Let B ＝ (B_t)_t≥0 be a standard Brownian motion started at zero, and let μ ∈ R be a given and fixed constant. Set B^μ_t ＝ B_t +μt and S^μ_t ＝ max_0≤s≤t B^μ_s for t ≥ 0. Then the process:

(x ∨ S^μ) － B^μ ＝ ((x ∨ S^μ_t) － B^μ_t)_t≥0

realizes an explicit construction of the reflecting Brownian motion with drift －μ started at x in R₊. Moreover, if the latter process is denoted by Z^X ＝ (Z^x_t)_t≥0, then the classic Lévy's theorem extends as follows:

((x ∨ S^μ) － B^μ, (x ∨ S^μ) － x ) ＝^law (Z^x, l⁰(Z^x))

where l⁰(Z^x) is the local time of Z^x at O. The Markovian argument for (x ∨ S^μ) － B^μ remains valid for any other process with stationary independent increments in place of B^μ. This naturally leads to a class of Markov processes which are referred to as reflecting Lévy processes. A point of view which both unifies and complements various approaches to these processes is provided by the extended Skorohod lemma.

View full abstract

We consider discrete-time linear stochastic systems with unknown disturbances and study a smoothing problem for those systems. We derive a smoothing algorithm by applying the optimal filter obtained in our previous work to an augmented system.

View full abstract

An estimation problem of the water quality of a river is considered. The BOD and DO are regarded as two major quantities to investigate the water quality. First, based on the couple of steady-state transport equations for substance concentrations, the stable water quality equations are derived, and then its state space representation is obtained. By formulating the observation process from which the observation data on the BOD and DO are measured, the Kalman filter is derived. Finally, simulation studies are presented.

View full abstract

We consider the problem of optimal transmission of correlated Gaussian signals through a set of parallel channels with feedback. We send the signals using a linear encoding with output feedback. It is well-known that the optimal output feedback which minimizes the power of the encoded signal is given by the least-squares estimate of the linear term. Under a constraint on the total transmission power, we consider the problem of computing a set of gains for the channels which maximizes the mutual information between the output and the signal.

View full abstract

In this paper we present the algorithms of land-vehicle INS (Inertial Navigation System) / DGPS (Differential Global Positioning System) In-Motion Alignment based on nonlinear filtering techniques (Quasi-Linear Optimal filter [1], Gaussian Sum filter [2], the Extended Kalman filter [3] and Monte Carlo filter [4]). We also show results of comparative numerical experiments, and evaluate the nonlinear filtering performances under various error sources such as the errors of GPS, accelerometers and gyros.

View full abstract

This paper proposes a PID parameter design scheme using Taguchi's robust design method. This scheme is then applied to a PID thermal control system under the influence of atmospheric temperature. Experiment with a cooking machine shows the effectiveness of the proposed scheme.

View full abstract

This paper considers a problem of designing optimal smoothing spline surfaces employing normalized uniform B-splines as basis functions. Assuming that the data is obtained by sampling some surface with noises, an expression for optimal smoothing surfaces is derived when the number of data becomes infinity. Then, under certain condition, we present the convergent properties of optimal smoothing spline surface. Moreover, they are extended to the case of periodic spline surfaces. The results are applied to dynamic contour modeling problem of living bodies, and the effectiveness is examined by numerical and experimental studies.

View full abstract

In this paper we are describing the first steps toward a new approach for processing various types of information that can be collected by moving object's (e.g. car) sensors. These moving cars are capable of communicating directly with each other and with the fixed wireless infrastructure. Furthermore, each car is assumed to have a GPS device and a camera, and can take snapshots of visual information, collect data about its location, speed etc. In this system, not only a single car can be a source of information, but also a set of cars can collaboratively generate new data about the surrounding conditions (or alternatively several sensors in the car can generate new data) [12, 11]. As the first step, we will focus on the image processing from a single camera. The 3D model is aimed to be restored by using the communication of wireless infrastructure.

