In this paper, the information compressed and transmitted and reconstructed software system (ICTRS) of Computer generated hologram (CGH) has been designed and established which can be widely applied by computer operators. This system structure can be applied to resolve the remote signal processing using the digital filter of CGH. It is a new system structure, and the size of image displayed is bigger (i.e., 7.11×7.11 (inc²)) In this system, LOCO-I image encoding technique and the image reconstruction technique using Fraunhofer transforming algorithm have been adapted. This processing method is a “lossless” or “near-lossless” compression algorithm whose compression ratio can be achieved to 1:5 (i.e., the compression ratio is 18.7806%), and the image of processed CGH can be effectively reconstructed by a computer. In experiments, we have mainly discussed and analyzed the amplitude and phase variation of CGH's pixels. In detail, we have explained why the image quality of processed CGH can be influenced by the pixel's amplitude and phase varied. The image quality of processed CGH has been compared with the image quality of original CGH, and the reconstructed image quality of processed CGH has been compared with the reconstructed image quality of original CGH in ICTRS. Finally, Compression ratio (R), Mean squared error (MSE) and Pear signal to noise ratio (PSNR) have been precisely calculated and analyzed to evaluate the reconstructed images quality of processed CGH and the image quality of processed CGH. The better compression and transmission and reconstruction algorithm model of CGH can be determined by the distortion measure. In this system, the image quality of processed CGH has been relatively improved. This method of processing CGH has been effectively verified by experiments.

抄録全体を表示

This paper discusses the optimal placement technique for LSI cells through genetic algorithms, and focuses particularly on the following points: (1) The algorithm has a two-level hierarchical structure consisting of “outline placement”, which partitions an LSI chip area into several areas, and “detail placement”, which determines cell positions in the partitioned area. The procedure for determining optimal cell positions is then explained. (2) For selection control, which is one of the genetic operations, new objective functions are introduced at each phase in addition to the usual objective function based on total virtual wire length. The objective function for outline placement controls dispersing wire congestion and that for detail placement controls eliminating local wire congestion. (3) Parallel processing systems suited to both outline placement and detail placement are studied. The experimental results prove that the proposed algorithm is effective for reducing local wire congestion on a chip, as well as total virtual wire length, and increasing the processing speed while preventing the deterioration of solutions.

抄録全体を表示

Document retrieval has been widely recognized as a key technology for obtaining efficiently useful infor-mation from a huge amount of documents. It is difficult, however, for conventional methods of document retrieval to select documents whose parts are solely relevant to a given query. In addition, users suffer from the problem of finding relevant parts from retrieved lengthy documents. In order to solve these problems, passage retrieval, which is to seek not whole documents but parts of documents or passages relevant to a query, has recently received much attention. In this paper, we focus on the passage retrieval method based on the fixed length passages (the fixed length method) and that based on the density distribution of keywords (the density distribution method), both of which employ a window function to obtain passages. By extending the fixed length method, we show that it is quite similar to the density distribution method from theoretical and experimental viewpoints, as well as they can be unified into the passage retrieval method with a window function. In addition, we discuss the conditions required to utilize the passage retrieval with a window function based on the experimental comparison to a simple keyword retrieval method.

抄録全体を表示

Genetic Algorithm (GA) is one of the remarkable global search methods. In GA using the initial individual population, which is generated by random number, the possibility that the solutions fall into local minimum is high and its solutions are not good sometimes. In this paper, we propose the method, which is called the elite GA (EGA), to improve the solution in GA. Here, some individuals with good adaptation are found by executing GA of a few alternations of generations. The elite population is achieved by repeating the operation several times. Because the variety of the solution can be secured in the elite population, the improvement of the ability to search is expected. Moreover, we propose the island model EGA (Island EGA, IEGA) which is the parallel processing of EGA.
In this paper, the simulation of minimizing LSI gate arrangement's area in the standard cell is executed by using each method which is the simulated annealing (SA), GA, EGA and IEGA. It is shown that the EGA and IEGA are more effective from these simulation results.

抄録全体を表示

With the advancement of high-frequency switching devices, the electro-magnetic interferences (EMI) have become problems in power electronic converter designs. It is necessary for an electro-magnetic compatibility (EMC) design to prospect and consider its possible EMI levels. This paper describes how to compute effects from a power converter to an object point to reduce conduction EMI noises effectively by an appro-priate design. First modeling techniques for converter elements are discussed for a model in the conduction emission frequency band by the parameter tuning method and for line constants by an analytical derivation. Then a derived model is simulated for harmonic distributions of loop currents and their magnetic fields.

