In this paper we show that the concept of a closed and stable population has inherent difficulties. We use a simple Markov model based on a variant of the Polya Urn to show that genetic diversity is not a reasonable expectation in such a setting.

View full abstract

This research is aiming at creation a visual supplement system. In the current researches of automatic extraction system of road traffic signs, the noise due to the variation of illumination condition, and the obscure image due to shielding affected negatively about the correct extraction of road traffic signs. In this research, robust characters were applied in the extraction of road traffic signs, adding with the new alternative method, to improve the road traffic sign recognition system. Audio warning and visual expression were added to remind the driver for a better warning.

View full abstract

Omnidirection camera is a special camera having 360 degrees view. Gripping the camera with fingers, and using the camera image with finger motion for human machine interface is proposed in [1]. In this paper, we track the motion of fingers in the image of the same setting in [1] by using a state space model, which represents state and image observation of the fingers. State estimation is conducted by CONDENSATION algorithm[3], which is one realization of particle filters[2] maintaining posterior distribution of the state by many samples (particles) according to the distribution, i.e. the distribution is approximated by degree of dense of the samples. Experiment of finger tracking has been conducted, and usability of the algorithm is demonstrated.

View full abstract

This paper discusses the image processing part in the hierarchical processing of traffic images of real-time. Since the camera is subject to pan and tilt, we first calibrate the camera position via image itself, and then extract the car imformation using stochastic model of the background. The moving objects like cars are treated as some occluding object. The processed result show a good performance.

View full abstract

We develop a 2-D image restoration and identification algorithm only using the blurred noisy images without any a priori knowledge of original images. A generalized semicausal stchastic image models are identified by applying the ARMA adaptive algorithm combined with the spectral factorization. Finally, we show experimentally the effectiveness of our proposed blind image restoration algorithm by using monochromatic real images.

View full abstract

An influence of a random noise on a mean curvature flow of a phase boundary is investigated. The phase boundary is defined as a curve or a surface separating different physical states such as a water-ice interface. The mean curvature flow is motion that the phase boundary moves with a normal velocity equals the mean curvature at each point on the phase boundary. Since we treat a stochastic mean curvature flow, we formulate it by a stochastic level set equation, which is a nonlinear stochastic partial differential equation with a uniques solution in a sense of a stochastic viscosity solution. In numerical simulations, behaviors of the stochastic mean curvature flows in two- and three-dimensional spaces are studied.

View full abstract

In the natural world, we often observe many kinds of spatio-temporal patterns such as patterns of animal coat, crystals of snow, etc. An analysis of generation processes of such patterns is very important as a basic problem in various fields of engineering including biological and material engineering. Among many spatio-temporal patterns, we especially study the pattern formation processes of sea shells from the viewpoint of biological engineering. As the mathematical model to generate the sea shells patterns, we consider an activator-inhibitor system, which is one of reaction diffusion systems. Taking into consideration the fact that some kind of disturbance exists to a greater or lesser degree in the natural world, we propose a stochastic activator-inhibitor system as the mathematical model of pattern formation processes of sea shells. We show that spatio-temporal patterns generated by the proposed model are robust for the random disturbance.

View full abstract

We derive a recursion formulae of transition probability of the noise-induced synchronization arising in a pair of identical uncoupled logistic maps linked by common noisy excitation only. The formulae has a delta-type stationary solution which represents the perfect synchronization with probability 1. The stationary solution maintains under chaotic bifurcation while the escape times to reach the perfect synchronization are increased in the chaotic region. A physical implementation of the synchronization is also demonstrated.

View full abstract

Networked control systems often suffer from what is called jitter, a transmission delay peculiar to the Internet. This is a phenomenon in which the delay time varies randomly due to network congestion, and is inevitable by nature. It is thus important to guarantee stability of the closed-loop system against jitter. This paper studies two types of design methods to this end. One is a phase margin guarantee method known in control theory, and the other is a packet sorting method in network architecture. The paper evaluates the effectiveness of these methods through experiment with an inverted pendulum.