View full abstract

In this paper, a clustering method based on an improved competitive process is proposed to segment the entire circumferential range data significantly. The segmentation technique is utilized as the preprocessing of 3-D shape modeling so that the modeling can be more easily achieved for the object that has arbitrary topology, in which the data points are divided into the several regions that represent the 3-D shapes of different quadric surfaces. The clustering method is implemented by evaluating a distance computed between each data point and each quadric surface. Furthermore, it consists of the following two stages. First, clusters are created on demand by using random sampling techniques. However, since a number of redundant clusters may be created, the dispensable clusters are secondly vanished by the competitive process that is realized by an optimal iteration time determined using MML criterion ;accordingly the remaining clusters are gradually agglomerated. Consequently, since the only appropriate clusters are remaining, the segmentation can be achieved by assigning the data points to these clusters.

View full abstract

Image restoration and enhancement are ones of fundamental problems in image processing. Many research activities, therefore, have been continuously carried out during almost three decades [1, 2, 3]. Inverse filtering, Wiener filtering with or without constrains, and Kalman filtering-smoothing techniques have been applied to the image restoration problem in the cases of either known or unknown blurring functions. Many methods in image restoration and enhancement rely upon the Fourier transformation and its discrete version; DFT (discrete Fourier transformation).

In this paper, we emphasize the usefulness of the DST (discrete Sine transformation) instead of DFT in case of image restoration and enhancement for many cases such as inverse filtering, Wiener filtering and blind deconvolution [4]. Especially, we concentrate here to design 2D (two dimensional) inverse filters for blurred images without observation noises. The main purpose of the present study is to provide the images to the presbyopic as well as the myopia persons based on inverse filtering.

View full abstract

Many kinds of spatio-temporal patterns appear in various fields of engineering including chemical and biological engineering. Among such spatio-temporal patterns, analyses of patterns created by the so-called self-organization are very important as basic problems in engineering, mainly because analyses of the spatio-temporal patterns in phase transitions of polymeric materials are essential to develop new materials and the analyses of epidermal patterns of animals and seashell patterns shed light on mystery in biology. In this paper, focusing on the chemotactic bacterial colony patterns as the spatio-temporal patterns created by the self-organization, we study the influence of the random disturbance on the colony formations.

View full abstract

This paper proposes a method for the blind deconvolution of all-zero mixed phase systems (also called mixed phase MA filters). A criterion based on higher order statistics is used while assuming non-Gaussianity of the unknown input signal. The inverse system used to perform deconvolution is not a finite impulse response (FIR) filter but a two-sided infinite impulse response (IIR) filter. The Z-transform ‘time-reversal’ property is used for developing an iterative blind deconvolution algorithm. An adaptation rule for the non causal part of the two-sided IIR filter is derived in the time-reversed case. An initialization scheme is also studied in order to avoid being trapped in local maxima of the criterion.

View full abstract

In this paper, we develop a subspace system identification algorithm for the Errors-In-Variables (EIV) model subject to observation noise with outliers. By using the Minimum-Covariance-Determinant (MCD), we identify and delete the outliers, and then apply the classical EIV subspace system identification algorithms to get state space model. In order to solve the MCD problem for the EIV model we propose the random search algorithm. In addition, we show that the problem of detecting the outliers in the closed loop systems is especial case of the EIV model. The proposed algorithm has been applied to heat exchanger data.

View full abstract

In this paper we discuss time series analysis of data measured by near infrared spectroscopy (NIRS, in short). NIRS is well known for its ability to observe the brain activity by means of three measures of hemoglobin density, oxidized hemoglobin (HbO), reconstructed hemoglobin (HbR) and total hemoglobin (HbT). We repeat each experiment several times to focus on how the subjects' experience of the tests affect their brain activity. We examine the results using three kinds of data analysis, (1) transition of time series data, (2) calculating correlation matrix of the data, and(3) estimation by a multivariable auto-regressive (MAR) model for the data.

View full abstract

This paper deals with a quasi-ARX modeling approach to nonlinear black-box systems. A quasi-ARX model consists of two parts: The first part is a macro-model, which is a user-friendly interface constructed using application specific knowledge and the nature of network structure; The second part is an ordinary neurofuzzy network, which is used to parameterize the coefficients. A dimensionality reduction technique based on principal component analysis is introduced to improve the quasi-ARX modeling. The modeling and the parameter estimation are described in details. Numerical simulations are carried out to demonstrate the effectiveness of the proposed modeling approach.

