In this paper, an FPGA circuit for estimating chaotic characteristics of unknown nonlinear systems is proposed. From particular solutions corresponding to chaotic signals of the unknown system, the proposed circuit approximates nonlinear functions. The approximation of these nonlinear functions are performed by using supervised learning. From the approximated functions, chaotic characteristics of the unknown system are estimated. An understanding of the behavior of unknown nonlinear system will be provided by utilizing the estimated chaotic characteristics. The proposed circuit is implemented onto an FPGA by using Verilog-HDL. This implementation confirmed that the proposed circuit can achieve high-speed operation and low-cost development.
Vanadyl-phthalocyanine (VOPc) thin films were prepared by an organic molecular beam deposition (OMBD) on Polyimide (PI) substrate. The morphologies and nonlinear optical properties of VOPc films were investigated with VIS/UV spectra and the third harmonic (TH) intensity measured by Marker fringe. The VOPc films prepared with different substrate temperatures (25°C and 150°C) showed the Phase structures of Phase I and II at 25°C, and Phase III at 150°C. The packing density of VOPc thin film having Phase III is much higher than that of Phase I and Phase II. Therefore, the VOPc film prepared at 150°C has higher packing density than that at 25°C. The VOPc films prepared on PI film charged with and no charged without corona charging were also investigated from VIS/UV spectra and TH intensity. The third harmonic intensity of the VOPc thin film enhances with the packing density in VOPc thin film. The third optical susceptibility (χ(3)) estimated from the peak of TH intensity of the VOPc thin film prepared on the PI film charged with corona charging is about 2.3×10-9 esu.
Vertical concentration profiles of SO2, O3, and NO2 in the lower troposphere were measured by a multiwavelength differential absorption lidar (DIAL) system. SO2 concentration profiles were measured for vertical range 2000-4000m by 3-wavelength dual-DIAL, and results showed that the effect of O3 can be eliminated by an appropriate selection of wavelengths. The measurement accuracy for dual-DIAL was estimated to be 1 ppb for 150m range resolution. O3 and NO2 concentration profiles were measured simultaneously for vertical range 1500-2500m with a measurement accuracy of about 6 ppb for 150m range resolution. These results show that the system is capable of high resolution measurement of a single species by dual-DIAL or simultaneous measurement of two species by conventional DIAL.
We have previously described newly-developed EEG derivation method from the free moving mouse. Briefly, EEG derived between an electrode placed on the dura mater through a burr hole on the skull and a reference electrode placed subcutaneously near the nose. This EEG derivation recording technique for conscious mice has enabled us to detect both EEG and sleeping status as baseline fluctuations caused by respiration. In the El-mouse, an established epilepsy model, maintained by sib-matings at the Department of Veterinary Physiology, Nippon Veterinary and Animal Science University, Tokyo, Japan, we have noted frequent appearance of spike discharges on EEG during sleep without ictal phenotype. Sleeping status was further monitored by animal's quinsence at the place with closed eyes and unawareness to environmental stimuli. These spikes resembled to those seen in interictal EEG and the pattern changed with age of mice. Thus, to reveal the nature of the spikes in sleeping EEG of El mice, we examined changes in power spectrum density (PSD) at different ages. And the wavelet decomposition was applied to EEG to identify the temporal changes of spikes and several components. PSD analysis indicated that the character of spike frequencies shifted toward higher frequencies with age of El-mice. And wavelet decomposition could assign spike discharges, ECG, respiration and body movement, respectively.
This paper proposes a new model of bacterial chemotaxis including not only intracellular information processing but also motor control on the basis of the molecular evidence. E. coli is chosen as a target bacterium, which has a simple molecular structure and is amenable to biochemical and genetic analysis. A computer model of the chmotaxis is developed in order to simulate its emergence. Parameters included in the model are regulated using the genetic algorithm in such a way that a fitness representing the chemotactic ability is maximized.
In this paper we consider a large number of wireless terminals that are interconnected by a multihop wireless network called an ad-hoc network. Design of routing protocols is a crucial problem in ad-hoc networks. Location information of wireless terminals is an effective measure for ad-hoc network routing. This paper presents a method to identify network topology implying terminal location and connections among terminals. A modified Self-Organizing Map (SOM) algorithm is proposed to apply to the network topology identification. This method exploits information of the received power levels of the signals that are transmitted by other terminals. This paper has evaluated how network topology is identified by using an example graph made by random numbers. The results show that only one bit information about the received power level in each terminal can identify network topology accurately with average error of about 10% for more terminals than a certain value.