抄録全体を表示

Since the number of automobiles has been increasing every year, it has become important to analyze traffic jam in recent years. If the problem like traffic jam can be solved, pollutions will be removed and the burden imposed on drivers will be lightened.
As the examples of traffic flow measurement, there are the methods of using loop-coil sensors or ultrasonic wave sensors. These methods have some disadvantages that construction of equipment become large-scale, expenses piles up and measurement points are fixed. They are not useful for traffic flow survey. By way of another method, though there is the method by manpower, it is difficult to measure such as distinction of the type of an automobile, and it is inferior in accuracy in long time measurement.
We consider constructing the real-time processing system of traffic flow measurement using a TV camera that is a general purpose use. The system consists of a handy TV camera, a personal computer, and a usual image processor system. We employed the method of using spatial frequency analysis of video picture plane to traffic flow measurement. This method has advantage of noise margin by employment of integral operation to calculate Fourier coefficients.

抄録全体を表示

This paper proposes a 3-dimensional information extracting method using multi-directional stereo images from satellite. In this method, we perform correlation analysis between pair of images with small angle difference and between pair of images with large angle difference. As a result, we can extract high accuracy 3-dimensional information than the conventional method. Multi-stereo imaging method causes the increase the amount. of the transmission data of the images. Therefore, we compress the transmission data using correlation of stereo images. Simulation results show the proposed method can extract high accuracy information without increasing amount of the transmission data.

抄録全体を表示

In order to obtain a reliable model of power system, identification of power system dynamics by employing a neural network is studied. A new method of combined use of a mathematical model and a neural network is proposed. The effectiveness of the proposed method is verified by applying to two kinds of one-machine infinite-bus system-an experimental system and a numerical simulation model system. In the conventional method, the neural network learns the generator terminal voltage of the system directly. On one hand, in the new method, the neural network is trained to learn errors between the generator terminal voltage of the system and that produced by the mathematical model. The results of the test show that good performance is obtained for the proposed method. Construction of more reliable model is demonstrated by combined use of the mathematical model and the neural network.

抄録全体を表示

Recent studies on global optimization using dynamical-system trajectories suggest that interaction among trajectories may lead to better performance than simple parallel search. In this paper, we closely examine and compare two forms of interaction applied to trajectories of second-order differential equations. Our results for various objective functions systematically illustrate the characteristics of individual trajectories as well as the overall performance attained by the best trajectory.

抄録全体を表示

In this paper, we propose a method of generating text about human behavior from video images using hierarchical concepts of motions of body parts including body, head and hands. In natural language, there are various expressions about human acts, so we firstly analyze and classify concepts of several motion verbs about each body part using case frame expressions which represent semantic structure of natural language sentences. By making correspondence between these concepts and estimated positions, postures and motions of human body parts from video images, appropriate verbs, objects, locus etc. are determined and expressed in case frames. Lastly these case frames of each body part are integrated into a total body expression, and translated into natural language text by applying case patterns and syntactic rules.

抄録全体を表示

In this paper, the new booster circuit is proposed whichi can generate the output voltages two times higher than a conventional on like a charge pump booster. The booster circuit is adopted both charge pump technique and series parallel alternating switched capacitor technique. The circuit is simulated by Spice software. The theoretical analyses and simulation results show the effectiveness of the proposed circuit. The proposed booster is suitable for embedded power supply circuit im MOS IC

抄録全体を表示

This paper proposes a method to classify three fatigue levels of bills using acoustic frequency band energy patterns by a neural network. The proposed method deals with an acoustic frequency band energy pattern from the acoustic signal which has been generated by the bill passing through a banking machine. The competitive neural network is used to classify a bill into one of three different fatigue levels. We use two types of the LVQ algorithm for training of the competitive neural network. The experimental results show the effectiveness of the proposed method.

