In this paper, in order to eliminate the effects of a background noise on an identification of sound insulation systems, two kinds of identification methods will be proposed especially on the basis of criterion of mutual information between input and output data conditioned by a constant high input level given in advance with few effects of the background noise. These methods will be experimentally applied to several type sound insulation systems, and then response probability distributions of the systems to a traffic noise input will be evaluated through the identification results.

View full abstract

This paper proposes a hierarchical modeling and identification approach to nonlinear systems. A general nonlinear system is represented as a hierarchical form called ARX Network. The parameters of ARX Network have a layered structure corresponding to the orders of system nonlinearity. This feature has been used to deal with nonlinear system with various complexity and to develop a hierarchical identification algorithm to estimate the parameters robustly. ARX Network holds a linear structure similar to linear ARMAX model which has been taken as advantage to apply the ARX Network to adaptive control and robust fault detection of nonlinear systems. Furthermore, the nonlinear nonparametric models are introduced to encompass the higher order nonlinearity of system hence the obtained hybrid ARX Networks has not only a good structure but also good flexibility. The effectiveness of the ARX Networks has been confirmed through experimental studies using both real data and simulated data.

View full abstract

The paper considers a problem of the identification for a dynamical system, on the assumption that the density of an observation noise is not available. The system parameters are estimated by a least squares method. The system structure is estimated by using a neural network.

View full abstract

This paper describes a stochastic identification for linear multi-degree-of-freedom structures, based on the noisy observations of excitation and seismic structural responses. The system equation and observation equation are formulated in the form of the Ito stochastic differential equation. The optimal estimates of stochastic state variables are given by determining the conditional mean. However, due to the difficulty of determining directly the conditional probability density function of the stochastic state variables, the procedure solving the conditional moment equations has been discussed here. Particularly, we will apply this method to response estimation of a real structure having observation records of the 1995 Hyogoken-Nanbu earthquake.

View full abstract

The wavelet-based algorithms are proposed for an on-line detection of a train of pulse signals corrupted by random noise. The key approach is to catch the time instances at which the pulse signal changes by computing the correlation between wavelet and observation data. The effectiveness of the proposed algorithms are confirmed by numerical simulation experiments.

View full abstract

In this paper we derive the correlation structure of wave packets coefficients. We provide an expression that describes the scale dependency for the correlation structure. In particular, we prove that de-correlation of fractional Brownian motion will decrease exponentially across wave packets scales. Based on what we derived, we discuss certain important properties of the correlation structure.

View full abstract

This paper presents a new non-model approach of fault detection via wavelet transform and its application to the fault detection of a long vehicle. It has good sharpness of response to vehicle faults and robustness to noises. Vehicle tyre and spring faults can be easily discriminated. Results of simulation and experimental investigation on a model vehicle are given.

View full abstract

This research is concerned with torque compensation for nonlinear servosystem using H-infinity robust estimator. The nonlinear friction is one of the important obstacles to high-speed and high-accuracy of servosystems. Until now, the effective methods have been proposed to carry out the compensation of nonlinear friction using observer. However, these methods are not applicable to the practical use, as the real system has not necessarily white noise processes and includes varying parameters due to the effect of load inertia. In order to overcome these problem, the H-infinity robust estimator is used to construct the servosystem with torque compensation. Through simulation experiments, it is verified that good estimation of torque can be obtained by the use of H-infinity robust estimator, and high-accurate position control can be realized by the torque compensation.

View full abstract

We developed a visual system. This system is able to track the moving objects within a certain range of errors. The visual system is driven by two DC motors. They are controlled based on the visual data obtained from a CCD video camera. The problem how to implement a fuzzy logic control method in this system is also considered in order to check the control performance of tracking. The fuzzy logic algorithm is a powerful control technique for nonlinear dynamical system.

View full abstract

Fuzzy identification methods have been successfully applied to systems where nonlinearity is one of the main constraint feature. However noise inference of the fuzzy identification has been discussed in detail in such studies. In this paper we propose an efficient fuzzy identification method which has two strong points. The one is how to reduce the number of implication to construct a better fuzzy model. The other is to remove the noise effects when the systems are embedded in noise. By using this method we demonstrate a better fuzzy modeling result for such systems.

View full abstract

In this paper, the author tries to reconstruct fuzzy random vector(FRVCs) by using set representation approach. Fuzzy random vectors as vague perception of random phenomena is proposed by author, which have intrinsically both properties of fuzziness and randomness and they are obtained as fuzzy perceptions of ordinary non-fuzzy random vectors. First, FRVCs are re-defined by using the set representation of fuzzy sets. Secondly, adopting the multi-valued logic, the expectation of FRVCs are defined. Finally, some of the statistical properties of FRVCs are studied theoretically.