In this paper, we propose a new system reflecting user's emotion in a virtual space. In the system, a user inputs his/her emotion using 6 emotional parameters in detail, and he/she chooses some shapes of objects to exist in a virtual space. From these information, the proposed system reflects user's emotion in 6 aspects using fuzzy logic and weighted average calculation. The aspects are 1) an extent of the space, 2) background colors of the space, 3) colors of the objects, 4) size of the objects, 5) moving speed of the objects, and 6) the number of the objects. In the created virtual space, user can go forward, go back and turn as he/she likes. The system is evaluated by objective questionnaires by 13 testees and the effectiveness has been shown.
Power Envelope Coding method (PEC method) was proposed to improve the clearness of explosive consonant for LSP-VCV speech synthesis. Here, a decision function plays significant role to decide the sampling position effectively of power envelope of residual signals of explosive consonants. Synthesized residual signal is composed by the product of power envelope and M-series signal. As a result of the subjective test for the synthesis speech, PEC method is effective for /t/, /g/, /d/, /b/. For/k/, /c/, the synthesized speech by using white noise excitation has sufficient clearness.
In this paper, we propose a method of character extraction for scene images based on identification of a local target area and adaptive thresholding. The proposed extraction method is performed as follows: A scene image is resolved into a lightness image and a saturation image using a HSL transform. Vertical and horizontal edge images are made from these two images, and these edge images are binarized and thinned. The corresponding prominent features in the saturation and lightness images are detected using the Hough transform. A region between straight vertical lines is then extracted as a signboard region candidate in reference to the edge histogram. The extracted signboard region candidate is binarized using a threshold value determined by adaptive thresholding for each character region in the signboard region. The binary image containing extracted characters is then analyzed and the linear region containing the most character strings is identified as the character string region. This technique was applied to 100 scene images in order to verify the reliability of character extraction. Of the 450 characters in all the images, 438 were extracted correctly, representing a 97.3% successful recognition rate. Correct character strings were extracted in 98 of the 100 strings examined.
We present a method of detecting the color document's horizontal(or vertical) direction using the local layout information of document's components. Different from conventional string or string-line detections which have been intensively studied in an image-based document processing, the presented method deals with a color document with or without strings included as document's components. The given color document is analized following three steps of (a)color subtraction, (b)detection of two line segments which wrap appropriate two components in the form of a single convex hull and (c)generating the histogram based on the frequency of above line segments. The overall direction for a given document is basically computed using this histogram. The experiment is carried out for 92 color documents including those with or without strings. The result shows that a satisfactory direction is obtained for 90 documents tested.
The problem of decentralized robust tracking and model following is considered for a class of large scale interconnected systems with uncertainties. A class of linear decentralized state feedback controllers are proposed for robust tracking of dynamical signals in such a class of uncertain large scale systems. The proposed decentralized tracking controllers can guarantee that the tracking errors between each subsystem and local reference model are uniformly ultimately bounded. Moreover, we modify the linear controllers by introducing some nonlinear parts so that the tracking errors decrease asymptotically to zero in the presence of uncertain parameters and interconnection terms. Finally, an illustrative example is given to demonstrate the validity of our results.
This paper proposes a method of fault detection for building air conditioning systems using relationships between measured data. In order to represent relationships, correlation analysis is applied to measured data. In air conditioning systems, variable machines are controlled to keep temperature of a room at a pre-set value. There are correlations among some pairs of measured data. If there are fault parts in the system, its correlations will be different from that in normal states. The fault detection can be realized by identifying the difference of correlations. At first, the parts which are controlled similarly in the system are picked up from the structure of a target system. By comparing correlation coefficient at each part, different parts can be nominated as fault. This method was applied to a Variable Air Volume system, which is one of typical air conditioning systems. From this result, it was confirmed that most fault parts can be detected correctly.
This paper focuses on a non-linear GPC (Generalized Predictive Control) algorithm by neural networks. More or less, many of the systems are non-linear systems, so the non-linear GPC algorithm is inevitable for the GPC application to the actual control systems. We use neural networks for determination of the parameters of non-linear GPC and Gauss-Newton method to obtain the control algorithm by the minimization of a cost function. The simulation results illustrate the usefulness of the technique presented in this paper.