抄録全体を表示

Two types of radiometers are generally used for microwave noise measurements. One is a total power type and the other is the Dicke type. The former is theoretically the most sensitive and more broadband in frequency, but it is slightly unstable and worse in the linearity of the receiver gain. From the viewpoint of realizing a broadband radiometer, the total power type has been adopted for precision noise measurements. In order to reduce the influence of non-linearity, a differential detection scheme has been employed in the detection section. And also in order to reduce the instability of the gain, the temperature of the front end section and the IF and detection section has been stabilized with circulated warmed water. Based on these ideas, a new microwave radiometer has been built. In addition a new equation of the output indication for the total power radiometer has been derived. By using this equation, an equation has been derived for making noise calibration and several equations have been derived for measuring parameters indicating the characteristics of the radiometer, which are important to perform precision noise measurements. And the effectiveness of the radiometer has been proven experimentally using these equations.

抄録全体を表示

Channel routing is one of the important tasks in the layout design of VLSI's, where a complex set of routing constraints should be satisfied, and at the same time, the number of tracks, the sum of wire length, and the number of vias should be minimized in order to improve the performance and production costs.
In this paper, we propose a new channel routing method using genetic algorithm based on the coexistence of heterogeneous populations. In the proposed method, an individual can be defined as a channel routing result, and classified into heterogeneous sub-populations according to its number of tracks. 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 tracks through many generations, and near-optimal solutions can be searched effectively. We have also applied the proposed method to several channel routing problems in order to demonstrate its effectiveness.

抄録全体を表示

This paper presents a fast dithering technique for MPEG viewers with a 256-color mode. It is difficult to employ pattern dither or error diffusion techniques for moving pictures with high frame rates, such as 20-30 frames per second. We overcome this difficulty by embedding a dithering function not into a display driver, but into an MPEG decoder. The proposed MPEG decoder incorporates a dither processing unit between an inverse quantizer and an inverse DCT (IDCT) unit. There are three reasons for the high speed operation of the proposed dithering technique. First, the proposed method yields a pattern dither signal by processing only 1/3 of DCT coefficients. Secondly, it does not processe R, G, or B signal but only Y signal before YUV-RGB transformation. Thirdly, only I-Picture is processed since the dither signal on it propagates to P-Picture and B-Picture automatically. By using this technique, we can reduce the number of additions to 1/150 of that needed by a conventional pattern dither method.

抄録全体を表示

This paper presents Reconfigurable Machine: RM-V developed to accelerate a wide range of applications on the flexible hardware architecture combining FPGAs and memories. RM-V consists of larger scale FPGAs, SRAMs, and SDRAMs offering 6.8 times as large gate capacity as the former prototype: RM-IV, and 170 times as large memory capacity as that. RM-V provides higher flexibility and scalability by using application specific base-board on which one to four module-boards are mounted. Experimental results with image processing applications have shown that RM-V works at the processing speeds 6 to 17 times as fast as software processing.

抄録全体を表示

This paper presents a high-level synthesis system, called RMAC-V (Reconfigurable Machine Application Compiler for RM-V), for applications using SDRAMs implemented on the flexible architecture of RM-V (Re-configurable Machine-V) combining FPGAs and memories. Given an application program written in the C language, RMAC-V produces an RT-level hardware description in VHDL. To reduce the clock counts needed to access memories, RMAC-V introduces two techniques: multi-clock scheduling and preloading row address. Experimental results on Wavelet Transform Engine (WTE) have shown 33% fewer total clock counts than those with conventional method.

抄録全体を表示

This paper is concerned with the application of the H_∞ filtering technique to the carier phase differential GPS positioning. This positioning method is based on measuring the range between the receivers and the satellites using the carrier phase of the satellites.
Generally, measuring the range with the carrier phase measurement involves the ambiguous cycles of the carrier signals. These cycles always take integer numbers, and are called “integer ambiguity”. Although many research activities have been focusing on the estimation of the integer ambiguity based on the least squares criterion, we combine, in this paper, one of these methods called “LAMBDA (Least squares ANIBiguity Decorrelation Adjustment)” with the H_∞ filtering technique.
Finally, the experimental results by using real receiver data obtained at static points are shown. The proposed method shows a robustness for uncertainty of the noise statistics.

抄録全体を表示

A landscape image can be classified by the two categories with a natural landscape and an artificial landscape by using the image characteristic of direction. In this paper, a method is proposed to extracting characteristic of the image by using morphological filter. The experiments have been done to verify the effectiveness of the proposed method for many landscape images to separate to natural landscapes and artificial landscapes. From the experiments, morphological “Opening” operation can extract the image characteristics of direction, and can be used landscape images to divide into natural areas and artifical areas by appling the difference in direction characteristic for picture elements.