View full abstract

Fuzzy control has been applied to various fields because it is easy to describe experience and knowledge of human directly. This paper deals with a fuzzy control for a fluid temperature system using optimized shapes of membership functions of fuzzy production rules. The shapes of membership functions are optimized through the learning based on the genetic algorithm. The effectiveness of the fuzzy control with the optimized shapes is confirmed by an experiment of a fluid temperature system.

View full abstract

This paper presents the latest results of running tests of a mobile vehicle which was devised in Prof. Sugisaka's laboratory in Japan in details. A FUZZY-PID controller was used in the control system of the mobile vehicle. This kind of controller means that the proportional gain k_p, differential gain k_d, integral gain k_i in a PID controller are adjusted continuously by fuzzy control method

during the running of the mobile vehicle. The correct running action which run along with a red type on the floor was expected and successful experiments had been made.

View full abstract

Statistical distribution of local expansion ratios and local q-order Lyapunov exponents are employed as indices for diagnostic techniques. These indices characterize change in statistical properties of dynamical systems that have internal disturbance for the systems' parameters; they give a foothold for detailed analysis of engineering systems.

View full abstract

A computational scheme is presented for detecting invariant feature associated with self-similar patterns. The patterns to be detected are assumed to be generated through successive application of a set of contraction mappings. By identifying the mappings with generated attractors, a conditional probability is represented on a diffusion system that yields a discrete invariant set. This invariant set can be expanded to generate associative attractors. The detection scheme is verified through computer simulation.

View full abstract

In this paper, one-dimensional chaotic difference equation, described by unimodal functions, are investigated, It is shown that folding or expanding of mapping generate the chaotic system with the invariant distribution. New approaches, demonstrated here, simplify the Frobenius-Perron equation.

View full abstract

An identification problem is considered for stable rational minimum-phase transfer functions without a priori knowledge of either numerator or denominator degrees. Recently, a new technique for ARMA-system identification has been proposed by making use of the powerful concepts of positive functions and bounded functions. We examine the ARMA-system identification method and propose a new criterion for determination of ARMA model order. It is shown that the developed criterion is more simple and practical by simulation experiments.

View full abstract

A vector autoregressive model(called VAR model) is constructed to investigate the behavior of GNP in Japan which is influenced by several important economic variables and real economic activity. The system model is expressed by
x(t)＋A₀x(t)＋A₁x(t－1)＋…＋A_kx(t－k) ＝ε(t)
where x is a vector with 5 variables; call rate, real money(m/p), money, real GNP and exchange rate. The unit root and cointegration are checked in association with stationarity of the system. In addition, candidates for variables in a model are checked by Granger's causality test. Responses of the each element of the x-process to shocks ε are considered under the condition such that A₀ is a non-zero matrix. In order to identify the matrix A₀ , structural restrictions are imposed on A₀ from the economic point of view. Such a system is called ”a structural VAR model”. Impulse responses are shown from the identified model.

View full abstract

The purpose of this study is to develop a numerical model to simulate the behavior of the head potential fluctuation in subsurface flow. Numerical and analytical solutions are presented. Impacts of the storativity and porosity on the head potential fluctuation are investigated in this paper. It is found, in one-dimensional subsurface flow, that the less the storativity, the easier the fluctuation of the head potential propagates. In addition, an asymptotic behavior of the head fluctuation does not reach until the head propagation travels more than 0.75 times the total aquifer length.

View full abstract

A new model based on time-varying peak frequencies of power spectrum (TVPP model) has been proposed. The model contains peak frequencies of power spectrum as time-varying parameters and time invariant factor of power spectrum. Nonlinear nonstationary state space modeling proposed by G.Kitagawa is used to estimate time-varying peaks, and quasi-Newton method with BFGS modification formula is used to estimate time invariant parameters. Application result to seismic wave data has been reported. Comparison to time-varying coefficient AR model (TVAR mode) has also been reported.

View full abstract

For state-space digital filter applying distributed arithmetic instead of an ordinary multiplier, the round off error is different from the result of ordinary analysis and is lost high accuracy in nature. In this paper, the relation of SNR and word length is shown by some examples and new analysis is presented. From the result of analysis, the noise whose characteristics and source are unknown exists in state-space digital filter with distributed arithmetic.

View full abstract

This paper presents an analysis of a subband adaptive filter structure using noncausal Wiener filter theory and multirate system concepts. The transfer function of the overall system is close to that of the fullband optimal Wiener filter whereas that of the conventional subband filter contains a delay of the filter bank.

