to these clusters.

抄録全体を表示

A detection scheme is presented for maneuvering affordance in noisy imagery. Under the assumption that image features to be observed are generated as fractal attractors induced by mental maneuvering process, environment features are represented in directional scale images. For eliminating background noise in this sampling process, pixelwise filter is designed based on probabilistic complexity analysis. The detectability of affordance patterns has been verified through experimental studies.

抄録全体を表示

In signature verification and/or character recognition problems, matching of two drawings has a certain significance [3]. For these problems, online equipments such as a tablet and an electronic pen are simultaneously used. In an environment where online equipment can be used, using these is more efficient than simply using a pen and a sheet of paper. The latter is called “offiine method”. Two advantages of using online equipments arise from here. One is that the analyser is free from the heavy computational load of image processing which is inherent to offline methods. Thus the analysis can also be “online” without waiting the availability of the whole signature. The other is that the analyser can use various information of writing which is usually unavailable in offiine method. An example can be shown in the signature verification. If the holder of the signature intentionally writes a part of the signature in a strange direction, others are not likely to be able to imitate it. Another simple technique for making the signature not likely to be imitated by other persons is, say, to put a dot in the right end and begin the signature from the left end. The drawings appearing on the screen does not explicitly show the secret on the screen, but the analyser or a computer can obviously detect this secret and is likely to find the forgery whose writer has not noticed the secret of the true signature. Thus we can believe that the online matching is especially useful in signature verification problems. It is clearly necessary to detect the difference between the genuine and fake ones written in a standard way of signing. This paper gives an analysis of a method proposed for this purpose.

抄録全体を表示

The recursive maximum filter (RMF) is an algorithm devised to solve the problem of detecting small moving objects in noisy image sequences. This problem arises in such applications as small target detection for infrared image sensors, and dim star detection for space-borne star sensors. In this paper, we analyze the performance of the RMF. First, we formulate the RMF as a Bellman equation for the dynamic programing. Then, false alarm probability is evaluated using the extreme value theory. Finally, detection probability is obtained as a function of input signal to noise ratios (SNRs). It is shown that input SNRs required for target detection can be reduced to less than half by the RMF.

抄録全体を表示

We develop the speech-displaying system called “KanNon” ^{[1, 2]} which supports speech communication of hearing-impaired people. The KanNon system displays a sound spectrogram and corresponding characters by speech recognition, which is scrolling on-line. To develop the KanNon system with the high accurate spectrum estimation of the speech, we adopt new spectral estimation method called “MCE-Burg^[3]”, which is the method combining the continuous minimum cross entropy method(C-MCEM)^{[4, 5]} and Burg method^[6]. Finally, we compare sound spectrograms estimated by the Burg method and the MCE-Burg method, and show that results by the MCE-Burg method has sharper contrasts and peaks of formant compared with Burg method.

抄録全体を表示

A de-noising using a Wavelet Packet transform is described. An optimum threshold for de-noising was obtained. A new selecting method of a transform level and an order of the wavelet function is also proposed. For the threshold of an adaptive de-noising. the noise estimated from the speech signals was used. The estimated noise was proportional to the optimum threshold. The other two de-noising methods were compared. The proposed method has the most prominent property among the three methods.

抄録全体を表示

A transcript algorithm suitable for multi-voice music which sounds several tones simultaneously has been developed. The algorithm to which Wavelet Packets analyses and Extended Kalman Filtering are applied is designed for log-normal state vector, distributed initial phase and observation noise under the Central Limit Theorem. Derivation of the algorithm as well as practical performance for transcription of chords, intervals and single tones are introduced.

抄録全体を表示

For the evaluation of actual environmental acoustic systems between inside and outside of the room, a new method predicting its system response probability distribution is proposed. More concretely, first, the extended regression functions (of involution) of the output on intensity scale are introduced as the system model matched to the prediction. To identify these system parameters, the recursive algorithm based on the criteria that there remains only accidental fluctuation in the model error when they are chosen optimally is newly derived. Since the parameters of the general form of the conditional probability distribution conditioned by input are determined, the expression of the probability distribution of the output without contamination of background noise derived from this can be employed for prediction with a new input. Finally, the proposed method is experimentally confirmed by applying it to actual indoor-outdoor acoustic system.