抄録全体を表示

In this paper, the explicit solution to the discrete-time LQ problem for singular weightings using a generalized interactor matrix will be considered. It will be shown that a special derivation of generalized interactor gives the minimum-phase image of the plant. A derivation method of generalized interactor using the spectral factorization is also shown.

抄録全体を表示

A method of model matching control for continuous-time systems in the presence of arbitrarily bounded disturbances is proposed. The proposed control system consists of two feedback loops: one performs the model-matching control and the other includes an error feedback controller which could be fully utilized to reduce the effect of external disturbances. In this paper, an explicit design procedure for achieving perfect model matching control with minimal sensitivity will be proposed. Furthermore, we can assure that the system input and output remain bounded at any time for continuous-time system in the presence of arbitrarily bounded disturbances. Finally, the results of computer simulation are presented to illustrate the effectiveness of the proposed method.

抄録全体を表示

At present, a classification of the radial pulse waveform is done subjectively by the oriental medicine. Quantitative recognition of the radial pulse waveform has not been presumed to establish until now. For this purpose, we have investigated the correspondence between the radial pressure pulse waveform of 80 episodes (80 adults with normal physical findings) and the classification by three pulse diagnosis experts. It was found that, (1) the radial pressure pulse waveform quantification is related to ΔBP_P and ΔBP_D, and that, (2) the fundamental radial pressure pulse waveform classification in three groups; HUA mai, PIWG mai and XIAN mai in the Oriental pulse diagnosis is well correlated to ΔBP_P and ΔBP_D, that is presented here, “Locality of the waveform”.

抄録全体を表示

In this paper, we propose a new method to detect object and estimate the position and the orientation from an image using a three-dementional (3-D) model having feature points. In the proposed system, a 3-D model is employed in order to represent a 3-D object, and the system can generate the 2-D models from various viewpoints. Then, the system compares the model that projected onto 2-D with the object in the input image. In the comparison, the system gives attention to some partial feature points of the object, and computes the fitness value of the 2-D model in the input image. The system can estimate the position and the orientation of the object by maximizing the fitness of the 2-D model. In the proposed system, a genetic algorithm (GA) is employed for this optimization of the fitness. In the experiment, we took up detection and position and orientation estimation of a car. We have tested various kinds of images and obtained excellent results.

抄録全体を表示

This paper proposes a framework that a genetic algorithm is applied to determine and construct an organ, especially a neural network of a creature. The vision system of a creature is a result of genetic evolution. Thus, we are trying to realize it on the computer.
By using the method, we search how the visual organ of creature is evolved under a special environment, and how many variations of neural networks exist. Besides, we think it is possible to generalize the method to an automatic generation of various kinds of visually recognizing system by giving the various kinds of evolution directions. Here, we assume that the visual organ of a target creature is composed of a neural network with arbitrary structure and connection weights, and that the target creature will be eaten by the snake if it misrecognizes, and it will multiply if it correctly recognizes the food when it meets it. In these environments, we are trying to examine how the vision system of the creature is evolved by using the genetic algorithm. Furthermore, we examined how noises effected to the evolution of the vision system.

抄録全体を表示

This paper describes a method to reduce a variance of measured headway distance using stereo vision. In this method, first, headway distances are measured by both the stereo disparity and the difference of the size of a preceding vehicle on image, and then one of them which has smaller variance is chosen as the headway distance. The size of a preceding vehicle on image is measured by a new method of detecting the edges on the preceding vehicle which alleviates the bad influence of change in back ground. A headway distance measuring system using this method has been developed, showing the effectiveness on tracking a preceding vehicle within 25m headway distance.

抄録全体を表示

In this paper, a method of enhancing the proxy server is proposed, by which the relay function for com-pensating the connection can be established after the host move. This system consists of Mobile Hosts (MH) with the function of automatic sockets' switching, and Proxy/Relay server operated as an Access Point (AP). TCP connection is divided into plurality by this, and congestion control and flow control for each medium are separated. As a result, communication performance is improved, and maintenance of TCP connection can be realized by a slight modification to a wired network. After the implementation of the proposed server, we measured throughput, and for users movement from cell to cell, validity was confirmed for this methods.