View full abstract

In this paper, we develop a subspace system identification method for linear stochastic systems subject to observation noise with outliers, where the observation noise contains large values with a low probability. By using the leave-k-out diagnostics method, we detect the outliers and substitute the median of the output data for them. Then we apply the orthogonal decomposition (ORT) based method [3] to get state space models. A numerical example demonstrates the effectiveness of the proposed method.

View full abstract

A stochastic subspace identification method has been developed based on a “block LQ decomposition” [1] with the help of stochastic realization theories. The algorithm, however, does not guarantee that an estimate of the forward innovation representation is stable and of minimum phase. In order to overcome this difficulty, this paper discusses a method of estimating the spectral density function of the output process, and present a prototype of a subspace identification algorithm guaranteeing minimum phase via a spectral factorization technique.

View full abstract

Using a unified approach, recursive subspace identification algorithms are derived for the MOESP type of subspace methods. The proposed approach is based on the fact that the subspace extraction amounts to computing singular value decomposition of the Schur complement (SC) of the input submatrix in data product moments and the SC can be interpreted as the least squares residuals.

View full abstract

We present a new algorithm based on reinforcement learning for packet scheduling in routers with QoS requirements. In our approach, reinforcement learning is used to learn a scheduling policy in response to feedback from the network about the delay experienced by each traffic priority class. We construct a new traffic regulator with the stochastic learning automaton, which does not require prior knowledge of the statistics of each traffic flow and can adapt to changing traffic requirements and loads.

View full abstract

This manuscript introduces a new algorithm for shaping the traffic of the network. The traffic shaping method is effective for avoidance of the congestion of network. Our proposed algorithm employ Reinforcement Learning method with Ants Computing System scheme. Every router obtains the routing information from agents which called “Ants,” and uses it for updating the routing table. We show that our algorithm gives good proportion for distributing each traffic flow.

View full abstract

This paper deals with quasi-ecosystem and evolutionary learning method for agents in the quasi-ecosystem. The quasi-ecosystem is composed of plant and herbivore agents, which are in a relationship of parasitism. The selfish behavior which satisfies themselves leads to extinction because resource in the environment is finite. Evolutionary computation is adopted to update feed rules of herbivore agents. Extinction can be avoided by maintaining numerical balance between each species. The effectiveness of the proposal approach is demonstrated by computer simulations.

View full abstract

In this paper, a recursion of the subspace prediction (SP) algorithm which has been derived recently by the authors is investigated for time-invariant/varing stochastic systems. The proposed algorithm is tested by simulation experiments.

View full abstract

In this paper, we present identification algorithms for parametric canonical-formed Multiple-Input Multiple-Output (MIMO) state-space model and also for determining the order of models, based on maximum-likelihood and Akaike's Information Criteria (AIC). For obtaining maximum-likelihood estimate of 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 N4SID[3] or MOESP[2] methods where the similar transformation plays a key role.

View full abstract

The identification problem for a class of movements in the space by using the perspective observation, where the parameters are all time-varying, is considered in this paper. The motion equation can cover a wide class of practical movements in the space. The estimation of the position and the motion parameter are simultaneously developed in the proposed algorithm. First, the parameters relating to the rotation of the motion are identified, where only one camera is needed. Then the position of the moving object is identified, where the stereo vision is necessary. In the third step, the parameters relating to the straight movement are identified. The assumptions about the perspective system are reasonable, and the convergence conditions are intuitive and have apparently physical interpretations. The proposed method can cope with a much more general class of perspective systems, and is robust to measurement noises. Simulation results show the effectiveness of the proposed method.

View full abstract

In this paper, two bias-compensated least-squares (BCLS) methods (BCLS-α method and BCLS-β method) are proposed for identification of linear discrete-time system in the case where the input measurement is corrupted by white noise and the output measurement is corrupted by colored noise. It is well known that BCLS method is based on compensation of asymptotic bias on the least-squares (LS) estimate by making use of noise variances estimates. The main feature of proposed algorithms is to introduce an auxiliary multivariate estimator using filter (α-filter or β-filter) in order to estimate input noise variance and output noise covariances. Some simulation results indicate that the proposed methods provide good estimates.