抄録全体を表示

In this study, as a concrete example of simplified nonlinear inverse model in an indoor sound field, a complex sound fluctuation leaked through an aperture in the wall of coupled room under an actual sound environment has been discussed. First, a probability evaluation theory introducing a simple stochastic inverse system model has been proposed especially from an object-oriented viewpoint. More concretely, in order to evaluate the output response probability distribution of the acoustic intensity, after introducing the time-averaged stochastic variable, we have identified functionally the leakage property for any variability of the transmission factor. Then, we have predicted the output response probability distribution for a new arbitrary stochastic input once after establishing a prediction theory for the probability distribution of a multiplicative model. Finally, an effectiveness of the proposed theory has been experimentally confirmed too by an application to an actual complex indoor sound field.

抄録全体を表示

This paper presents some convergence results of an adaptive filter algorithm for the feedback-type active noise control (ANC) system with feedback path modeling using a dither signal and that of hearing aids. In these systems there is an adaptive filter in the feedback path so that the conventional method for analysis does not work. A frequency domain method is used for the analysis. It is based on the averaging method and the frequency domain expression of the adaptive algorithm with the causality constraint. The convergence conditions are derived for both problems and a bias expression of the weights for the latter problem is presented. The positive realness of the transfer function of the optimal predictor of the incoming signal is required for stability. Finally, the convergence condition and the bias are explicitly given for simple examples and their validities are shown by some simulations.

抄録全体を表示

The present paper is concerned with the adaptive cancellation of coupling wave effect in the Single Frequency Networks (SFN) relay station of the Orthogonal Frequency Division Multiplexing (OFDM) digital terrestrial broadcasting system. Due to lack of the PE condition, the coupling wave channel dynamics outside the signal bands cannot be identified precisely and that makes the stable cancellation very difficult. It is illustrated that both the estimation of the system model and the evaluation of the modeling error are important in the stable canceller design procedure.

抄録全体を表示

We consider the problem of optimal transmission of correlated Gaussian signals through a set of parallel channels with feedback. The original signals are assumed to be given by a solution of a multi-dimensional linear stochastic differential equation and are sent by a linear encoding with output feedback. It is well-known that the optimal output feedback which maximizes the mutual information between the original signals and the channel output is given by the least-squares estimate of the linear term. Under a constraint on the total power of the encoded signals, we consider the optimal transmission problem of computing a set of gains for the channels which maximizes reduction of the estimation error.

抄録全体を表示

Conventional algorithms for blind source separation do not necessarily work well for real-world data. One of the conceivable reasons is that, in actual applications, the mixing matrix is often almost singular at some part of frequency range and it can cause a certain computational instability. This paper proposes a new algorithm to overcome this problem. The algorithm is based on the minimal distortion principle proposed by the authors and in addition incorporates a kind of regularization term into it, which has a role to suppress the gain of the separator for the frequency range at which the mixing matrix is almost singular.

抄録全体を表示

In open models of systems with several types of agents, new types of agents that are unexpected or unanticipated may enter at some future time. This possibility is largely ignored in the mainstream economic literature. This paper introduces the results of the two-parameter inductive methods of Ewens, Pitman and Zabell to deal with this problem. These methods postulte entries by unknown types of agents, goods, behavioral rules, or strategies into the model, conditional on the observed data. Ewens one-parameter distribution has been introduced in Aoki (2000c). In economic applications we impose the notion of exchangeable random partitions of a finite set. This produces simple rule of sucession, that is, the expressions for probabilties for entries by new or known types, conditional on the observed data. Using the notion of exchangeable random partitions of a set of finite number of agents, we specify the continous-time Markov chains as economic models with transition rates to reflect the above constraints.

In addition to the usual state vector which lists the number of agents by types, we explain the use of partition vector as state vector in models composed of a large number of exchangeable agents of possibly many types. We illustrate the use of this new state vector by describing the equilibrium distribution of agents by types in partition vector form. We suggest that this approach is also useful in agent-based simulations once we use the notion of holding times to randomly select agents that "acts" first.

抄録全体を表示

Many spatio-temporal patterns such as target and spiral patterns in the BZ reaction are observed in the various fields of engineering, including fluid and chemical engineering. The analysis of such patterns is one of important problems of engineering. In this paper, focusing our attention on the Turing patterns, the stochastic modeling of Turing pattern formations and the influence the disturbance on the patterns are studied through numerical simulations.