抄録全体を表示

This paper considers a design method of the filtered inverse system for time-invariant system, which produces the input signal to the system approximately. The inverse system must be stable in practical application. It is well known that the inverse system is unstable if the system has invariant zeros in right half plane. Yamada and Watanabe presented a state-space design method of stable filtered inverse system for the strictly proper system with some right half plane invariant zeros. But the method was explained with a bit complicated process.
The purpose of this paper is to give simple explanation to construct a stable filtered inverse system. We constract the preliminary filtered-inverse system by usual method and derive a puseud proper plant from it. The stable filtered inverse system is obtained by using transformation toward the filtered inverse system. An application of the stable filtered inverese system to a design method of inner-outer factorization for strictly proper systems without using descriptor form and solving generarized Riccati equation.

抄録全体を表示

This paper introduces an adaptive hypermedia system that provides a means for information providers to direct their users through hyperspace. With our system, information providers are able to easily direct a user to his own personalized navigation path. Destination options are determined by hiding links and by applying rules that guarantees that a user is offered the best path for him. In order to reduce the effort in creating navigation rules, we simplified the format of these rules and offer a navigation support tool that verifies the navigation rules and reports any errors. This tool not only allows information providers to easily write navigation rules but also guarantees the adequacy of the navigation path.

抄録全体を表示

Researches related to “writer recognition” are recently watched. But the former researches are mostly “text-dependent” methods, and a methods called “text-independent” which are able to apply to the comparison between different characters are hardly reported. Because the text-independent writer recognition has many applications, for instance, judgment of testaments, threatening letters, it is much in social demand. In this paper we proposed a text-independent writer verification method based on the feature of inclination of strokes. Using this method we made an experiment on 20 people and 4 different sentences. The result of correct acceptance rate was 95.00%, and correct rejecting rate was 95.04%.

抄録全体を表示

Many techniques have been extensively researched into person identification using physical features of the human body. In this paper, we propose aa system for person identification using outline information of a hand. This method collects the data of many persons at the same time without using special devices. We used the higher order auto correlation features for extracting shift-invariable feature and separated a hand into 5 fingers to arrange the direction of every finger. The experiments using images converted into lower resolution to obtain more global correlation and using virtual training data to increase training data effectively are done. Validity of the methods was confirmed.

抄録全体を表示

In the field of oil or natural gas well drilling, Measurement While Drilling (MWD) is desired for the sake of safe and economical drilling operation. Among them, Electromagnetic-MWD (EM-MWD) has the advantage of higher data transfer rates. However, electromagnetic waves attenuate during propagation in the soil and are received as the weak signals that have been corrupted by environmental noises. The properties of the drilling field noise have been investigated. In order to increase transmission reliability, a new communication method has been established and a real-time adaptive noise canceller (ANC) has been developed. It has been verified by experiments in the drilling fields that the ANC reduces the noise by up to 20 dB and an underground transmission from a depth of 3, 300m has been successfully achieved.

抄録全体を表示

This study deals with a human-machine cooperative control system to ease recognition of work dynamics variation to facilitate skilled work. At first, a generalized model is introduced for the human-machine cooperative control systems which comprise a human worker co-operates with a robot manipulator holding a tool. The robot manipulator is assumed to be position-controlled using two measured interaction forces, one between the human and the manipulator and the other between the manipulator and the work. Secondary, it is indicated that two typical functions of the system, the transfer function from human force to tool position and its sensitivity function to work dynamics variation, are independently specifiable. Finally, experiments are carried out on an experimental human-machine cooperative control system that is designed to emphasize the work dynamics variation. The results illustrate that the system ease the work which is difficult to accomplish for a human worker alone.

抄録全体を表示

An extended Test Feature Classifier (TFC) for classification of patterns with many-valued features is proposed, which is a non-parametric method that is based on combination of features called prime-test features and on voting-based discrimination. In the binary-valued TFC, the features should be coded into some binary representations. Using the classifier was difficult in cases having many features because the long binary representation leads an increase in the computational cost of learning and discrimination. By introducing the Euclidean distance as a metric to define prime-test features, the classifier can directly handle manyvalued features. To confirm the effectiveness of the new prime-test features and of the proposed TFC several quantitative evaluations are performed using an artificial dataset having heavily overlapped distributions and non-linear separability, as well as a real dataset having a large number of patterns and difficult classification tasks.

抄録全体を表示