View full abstract

This paper presents two novel radial basis functions and their comparison. Both radial basis functions are based on an idea of Support Vector Machine (SVM) by mapping data into a high dimensional feature space, which is known as Reproducing Kernel Hilbert Space and then performing Radial Basis Function (RBF) network in the feature space. Orthogonal Least Squares (OLS) method is employed to select a suitable set of centers (regressors) from a large set of candidates in order to obtain a sparse regression model in the feature space. The proposed method is applied to a nonlinear system identification problem by simulations.

View full abstract

It is well-known that the least squares (LS) method gives biased estimates for infinite impulse response (IIR) model in the presence of output noise [1]. One of the methods which delivers consistent estimates is the total least squares (TLS) method. In this paper, two types of the recursive TLS algorithms for noisy IIR estimation are discussed.

View full abstract

In this paper, we present an identification algorithm for continuous-time Multiple-Input Multiple-Output (MIMO) state-space models and also for determining the order of models from the samples of the input-output data. In the proposed algorithm, from the sampled data first an equivalent discrete-model is identified, then the model is transformed to the corresponding continuous-time model. The parametric discrete-time canonical-formed MIMO state-space model is identified based on Maximum-Likelihood (ML) and Akaike's Information Criterion(AIC). For obtaining the maximum-likelihood estimates of the model parameters, we apply Expectation-Maximization(EM) algorithms which are iterative methods such that the choice of the initial estimates is most important. The initial estimates of parameters in canonical-formed state-space models are obtained by MOESP[1] or N4SID[2] methods where the similarity transformation plays a key role.

View full abstract

Recursive updating algorithms of error covariance matrices in subspace identification methods for time-varying systems are derived. The proposed algorithms can be applied to estimate the system parameters which are slowly time-varying. The algorithms are based on the fact that the subspace extraction amounts to computing singular value decomposition of the SC of the input submatrix in data product moments and the SC can be interpreted as the least squares residuals. We have proposed an unified framework for the MOESP type of the subspace identification method by using the SC. In this paper, we show the aforementioned time-varying case can be also treated in the proposed framework.

View full abstract

Radial basis function (RBF) network's structure determination method by using multiobjective evolutionary algorithms is considered and an ensemble network constructed by the Pareto optimal networks is also proposed in this paper. The candidates of RBF network structure are encoded into the binary coded chromosomes. Then, they evolve toward Pareto optimal front which is defined by the several objective functions corresponding to the model accuracy and model complexity. The application to the nonlinear system identification is studied by numerical simulations, the applicability of the proposed approach is discussed.

View full abstract

In this paper, the author will investigate a class of fuzzy random vectors, where FRVCs are considered as vague but smart perceptions of some complex random phenomena.

First, based on the result previously proposed by the author[1], and inspired by the recent researches, especially Krätschmer[2], the definition of FRVCs are reconsidered form the viewpoint of vague but concise description of the state of very complex random phenomena.

Secondly, the expectation of FRVCs are derived from the viewpoint of mult-valued logic approach.

View full abstract

In this paper, an assist method for human's operation of electric-powered wheelchairs is developed. The purpose of this research is to make powered wheelchair intelligent and to develop a mobility aid for people, who find it difficult or impossible to drive a conventional wheelchair. On a prototype of our group, a neural network produces an obstacle avoidance function. In this research, by the approach that connection weights of the neural network vary according to the condition of obstacles in the neighborhood of the wheelchair and the running state of wheelchair, we improve the obstacle avoidance function First, neural networks evolve by using digital computer simulator. Secondly, experiments, using a prototype with six wheels implemented neural networks whose connection weights are given by numerical studies, demonstrate that the neural network with variable connection weights exhibits the more excellent ability of obstacle avoidance.