抄録全体を表示

We study the filtering problem for the stochastic volatility model of Heston by using the nonlinear estimation theory. To solve the estimation problem for the stochastic volatility process, we use the random time change method. The derived basic equation for the filtering is the so-called Zakai equation and its numerically realized algorithm is proposed with the aid of the splitting-up method. Regarding the European call option problem, the identification of the market price of the volatility risk is also studied. Some numerical simulation studies are demonstrated to show the advantage of the proposed method.

抄録全体を表示

The authors previously developed a dynamic model based accurate pipe length measurement system using stationary waves generated spontaneously or by hitting the pipe. However, two problems rise when the system is applied to the length measurement of steel pipes at factories. The first is that a considerably large noise is continuously generated around the measuring spot and thus it causes low SN ratio in the observation. The second is that high order modes of the stationary wave cannot be available because the diameter of the steel pipes is relatively large compared to the pipe length. This paper thus proposes an accurate length measurement of a pipe under noisy environment by considering the selections of an optimal sensor location and an optimal data window.

抄録全体を表示

In Carrier Phase Differential GPS (Global Positioning System) ionospheric and tropospheric delays are deteriorated the accuracy of positioning.

In this paper, in order to achieve accurate positioning, ionospheric and tropospheric delays estimated and corrected.

抄録全体を表示

In this paper, we propose a new method for land-vehicle In-Motion Alignment by integrating Inertial Navigation System (INS) and the transfer Differential GPS (DGPS). Our aim in this work is to integrate the advantages of these systems and to develop the navigation system that does not require initial attitudes information of the vehicle by applying recursive nonlinear filterling techniques.

抄録全体を表示

In a case of flood, earth and sand at the bottom of bridge piers are sometimes scoured and thus yields unstableness of the bridge piers. Therefore, constant monitoring of bridge piers is of great importance in keeping the safety operation of trains. The paper thus proposes a simplified monitoring system which uses an acceleration pickup and analyzes the vibration signal of bridge piers generated with a light hammer. The system is a dynamic system model based monitoring system for the bridge piers.

抄録全体を表示

Dependence of spatiotemporal response of the alpha wave on light intensity of white LED flicker stimuli is examined using a complex demodulation method and the Fourier transform. In spite of inter-individual differences in the spontaneous alpha waves, most of the alpha waves show similar response dependency on the light intensity: the alpha wave is not only attenuated but also destabilized to the weak stimuli with intensity less than 10lx, whereas it becomes not only enhanced but stabilized with increasing the stimulus intensity. The attenuated response to the weak stimuli is a novel observation on the dynamics of the alpha wave, while the enhancement of the alpha wave with increasing the stimulus intensity seems to be a reasonable result from the viewpoint of a nonlinear oscillators system. These results may indicate that the interaction or coupling strength between endogenous alpha-generators can be affected by the strength of the external photic stimulation. Furthermore, the observations obtained are useful for modeling the alpha wave by a coupled nonlinear oscillators system.

抄録全体を表示

In this paper, we propose the Elite Correlation Selection (ECS) operator as a new selection operator to improve the genotype diversity in genetic search process. In the proposed method, the candidate for selection are evaluated the genotype similarity among individuals, and the individual is selected based on the above new evaluated values. As a measure of the similarity, the authors use the hamming distance of the genotype. The proposed method also is the improvement of the selection operator of the Minimal Generation Gap (MGG) model. The performance of the proposed method is evaluated with the standard test functions. The performance of the proposed method are compared with the MGG model and the Simple Genetic Algorithm (SGA). The simulation results show that the proposed method is very effective.

抄録全体を表示

This paper proposes an adaptive random search algorithm called RasID. A sophisticated probability density function is introduced for generating random search vectors. The new algorithm has powerful tuning capabilities such that it can perform intensified searches in perspective areas and diversified searches to escape from local minima.

抄録全体を表示

This paper presents a relatively simple but quite effective identification methodology based on absolute acceleration response measurements. The proposed methodology consists of two stages. In the first stage, the first few modal vectors of a structure are estimated by utilizing the proper orthogonal decomposition technique. In the second stage, the bandpass-filtered input ground acceleration is initially estimated from the obtained modal vectors. Then, with the obtained input ground acceleration employed as an input, absolute acceleration responses subjected to the bandpass filter are used as an output vector in the ARX model-based identification. This ARX model provides more precise modal characteristics of a structure and updates the input ground acceleration time history. The concept of the proposed approach is presented and its validity is demonstrated by conducting numerical simulations and scale model experiments.