Evolvable hardware can change its own hardware structure to adapt itself to new environment, and is realized by reconfigurable hardware and genetic learning such as genetic algorithms. Automatic logic circuit synthesis methods based on the evolvable hardware have been proposed in several recent years. These existing methods, however, involve some major problems that are how to reduce time of circuit synthesis, and how to synthesize a compact circuit which does not contain surplus logic gates. To solve the problems, in this paper, we propose a new circuit synthesis method using genetic algorithm based on the coexistence of heterogeneous populations. The object synthesized by this method is the sum-of-products formed logic circuit. In the proposed method, an individual can be defined as a logic circuit, and classified into heterogeneous sub-populations according to its number of contained logic gates. In addition to common genetic operators, namely, selection, crossover, and mutation, a new operator “movement of individual among heterogeneous sub-populations” is applied. The majority of individuals concentrate in a sub-population of fewer logic gates through many generations, and compact logic circuits can be efficiently synthesized at short time. The experimental results in several logic circuit synthesis problems show the effectiveness of the proposed method.
Wavelet neural networks employing wavelets as the activation functions recently have been researched as an alternative approach to the traditional neural networks with sigmoidal activation functions. In this paper, we proposed a new type of wavelet neural network by introducing local linear models, which are used in some neuro-fuzzy systems, as powerful weights instead of straightforward weights employed in the previous wavelet neural networks. The proposed network is called the local linear adaptive wavelet neural network. Its effectiveness is examined by the network performances on function approximation and chaotic time series prediction problems. In these experiments, the proposed local linear adaptive wavelet neural network performed well and compared favorably to the previous wavelet neural network.
This paper describes a new system for both expansion of applicable environment and destinction of bikes in traffic measurement by image processing in order to obtain traffic state exactly. We treat image sequence as input, and decide weather and time slots based on both environment parameter and information of back-ground image. And we get area of moving object by subtracting background image from input image under dicision environment. Moreover, this system detect relebant vehicle area by dealing with grouped labels with feature in one frame and continuous frame. Experimental results with this system show efficience on traffic measurement.
This paper describes a methodology based on domain reference models (DRM) for developing object-oriented software that has few software errors. We have already proposed this concept, which are reference models for modeling domain-specific objects. As the aim of this paper, we give the methodology based on DRM for picking up risk factors that cause software errors. A DRM define logically the static aspect of a domain independent of infrastructure environment, and represent guidelines of object extraction. For example, we define the DRM that we call the Presentation-Entity-Relay-Service (PERS) model for an industrial monitoring domain. Software engineers classify hot spots, which we redefine as the parts of software that is customized to all unpredictable risk factors, into related components of the DRM. This activity is performed in three phases of system analysis, and object-oriented analysis and design. We apply our approach to develop communication software of video-exchange systems in Intelligent Transport Systems (ITS). From the results of this application, we could reduce several software errors in the connecting test of this system. Therefore, it is effective for software engineers to develop software with our approach.
We studied a pleasantness estimation matrix which was constructed by the correlation coefficients between the electrodes of EEG of its three frequency bands. In this report, we examined the selection method of EEG data as the input vector. As a result, by using the matrix that was constructed by the data which has significant differences (p<0.01), we estimated subjects' pleasantness which corresponded to the stimli.
The generation of handwritten character by computers is tried in this study. First, the features of handwritten characters have been analyzed in the frequency region. It has been understood that there are some frequency characteristics, which look like 1/f or white spectra, at the positions of the edge and the center of gravity of the character. Next, the character is approximated with B-spline curve, and the control-point of B-spline curve is fluctuated base on the spectrum analysis. The character generated in this study looks like that generated by handwritten.
We introduce the concept of perceptual position to cluster a set of feature points (FPs) in the same way as that performed by human subjects. The perceptual position is based on a physical position of a FP. It is a fuzzy set whose domain of membership, being greater than 0, is determined by the FPs within a hypershpere centering around that FP with radius is L. A clustering algorithm for the perceptual positions is presented using fuzzy c means (FCM). As the results of experiments, clustering using this new algorithm is closer to that performed by human subjects than is that using the conventional FCM algorithm when we choose suitable L.