View full abstract

We develop the real-time speech visualization system called “KanNon”[1, 2] which supports speech communication of deaf people. The KanNon system presents several information of the speech such as loudness, pitch, sound spectrogram and characters by speech recognition system in real-time. In the present system, we are adapting a word unit speech recognition system using large-scale dictionary. However the KanNon system is required quick and simple display of speech contents for smooth communication. For this purpose, we apply phonemic speech recognition system for Japanese 5 vowels using “Time-Delay Neural Network (TDNN)”. Further, we developed speech detection, voiced/unvoiced (v/uv) detection and change detection algorithms in the KanNon system. Finally, we show experimental results using real speech data.

View full abstract

In this paper, the revised Group Method of Data Handling (GMDH)-type neural network algorithm with a feedback loop identifying sigmoid function neural network is proposed. In this algorithm, the optimum sigmoid function neural network architecture is automatically organized so as to minimize the prediction error criterion defined as Akaike's Information criterion (AIC) or Prediction Sum of Squares (PSS) by using the heuristic self-organization. The structural parameters such as the number of neurons in each layer, the number of feedback loops and the useful input variables are automatically determined by using AIC or PSS criterion. Therefore, it is easy to apply this algorithm to the identification problem of the complex nonlinear system and to obtain a good prediction results.

View full abstract

In this paper, the revised Group Method of Data Handling (GMDH)-type neural network algorithm with a feedback loop identifying sigmoid function neural network is applied to the medical image recognition of the brain. This revised GMDH-type neural network automatically selects the structural parameters such as the number of neurons in each layer, the number of feedback loops and the useful input variables by using Akaike's Information Criterion (AIC) or Prediction Sum of Squares (PSS) criterion. It is shown that this revised GMDH-type neural network is a very useful method for the medical image recognition because the neural network architecture is automatically organized so as th minimize AIC or PSS criterion.

View full abstract

This paper discusses the visual perception of a partner robot. The partner robot should classify not only a facing human, but also the expression of the human. Therefore, various features should be extracted from the image of the human. In this paper, we apply a cellular neural network to extract various features from the sequential images. Experimental results will show that proposed method can extract effective features to classify the human behavior from the original image.

View full abstract

Flexible manipulators in working space with environmental objects (walls, boxes, humans, other mobile machines, etc.) should avoid collision to achieve their mission. Unfortunately, if an unexpected obstacle collides with the manipulator, it is necessary to stop and replan the path for avoiding the obstacle when the collision is detected. However, in order to replan new path to avoid the obstacle, we have to know the flexible manipulator's spot where the obstacle collides. This paper presents a method to find a collision spot of a flexible beam based on the observation data measured by strain sensors pasted on it.

View full abstract

This paper presents a new method for combinatorial optimization problems. Most of the actual problems that have discrete structure can be formulated as combinatorial optimization problems. It is experientially known that Proximate Optimality Principle (POP) holds in most of the actual combinatorial optimization problems. The concept of Proximate Optimality Principle says that good solutions of most real combinatorial optimization problems have the structural similarity in parts of solution. In this paper we propose a new optimization method based on Tabu Search. In the proposed algorithm, POP is taken into consideration. The proposed algorithm is applied to some knapsack problems and traveling salesman problems, which are typical combinatorial optimization problems in order to verify the performance of the proposed algorithm.

View full abstract

By modeling charge-carrier balancing in ionosphere as a complex dynamixs, a method is presented for in-situ estimation of GPS residual. Under the constraint of Volterra's principle, the model is adapted to reduce the positioning residual. Adaptation scheme was verified to restore positioning rersidual through experiments.

View full abstract

Electroencephalograph (EEG) recordings during right and left motor imagery can be used to move a cursor to a target on a computer screen. Such an EEG-based brain computer interface (BCI) can provide a new communication channel to replace an impaired motor function. It can be used by e.g., handicap users with amyotrophic lateral sclerosis (ALS). In BCI system, the miss-recognition of subject's will causes the dangerous accident. Therefore, Error detection is necessary in order to avoid miss-operation of the machine. In this study, statistical pattern recognition method based on AR model was introduced to discriminate the EEG signals recorded during right and left motor imagery. Next, pattern recognition and spectrum analysis of EEG signals in the correct-recognition and miss-recognition were investigated in order to construct a detection system of miss-recognition. Finally, the possibility of Error detection was confirmed.