抄録全体を表示

In many cases it is not possible to remove the feedback during an identification experiment because this may lead to instability. The feedback may also be inherent in many dynamic systems (e.g., economical, biological). In this paper we shall consider the problem of identification of dynamic systems operating under feedback control and propose an indirect approach to closed-loop system identification based on combining with this basic idea and a bias-compensation principle. The main feature of the proposed method is that introducing auxiliary vector and matrix is helpful to find an approach to estimating the noise-output correlated variance. Therefore via compensating the consistent bias of LS estimate, the consistent BCLS estimates for closed-loop system can be given.

抄録全体を表示

This paper studies an abnormal behavior in vibration cutting system. Our experiments show that chatter-like unstable phenomenon occurs in the vibration cutting system. The unstable phenomenon changes its vibration type when time is varied. The purpose of this study is to determine the cause of the unstable phenomenon by using a vibraion cutting model described by nonlinear differential equations. Our numerical simulation shows that our model can produce similar behavior to the unstable phenomenon, when the vibration tool's amplitude A is reduced and Gaussian noise is superposed on A. This result suggests that the unstable phenomenon is caused by small and noisy movement of the vibration tool.

抄録全体を表示

Generally, cutting systems receive a bad influence from noise. In this paper, we study the influence of noises on cutting system by numerical simulations. We superpose Gaussian noises on the cutting speed parameter in Grabec's orthogonal cutting model. We show that the noises suppress chaotic vibration arised in the cutting model and reduce cutting forces. This result suggests that it is possible to suppress the abnormal behavior arised in real cutting systems by adding noises.

抄録全体を表示

This paper has concern with the analysis of the behavior of phase boundaries by stochastic level set approach. The stochastic level set equations have strong nonlinearity, so we formulate the system by introducing the stochastic viscosity solution, which is one of weak solutions of stochastic partial differential equations. Two types of motion of phase boundaries, growth with constant speed and mean curvature flow, are considered in this paper. And the influence of disturbances on the behavior of such phase boundaries is studied through simulations.

抄録全体を表示

This paper proposes a state estimation filter for detecting collisions with the flexible cantilevered-beam and obstacles. The beam is assumed to be subject to distributed random disturbance and unknown collision input. The purpose of the estimator is to detect the collision input and to estimate the strength of it. In this paper, a working model for the collision input in modal representation is considered for detection and estimation of the collision input and its strength. By applying the disturbance estimation technique, the state estimator for the collision input is constructed.

抄録全体を表示

A subspace identification method based on the orthogonal decomposition has been proposed in [3], where the stochastic realization of stationary process with exogenous inputs and its application to subspace identification have been discussed. In this paper, we consider two methods of identifying the deterministic part of the system. One is to estimate the deterministic part from the data obtained by the orthogonal projection of the output onto the space spanned by the input by the MOESP method [8], and the other one is to estimate the extended observability matrix based on the LQ decomposition of the data matrix [2]. We show by numerical examples that the identification results by the two methods are quite different for a system with sharp peaks in Bode plot.

抄録全体を表示

In this paper, an interpretation of subspace-based identification methods with Schur complement approach is proposed. MOESP (MIMO output error state space model identification) algorithms are well known as an elementary subspace method. The extensions of the MOESP with instrumental variables (IV) have been proposed in literatures, which can be useful to solve the stochastic identification problems. Data product moments corresponding to the data matrix in the MOESP algorithms are used. Then we show that the IV extensions in the MOESP-based methods can be expressed as modifications of the data product moment, and it enable us to treat the MOESP-based algorithms under a same framework even if the errors-in-variables (EIV) problems are considered.

抄録全体を表示

From the viewpoint of stochastic/statistical inverse problem, identification of transfer functions in feedback systems has been examined and applied to EEG/MEG analysis [1], since some transfer functions between output variables are invariant in a stochastic feedback system. However, there still remains an open problem of our method in scalar times transformations of output variables similar to filtering effect on transfer functions for pre-processing. From the viewpoint of normalization of data the minimum phase property of identified models is discussed to obtain transfer functions correctly. Scaling effects on transfer functions are examined in simulation study, and it is useful for diagnosis of feedback systems.