View full abstract

Two methods are proposed for generating the random wind velocity whose power and cross spectral densities are specified to reflect the real world. The one is based on the linear system with white Gaussian noise processes as the seed of randomness; while the other is derived based on the spectral representation of a stationary process from the Wiener processes defined on the frequency domain. The windward pressures are calculated from the wind velocities through aerodynamic admittance. Sample runs of windward pressure are illustrated by simulation works.

View full abstract

In this paper we consider the parameter identification problem for the Parabolic type factor model by using the US treasury bond data. First interpolating the yield data, we can estimate the covariance kernel of the system noise. With the aid of this estimate, the modified maximum likelihood estimates of the unknown parameters are obtained for the hyperbolic and parabolic models. Finally, comparing the obtained results, we can show that the parabolic factor model works well.

View full abstract

In this paper we study a pricing problem of an exotic Forward contract. Unlike the standard Forward, the contract is not fair, and like an option, it is compensated by the premium. Using standard arguments in the Black-Scholes economy, an explicit formula for hedging as well as pricing is obtained. This is possible because of an exotic way of settlement, which is another focus of this paper. Contrary to our intuition, simpler ways of settlement do not necessarily imply a simpler formula.

View full abstract

In the present paper, we will study the pricing of Asian options. The main contribution is the model construction; our model is compatible with market conventions, can be calibrated to observable market data, and computationally tractable, while it is 2-factor and the interest rate is stochastic. The model takes it into account that the interest rate risks can influence the prices of the options in the long run. Thus our model can be used to analyse the term structure, from short to long term, of the prices of the options

View full abstract

Monetary transmission mechanism in Japan is investigated in two kinds of time intervals [1981,1990] and [1992,1999] associated with the burst of the bubble economy in 1990. The VEC model is constructed for nonstationary I(1) variables (gdp, money supply, bank loans, price) combined with stationary or nonstationary real interest rate r(t). The principal line of attack is to use impulse responses in the growth rate model and accumulate them to obtain impulse responses of level variables. This accumulation gives us convergence property of level variables to nor-zero asymptotic states. We can calculate contributions to the asymptotic gdp from asymptotic money supply, bank loans and price in cointegrated and/or non-cointegrated systems. We show that the money channel has a stronger influence to gdp in [1981,1990] compared with credit one, while in [1992,1999] containing the period ”after the bubble” the importance of credit channel dramatically increases. Furthermore, by introducing real interest rate instead of nominal interest rate, a strange behavior of the price called ”price puzzle” was resolved.

View full abstract

During the last several decades, the soft computing techniques such as NNs & GAs have been studied quite extensively by many researchers and have been applied to various real world problems. They have been successfully utilized for constructing a large number of intelligent systems. One of the most remarkable application areas may be the financial market. Nowadays, comparatively large numbers of researchers show keen interest in the application of soft computing techniques to the financial market. In this paper, we shall propose a DSS for dealing stocks which improves the traditional technical analysis by utilizing soft computing techniques. In the proposed DSS, soft computing techniques are utilized in order to detect “Golden Cross” and “Dead Cross" several weeks before it occurs. We shall also give several computer simulation results which confirm the effectiveness of the proposed DSS.

View full abstract

A system which transforms signals from the brain into control signals is known as a brain-computer interface (BCI). Classification of EEG signals is a difficult task, especially when the derived classification result is to be used to control an electronic device, because in this case the classification has to be performed on a single-trial basis (i.e. not averaged).

In this paper, statistical pattern recognition method based on Quasi-AR model is introduced and is applied to discriminate the EEG signals during right and left motor imagery.

View full abstract

In this paper, a new real-time kinematic (RTK) algorithm for carrier-phase-based precise point positioning based on Global Navigation Satellite System (GNSS) regression models which the authors firstly introduce in the present paper. The derived RTK positioning algorithm may achieve the positioning accuracy in decimeter level, and will be possibly centimeter level.