View full abstract

The Particle Swarm Optimization method is one of the most powerful optimization methods available for solving global optimization problems. However, knowledge of autonomous and adaptive strategies for tuning the parameters of the method for application to large-scale nonlinear non-convex optimization problems is as yet limited. This paper describes an autonomous and adaptive strategy for tuning the parameters of the PSO method based on some numerical analysis of the behavior of PSO. The proposed autonomous and adaptive tuning strategy is based on self-tuning of the parameters of PSO, which strategy utilize the information about the frequency of an updated global best of a swarm. The feasibility and advantages of the proposed autonomous adaptive PSO algorithm are demonstrated through some numerical simulations using three typical global optimization test problems.

View full abstract

This paper proposes an adaptive Particle Swarm Optimization (PSO) algorithm using the information defined as the average absolute value of velocity of all of the particles, which information can be used as an index to understand the activity of all of the particles. While a stability analysis of PSO algorithm is carried out based on the stability theory of modern control theory, an adaptive strategy for tuning one of its parameters is introduced so as to follow a given ideal average velocity by feedback control. The feasibility and advantages of the proposed adaptive PSO algorithm are verified through numerical simulations using some typical global optimization problems.

View full abstract

In this paper, the effects of the global influence with random element are examined. We consider two types of agents, pioneers and followers, in a two-dimensional grid world. Agents decide their attitudes by local influences (effects of the neighbors) and global influences (effects of mass-medias), then change their attitudes. So as to consider the variance of effects which global influences give agents, we propose the global influence with a random element. Simulation studies show that the results of agent-based simulations are very sensitive to the global influence including the random effect.

View full abstract

This paper analyzes the effects of first-degree stochastic dominance (FSD) changes in multiplicative background risk on the risk-taking attitude of a decision maker. First, we consider contractive FSD changes in multiplicative background risk and analyze the effect of these changes. Then we consider general FSD change in multiplicative background risk. Also, within the context of coinsurance, we examine the effects of simple FSD changes and monotone likelihood ratio (MLR) changes in multiplicative background risk.

View full abstract

We consider the dynamics of forward rate process which is modeled by the parabolic type infinite-dimensional factor model with stochastic volatility. The parameters included in this parabolic model are estimated by using the yield curve as the observation data. In this paper, we propose the filtering and identification method for the parabolic type factor model by using the maximum likelihood technique.

View full abstract

This paper studies a model of capital adjustment cost and investment behavior of a competitive firm. Our goal is to examine the effects of uncertainty on the optimal rate of investment. The framework for this analysis is a stochastic version of the dynamic factor demand model. The result is induced from the data of 9 privately owned Japanese gas firms for the period 1981-1995. The model with the Euler-Lagrange equation of the stochastic version shows reasonable results. It means that the stochastic style of the dynamic factor demand model is successfully estimated.

View full abstract

Financial anxieties are modeled as psychological change of people due to financial shocks. Financial position (easy or tight) is regressed by interest rate of lending (rise.or fall) and the state of ‘’tight” position under the ‘’fall” interest rate is regarded as financial anxieties. Such anxiety is quantified by conditional variance of TARCH model. Precautionary demand is estimated as a function of financial anxieties in VEC model and the long-run equilibrium relationship of related monetary system of (money, real GDP, interest rate) is analyzed.

View full abstract

In recent years, according to very rapid increase of various kinds of information and communication equipment like personal computers and portable radio transmitters, not only a problem of the technical characteristics of equipment and propagation but also a problem of the influence to the environment including human body has become gradually important. In this study, some evaluation and/or measurement methods for compound effect based on the inter-subjective relationships between sound (with other environmental factors) and psychological or physiological factors have been proposed, especially through some examples of the physical factors (sound, lights, magnetic field and temperature) and the psychological or physiological factors (noise evaluation values, mean blood pressure and pulsation) surrounding an indoor electrification environment. Finally, with an application to actually measured data, a part of the effectiveness of the proposed method has been also experimentally confirmed.