View full abstract

A computational method of the formal linearization using a piecewise cubic Hermite interpolation is proposed. We introduce a linearizing function that consists of the state variables, their squares, and the cubes. The nonlinear terms are approximated by the method of the piecewise cubic Hermite interpolation and linearized with respect to the linearizing function. Numerical computations are easily carried out with the aid of computers. A nonlinear filter is synthesized as an application of this method and is verified through numerical examples.

View full abstract

What an artificial brain would do to determine optimal solutions and make decisions has been worked out over recent decades. Some of the work is reviewed and a model of an artificial brain is presented. Optimization and ecisionmaking functions are discussed.

View full abstract

In the present paper, it is argued how to deal with analytical and computational solutions of searchlight problem on flat terrain bounded below by a diffuse reflector. It is shown that the angular distribution of radiation emergent from a point source on top of the turbid plane-parallel atmosphere of finite thickness is discussed by embedding the problem within a family of parameters.

View full abstract

In this paper, a variable structure fuzzy logic controller(VSFLC) is proposed for the control of robot manipulators. Based on the variable structure system(VSS) theory, two fuzzy control schemes are presented: first, the simple VSFLC; second, VSFLC based on the knowledge for the upper bound of system uncertainties. The proposed scheme always guarantees the robustness in the presence of parameter uncertainties and external disturbances. In other words, the system output shows the almost same performance/trajectory for the same initial condition. In addition, since the proposed scheme uses concept of a sliding mode control, it can be shown that the system error is globally uniformly ultimately bounded (g.u.u.b.). The efficiency of the proposed method has been demonstrated by simulations for the control of a two-link robot manipulator subject to parameter uncertainties and external disturbances.

View full abstract

This paper analyzed the drawback of classical statistical discrimination method and clustering method. Also, it showed two layer fuzzy classification method with validity measure. Then, the conception of augmented variable was proposed and a method of classification and creating Hidden Markov Model based on the augmented variable was presented, two applications based on the method mentioned in this paper was introduced.

View full abstract

In this report, we propose a theoretical essence of the stabilization of flexible structures by using the static output feedback control.This shows that not only strain outputs but also other outputs are valid in this control method for flexible structures of many boundary conditions including cantilevers, free ends, and other general boundary conditions. Sufficient conditions of structures and observation mechanisms are obtained for the closed loop systems to be asymptotically stable. When the observation mechanism is nicely chosen, the closed loop system becomes exponentialy stable.

View full abstract

We study a model of stochastic hyperbolic equation with free boundary. This model is motivated by a micro-tunnelling machine which is used for the conduit instollation for communication cables. By using this tunnelling machine , a flexible pilot pipe is thrusted into a ground and hence the total length of the inserted pipe becomes a function of time. Furthermore, the thrust force generated by a hydraulic pum is stochastically perturbed. The transverse displacement of the inserted pipe is firstly shown to be governed by a hyperbolic equation with free boundary by using the Hamilton principle. Secondly, the existence of a solution of this equation is studied and two digital simulation examples are demonstrated for checking the buckling phenomena to the inserted pipe.

View full abstract

As a typical example of a nonlinear problem for a distributed parameter system, a free boundary problem is considered. The free boundary problem appears in many fields of industry. For example, the solidification problem in a steel industry is a very important problem. In the solidification problem, the interface between the liquid steel and solid one becomes the free boundary. In this paper, focussing our attention to the free boundary problem in the electron beam processing, the state and the free boundary estimation problem is considered under the noisy observation.

View full abstract

Sophisticated procedures and fast analyzing method have been needed for analyzing the stop consonants. Because the features of the stop consonants are involved in the short transitional portion between the consonant and the vowel segments. An adaptive signal processing algorithm only needs a simple procedure. Moreover, a fast converging adaptive algorithm is obtained by utilizing the ASA method which is improved by authors. The estimated parameters are converted into the critical band spectra which are useful for recognizing speech. And, the critical band spectra are converted into dynamic critical band spectra with time-varying characteristics. These spectra are inputted into a three-layered neural network for recognizing the stop consonants. In this paper, an improving method of the learning properties of the neural network is also described.

View full abstract

We consider the optimal transmission problem for a set of correlated Gaussian signals which is sent through parallel channels with feedback. The signals are assumed to be given by a solution of a multidimensional linear stochastic differential equation. Under a constraint on the total power of the signals, we will obtain the optimal gains for the set of channels which minimizes the steady-state estimation error.