View full abstract

we are concerned with a problem of the optimal selection of the gain matrix of a linear observation mechanism for the Kalman-Bucy filter. By introducing an information theoretic constraint, we obtain a gain matrix which maximizes the reduction speed of an weighted estimation error. In this paper, we are especially concerned with the case where the weighting matrix is not positive but nonnegative definite. By this condition, we can treat an observation with any dimension. This result is more general than the one obtained by one of the authors using a formulation in the optimal transmission framework.

View full abstract

Recent significant increase of the computational power leads to rebirth of Monte Carlo integration and its application of Bayesian filtering, or particle filters. Particle filters evaluate a posteriori probability distribution of the state variable based on observations in Monte Carlo simulation using so-called importance sampling. However, the filter performance is deteriorated by degeneracy phenomena in the importance weights. To circumvent this difficulty and to improve the performance, a novel filter called the Evolution Strategies (ES) based particle filter, has been proposed by recognizing the similarities and the difference of the processes between the particle filters and ES. In this paper, the proposed filter is applied to simultaneous state and parameter estimation of nonlinear state space models. Results of numerical simulation studies illustrate the applicability of this approach.

View full abstract

The purpose of this paper is to propose a pulse neuron model named Pulse Neural Network incorporating Short-term Synaptic Plasticity (PNNSSP). PNNSSP extends PNN so as to change synaptic efficiency in accordance with pulse frequency. Results of computational experiments indicate PNNSSP requires less number of neurons and synapses which are necessary to build XOR function than PNN. In addition, we make experiments in which PNNSSP is used for a controller of an autonomous mobile robot as an engineering application of PNNSSP. As a result, it becomes clear that efficiency of the robot which uses PNNSSP for the controller is superior in comparison with PNN.

View full abstract

In this paper, the author will investigate variances for a class of fuzzy random vectors(FRVCs for short), where FRVCs are the improved ones of the previously proposed by the author[I] and partly investigated[2, 3] 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 proposed FRVCs and their expectations are reviewed, where the set representation approach of fuzzy sets with the help of the multivalued logic are applied for deriving their reasonable definitions. Secondly, the variances of FRVCs are derived from both viewpoints of multvalued-logic approach and the Frëchet approach, and some preliminary results are studied theoretically.

View full abstract

The performance of an Electro Chemical Discharge Machining (ECDM) process, in terms of surface finish and rate of machining is significantly influenced by the performance of the feed control system, the process control system, and the pulse generator. This paper presents the design stages of an Adaptive Neuro Fuzzy Inference System (ANFIS) for the Pulse Classification of the ECDM process. Inter-electrode gap signals such as voltage and working current waveforms were aquired and analysed. Five types of pulses were identified. An Adaptive Neuro Fuzzy Inference System (ANFIS) was developed to classify the pulses.

View full abstract

In this study, First, Milanka Price Index (MPI) of the Colombo Stock Exchange (CSE) was forecast using Linear Moving average (LMA), Simple Exponential Smoothing (SES) and Adaptive Response Religh's Single exponential Smoothing (ARRSES). Then, a Feed - Forward Artificial Neural Network (FFANN) approach was developed where the inputs of the neural network are the forecasted values from conventional forecasting techniques. Mean Absolute Percentage Error (MAPE) and Prediction Error Variance (PEV) were employed to measure the performance of LMA, SES, ARRESES and the proposed FFANN method. Finally, the results of the conventional approaches have been compared with that of the proposed FFANN approach.

View full abstract

This research describes a high-speed method, which detects automatically whistler dispersion, to be used for determination of a global electron density distribution in the ionosphere and plasmasphere. It is required in order to know the detailed state of plasma. It is necessary to observe the electron density distribution in the ionosphere in a short time scale, a day or shorter. A Japanese scientific satellite, “Akebono (EXOS-D)” has been used for various observations in the ionosphere and plasmasphere. The electron density distribution in these areas can be estimated from the propagation delay time of the whistler phenomenon, which is generated by electric discharge of the lightning on the ground and is observed by the satellite, the procedure to determine the distribution is to find a theoretically deduced model in which calculated propagation delay time fits the observed delay time. The purpose of this study is to reduce the analysis time of the delay (dispersion) of whistlers observed by Akebono, using a patterm recognition technique, so called “Hough Transform”

View full abstract

This paper provides a method of detecting cracks in concrete pillars by using an acceleration pickup. By giving a shock on the pillar with a hammer, several types of stationary waves generated in the pillar are observed with the sensor on one edge of the pillar. The method identifies the location of the cracks by modeling the sensor output as an output of a linear dynamic system with unknown parameters and applying a Kalman filter and the maximum likelihood method. Experiments show that two cracks in a concrete pillar of length 0.91m are detected within the localization error of 3～10mm.