View full abstract

In order to keep reliability of industrial systems, maintenance of system is required. Through a maintenance plan, it is expected not only to keep the system reliability but also to reduce the total cost. In this research, maintenance scheduling method is proposed. The method reduces the total number of maintenance by flattening element reliability while keeping the current system reliability level. To obtain the optimal maintenance scheduling, we assume that maintenance improve the system reliability. In order to satisfy the assumption, the segmentation of the maintenance interval is proposed. By using the proposed segmentation of the maintenance interval, the monotone increase relationship between the maintenance interval and the failure rate can be fulfilled. The proposed optimal maintenance interval is applied to the actual data of equipment in thermal power stations and it gives satisfactory results.

View full abstract

The resolution of strain measurements using Brillouin scattering in an optical fiber depends on the pulse width of a pump laser pulse. It was considered that there was a lower limit for pulse width, since the line width of the Brillouin spectrum was broader for shorter pulses. This lower limit is about 10 ns and corresponds to a spatial resolution of 1 m. However, on the contrary, Bao et al. recently discovered a phenomenon in which the line width of the spectrum decreased for a very short pulse with a width of 1 ns when there was light leakage. In this paper, we analyze this phenomenon by solving coupled wave equations using a perturbation method, and clarify the cause of the phenomenon. We then propose an improved Brillouin measurement method using a two-step pulse inspired by the phenomenon and show that its spatial resolution is about 10 cm, which is one digit smaller than the present lower limit.

View full abstract

For the state estimation of acoustic field in a reverberation room fluctuating randomly in a non-Gaussian distribution form under the contamination of the background noise, some wide sense digital filter based on Bayes' theorem and a stochastic type Hermite series expansion form of the probability distribution is newly proposed. The proposed method is experimentally confirmed by applying it to the actual reverberation room and its result is compared with the one using the Kalman filter.

View full abstract

Frequency domain ICA (Independent Component Analysis) has a scaling indeterminacy and a permutation problem. The scaling indeterminacy can be solved by use of a decomposed spectrum. For the permutation problem, we have proposed the rules in terms of gain ratio and phase difference derived from the decomposed spectra and the source's coarse directions.The present paper experimentally clarifies that the gain ratio and the phase difference work effectively in a real environment but their performance depends on frequency bands, a microphone-space and a source-microphone distance. From these facts it is seen that it is difficult to attain a perfect solution for the permutation problem in a real environment only by either the gain ratio or the phase difference. For the perfect solution, it is important to devise the correction scheme, taking the frequency bands, the microphone-space and the source-microphone distance into the consideration.

View full abstract

An acoustical estimator for gas temperature distributions is proposed. It is based on the MAP (Maximum A Posteriori) estimation with local iterations that is referring acoustic time-of-flight through non-linear observation process. The system is also featured a matched filter with complex absolute detection that is free from unknown phase shift. In addition, for fast measurement, the Gold sequence P.R.K. (Phase Reversal Keying), which has sharp autocorrelations and low cross-correlations, is applied to the acoustic signal for simultaneous time-of-flight identifications on paths. It has been shown to effectively estimate the temperature distributions in a gymnasium and outdoor fields. The estimator has been applied to evaluation of watering effect.

View full abstract

A method of modeling the seismic waves is proposed using the chirplet-based approximation incorporating with the spectral representation of the real earthquake-induced ground motion. Detail algorithm for the chirplet-based approximation is presented and three typical earthquake data are modeled. Furthermore, the seismic waves are generated artificially using the spectral representation with evolutionary power spectrum due to Priestley.

View full abstract

Exact convergence analysis of the RLS and LMS algorithms in adaptive filtering is presented for the case of sinusoidal signal cancellation without the persistently exciting condition. This situation occurs when the number of tap coefficients of adaptive filter exceeds that of the complex sinusoids in the input signal. The convergent point of both algorithms is shown to be the one determined by the pseudo inverse of the deterministic covariance matrix. The convergence proof for the LMS algorithm is based on the Lyapunov function method.

View full abstract