View full abstract

P300 is one of the most prominent responses related to recognition of certain stimulation, and appears around 300 [ms] after stimulus in the time series recorded from a human scalp. The peak latency of P300 for each stimulus varies according to the recording conditions. It is important to know whether the variability is caused by physiological factors (physiological variability) or only by the noise in the background EEG (noise variability), when the variability in the features of the waveform is detected. We proposed a method for judging whether the physiological variability exists or not, and if exists extracting the physiological variability of the peak latency of the P300 from the measured variability based on the probability density functions derived in this study. The proposed method was evaluated by using simulation data, and was applied to actual ERP data of 12 healthy subjects.

View full abstract

To construct the model of reinforcement learning systems, we presented a theoretic model of stochastic learning cellular automata (SLCA) in our previous paper. The SLCA is an extended model of traditional cellular automaton, defined as a stochastic cellular automaton with its random environment. There are three rule spaces for the SLCA: parallel , sequential and mixture. This paper suggests a parallel SLCA with a genetic operator and applies it to the combinatorial optimization problems. The computer simulations of graph partition problem show that the convergence of SLCA is better than parallel mean field algorithm.

View full abstract

In high-capacity cellular radio communication systems, efficient channel use is achieved by covering the service area with large numbers of small cells. In such environments, a cell boundary crossing by moving users can generate many hand-off attempts. To reduce the forced termination of calls in progress, this paper considers such a priority scheme that some channels are reserved for hand-off calls and a finite buffer-capacity of N is offered to hand-off calls. We present an exact analysis and an efficient matrix-analytic procedure to numerically evaluate the performance of the system such as blocking probability, channel utilization, queue length and delay. Numerical results show how priority can be provided to hand-off calls according to the number of reserved channels and buffer size.

View full abstract

The corrected least squares (CLS) approach using an overdetermined model is investigated to decide the number of sinusoids in additive white noise. The CLS estimation is different from the ordinary LS and the TLS approaches in that the noise variance is subtracted from the diagonal elements of the correlation matrix of the noisy observed signal data. Therefore the inversion of the matrix becomes ill-conditioned and then adequate truncation of the eigenvalue decomposition (EVD) should be done, since the rank of the matrix is equal to the number of sinusoids. Thus, in the overdetermined CLS approach, the estimation of the noise variance and the truncation of eigenvalues are mutually dependent, therefore it is required to decide them simultaneously. By introducing a multiple number of regularization parameters and determining them to minimize an MSE of the model parameters, we can give an optimal scheme for the truncation of eigenvalues to detect the number of sinusoids. Further, an iterative algorithm using only observed data is also given to determine the noise variance and the truncation number simultaneously.

View full abstract

In the field of medical X rays image, it is important how to seize a quantity of being affected by radioactivity to the human body. Then, in the case when dropping affection by radioactivity dose, quantum noise increases on inevitability, and some counter-measure methods become central problem. In this paper, a new trial of wide-sense digital filter for estimating adaptively the non-Gaussian stochastic systems will be proposed in the hierarchical form based on the orthogonal series expanded Bayes' theorem. Then, the effectiveness of proposed recursive algorithm will be experimentally confirmed through several applications to real medical X-ray images.

View full abstract

Position emission tomography(PET) is a medical imaging system. It visualize an image of the density of tracer, that is a radioactive element injected into patient, with the noisy counting data of the emitted photons. This image reconstruction in PET can be considered as an estimation problem of the number of photons emitted from each pixel, and maximum likelihood(ML) estimation and Bayesian estimation are applied for PET. In this paper, we discuss the Bayesian estimation approach for image reconstruction of the PET and examine the effect of several priors including the Wiener filter type prior. Monte Carlo simulation results indicate that the Wiener filter type prior provide better image.

View full abstract

In this paper, we introduce a parameter estimation method of the Markov Random Field (MRF), and also present some new energy functions associated with Gibbs Distribution. In applying a MRF as a statistical image model, we can describe the probability density function of image gray-levels as a Gibbs distribution. Its exponential part, namely, the energy function affects the statistical property of the image. In most cases the energy function is defined without being considered an effect caused by the difference of gray-levels between neighboring pixels. In this paper, we propose new energy functions which consist of the difference of gray-levels in the neighborhood and simultaneously present an estimation method of the unknown parameters comprised in the energy function.

View full abstract

This paper considers a recursive optical flow approximating approach to recognizing the shape of a three-dimensional object. After the object model is derived here under the assumption that the shape of the object model is not intricate, a shape-recognition algorithm is demonstrated within the framework of the gradient method. Finally, the convergence of the algorithm is shown, considering the cost function.

View full abstract