View full abstract

This paper proposes a simple inter-symbol interference (ISI) and inter-block interference (IBI) elimination scheme for a single carrier block transmission with cyclic prefix (SC-CP) system with an insufficient guard-interval (GI). In the SC-CP system, only limited number of symbols in a transmitted block cause the interferences, while all the information data contribute to the interferences in the orthogonal frequency division multiplexing (OFDM) system. Therefore, in the SC-CP system, the interferences can be exterminated by only setting transmitted signals to be 0 at certain time slots in a transmitted signal block without changing any parameters or configuration of the system, such as the length of GI. The proposed scheme also can be considered as a simple loading scheme in temporal domain, which tries not to use time slots contaminated by the ISI and the IBI. Moreover, this paper also proposes a pilot signal configuration for the SC-CP system, which enable us to estimate channel impulse response in the discrete frequency domain even when the channel order is longer than the GI length.

View full abstract

This paper reports a technique for discriminating double talk and echo path change using the stochastic characteristics of power change for an adaptive noise canceller. The causes of rapid error increasing are double talk and echo path change. When the echo path is changed, the system corrects the impulse response in order to reduce the error. However, in the case of double talk, the system has to suspend the updating impulse response in order to maintain the quality of the voice signal. In the conventional system, it was difficult to discriminate between the two situations. In this research, the stochastic characteristics of the voice power change in the double talk period were experimentally verified to be different from the power change during echo path changing. Based on the results, a new double talk detection method is proposed.

View full abstract

Recently, adaptive arrays have been studied in order to improve communication quality of mobile communication systems. The least mean square algorithm is one of techniques which can be applied to adaptive arrays. However, this method does not sufficiently exploit the degree of freedom of the step parameter of the recursive rule in the case where a priori information on the input signal is available. This paper proposes a design method of the step parameter such that the convergence of the recursive rule is guaranteed against the uncertainty of the input signal. The design problem of the step parameter is reduced to a semidefinite program which is efficiently solvable.

View full abstract

This paper proposes an anomalous measurement detection algorithm in carrier phase differential GPS(Global Positioning System) positioning in order to attain continuous precise positioning. In this paper, firstly the characteristics of the errors in the measurement data are briefly reviewed such that main error sources affecting the quality of the measurements are identified. Then a new algorithm which comprehensively detects these errors using the innovation process of the Kalman filter is presented.

View full abstract

In order to keep reliability of thermal power plants, maintenance of system is required. Through a maintenance plan, it is expected not only to improve the system reliability but also to reduce the total cost. In this research, a unified maintenance plan based on prediction of deterioration is proposed. By a data classification technique, the unified maintenance plan was improved.

View full abstract

This paper proposes a method for detecting the change of mechanical vibration modes by means of a blind system identification method based on statistic independence. If a fault breaks out in a machine under disturbance, the method makes it possible to find the change in system parameter without sensing the vibration source directly. Experiment with a flexible structure has been carried out to verify the effectiveness of the method.

View full abstract

The paper considers a sequential inspection policy in an imperfect production process which shifts randomly from an ‘in-control' state to an ‘out-of-control' state following a general probability distribution. The manufacturer is in a contractual agreement with the customer to provide free minimal repair service until a certain (warranty) period from the time of initial purchase. The objective is to determine the optimal number of inspections and inspection time sequence during a production run which minimize the manufacturer's future expected costs in present term or average cost in distant future.