抄録全体を表示

Many electric power companies in Japan are installing the SCR ( Selective Catalytic Reduction) system for reduction of NOx(Nitrogen Oxide) emissions. While these systems are effective, control and maintenance problems often result in higher reagent ammonia (NH3) usage and NH3 slip due to a long time delay in the NOx signal from the stack and the nonlinearity of the process. This paper discusses significant aspects of the technical and economic evaluation by desktop simulation. The significant advantages of the proposed predictive control for the nonlinear process are that the multi-step ahead predictor is formulized as a nonlinear version of future control based on the RBF-type ARX models. The RBF-type ARX model-based predictive control is one of solutions for the SCR at the GTCC optimization and control. The benefits of the proposed solution are qualified in both economic and environmental terms.

抄録全体を表示

In this study, we proposed Kernel Principal Component Analysis (KPCA) which is applied for feature selection in a high-dimensional feature space which is nonlinearly mapped from an input space by a Gaussian kernel function. By using Mercer Kernels, we can compute principal components in a high dimensional feature space. Then, the extracted features are employed as preprocessing step for an ordinary least squares regression in the feature space which is Reproducing Kernel Hilbert Space (RKHS).

抄録全体を表示

An effective method of active control is proposed for suppressing random vibration of high-rise buildings caused by wind and seismic disturbance forces. Keeping in mind the mutually different characteristics of their disturbance forces, the active control is realized by constructing two types of controllers one of which is against the seismic force and the other against wind disturbance.

抄録全体を表示

The inverse Nyquist array method is one of the practical design methods in frequency region for the multi-input and multi-output system. In this method, the design of the noninteracting precompensator is reduced to the eigenvalue problem by the pseudo diagonalization method. However, the pseudo diagonalization method is applicable only to the system where numbers of input and output are equal. In this paper, we generalize the pseudo diagonalization method to be applicable to the system where numbers of input and output are not always equal.

抄録全体を表示

This paper first proposes a model of traffic signal control systems for avoiding traffic congestion in urban road networks. The model is controllable and observable, and therefore, modern control techniques such as LQ and H^∞ control are applicable. In addition, disturbances and stochastic changes of traffic situations are taken into account in the model. As a result, the proposed model is described as a system with multiplicative noise which is a kind of stochastic system. Next, a state-feedback control method based on linear matrix inequalities is applied to this system, and is shown to be effective in a numerical example.

抄録全体を表示

In this paper, we propose an efficient procedure of physical parameter identification of a magnetic levitation system, where the levitated steel ball is controlled by a robust nonlinear controller which is designed based on rough nominal parameters. Design techniques of the robust nonlinear controller are described and parameter identification results are included. Finally, it is shown that position tracking performance can be improved by using the identified parameters.

抄録全体を表示

A general methodology for designing chaotic and hyperchaotic cryptosystems was developed using the control systems theory. Grassi proposed a nonlinear-observer-based decrypter for the state of the encrypter. If we can design the decrypter without the knowledge of the parameters of the encrypter, the chaos-based secure communication systems are not secure. In this paper, we propose an observer-based chaotic communication system, which allows us to assign the relative degree and the zeros of its encrypter system. Moreover, we design a robust synchronizer based on linear controller parameterization.

抄録全体を表示

In this paper, the author proposes a class of fuzzy random vectors(FRVCs for short), which is the improved one of the previously proposed by the author[1] as the vague perception of crisp random phenomena. The FRVCs proposed here have intrinsically both properties of fuzziness and randomness and they are considered to be obtained as fuzzy perceptions of ordinary non-fuzzy random vectors.

First, the improved FRVCs are introduced by using the set representation approach of fuzzy sets and some their properties are investigated. Secondly, applying the multi-valued logic, the expectations of FRVCs are proposed and some their basic properties are investigated. Finally, the reasonable definition of conditional expectations for the proposed FRVCs is studied theoretically.

抄録全体を表示

Recently, chaos can be found in each field of engineering. And chaos control is one of the chaos applied researches in control engineering. It is to stabilize the state of the system into the unstable periodic orbit (UPO ) embedded in the strange attractor. Conventional chaos controls like deleyed feedback control (DFC ), prediction-based feedback control (PFC), etc. are performed for known systems in many cases. On the other hand, although there is the technique for the unknown system using the output from the neural network as the prediction value of the state, there exists the problem about the error of the prediction value.

Then, the method for the unknown discrete time system described by the fuzzy model (afterward, PDFC, prediction-based delayed feedback control ) is proposed. This is the technique of determining the prediction value of the state and control input parameters analytically using the fuzzy model and performing the hybrid type control of the DFC method and the PFC method.

抄録全体を表示