View full abstract

This paper deals with a system of a serial production line consisting of machines and intermediate buffers with finite capacities. In some of such systems the machines will become idle, even when they are ready to work on items, because of starving and blocking. While the machines are idle, they can carry out some auxiliary jobs with economical benefits. We call the periods carrying out such auxiliary jobs as vacations. By utilizing the idle periods in this way the system will be more productive. The present paper proposes a method of evaluating performances of such production lines with vacations. We will construct a model of the systems as a stochastic process. A common difficulty in those situations is that the system of equilibrium equations will easily become huge even for practical scales of production lines if we formulate the whole system as one entity, i.e. as one system of equilibrium equations. We will first develop an exact solution method for the 2-stage system with two machines and a buffer in between. For larger systems we will develop an approximate method decomposing the system into a set of 2-stage systems. The approximate method is an iterative one. Numerical experiments show that the proposed method is very accurate as well as efficient.

View full abstract

Under a multi-source and multi-microphone configuration, a new method is proposed for solving the permutation problem inherent in frequency domain independent component analysis (ICA) by utilizing prior informatin about the relative position of sound sources. The proposed method can also solve the channel selection problem for selecting the only channel for the target sound estimate, which is critical to a practical application of ICA.

View full abstract

In this report, at first, an algorithm for CT based on the MAP (maximum a posteriori) estimation is derived for the systems, and a matched filter with complex absolute detection is introduced for the optimal signal processor to be free from non-linear phase shift. Then, to reduce the measuring time, the Gold sequence P.R.K. (Phase Reversal Keying) is proposed to the signal, since it provide sharp autocorrelations and low cross-correlations, being useful for simultaneous time-of-flight identifications on many acoustic paths. Finally, the total performance of the proposed methods is illustrated.

View full abstract

We have developed the new real time speech-displaying system called “KanNon” which helps deaf person to understand speaker's speech contents by spectrogram-reading. The term of the spectrogram-reading means understanding the speech contents by seeing formant transition which appears in the sound spectrogram. This idea is based on that the speaker's speech is characterized by formant transition. The KanNon system displays not only sound spectrogram, but also pitch frequency and loudness of speech and characters by speech-recognition system as real time scrolling image. For the purpose of displaying formant patterns clearly with high accuracy, we applied Burg method combining with the minimum cross-entropy (Burg-MCE) method to estimate the autoregressive model parameters, and also applied the mel-scale to the frequency axis in the sound spectrogram instead of the hertz-scale. Finally, we show more effective display for the spectrogram-reading in the KanNon system.

View full abstract

Two universal imaging rules, self-similarity and perspective projection, are combined for detecting almost horizontal space in complex scene. By extracting scale shift due to perspective projection through Laplacian-Gaussian image analysis, the expansion of maneuvering area is estimated in terms of the invariant measure associated with a fractal attractor. Estimated maneuvering area is shown to be visualized on the scene image without explicit geometry of the roadway. The representation scheme was verified through experimental studies.

View full abstract

In this paper we show how Principal Component Analysis can be mapped to a quantized domain in an optimal manner. In particular, given a low-bandwidth communication channel over which a given set of data is to be transmitted, we show how to best compress the data. Applications to image compression are described and examples are provided that support the practical soundness of the proposed method.

View full abstract

We consider the problem of designing optimal smoothing spline curves and surfaces for a given set of discrete data. First we derive concise expressions for the optimal solutions using normalized uniform B-splines as the basis functions. Then, assuming that a set of data in a plane is obtained by sampling some curve with or without noises, we prove that, under certain condition, optimal smoothing splines converge to some limiting curve as the number of data increases. Such a limiting curve is obtained as a functional of given curve to be sampled. The case of surfaces is treated in parallel, and it is shown that the results for the case of curves can be extended to the case of surfaces in a straightforward manner.

View full abstract

This paper focuses on a super distributed energy system that consists of a number of dispersed generation systems such as fuel cells, micro gas-turbines, and so on. The behavior of a customer with a dispersed generation system is modeled on the Ising spin model in statistical mechanics. The validity of the modeling of the behavior of a customer is verified based on the Monte Carlo simulation. The feasibility of decentralized autonomous control using vicinity information is also investigated on the basis of stability analysis of the Hopfield neural network model.

View full abstract