電気学会論文誌C(電子・情報・システム部門誌)
Online ISSN : 1348-8155
Print ISSN : 0385-4221
ISSN-L : 0385-4221
122 巻, 6 号
選択された号の論文の24件中1~24を表示しています
  • 石川 博康, 江川 孝志, 神保 孝志
    2002 年 122 巻 6 号 p. 910-915
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    Properties of metalorganic-chemical-vapor-deposition-grown AlGaN/GaN hetero-structures on sapphire substrates were studied by the Hall effect and atomic force microscopy. The effects of the Al composition and roughness of the surface on the two dimensional electron gas (2DEG) mobility were investigated. 2DEG mobility of undoped-Al0.11Ga0.89N/GaN heterostructure on sapphire was 1100 and 9260cm2/V-s at 300K and 4.6K, respectively. We also fabricated recessed gate and non-recessed gate AlGaN/GaN high electron mobility transistors (HEMT). The recessed and non-recessed gate devices showed the maximum transconductance 146 and 93mS/mm, and drain-source current 900 and 720mA/mm, respectively. The recessed gate AlGaN/GaN HEMT exhibited the very weak temperature dependence of the threshold voltage. Our results demonstrate significant performance improvement for AlGaN/GaN HEMT by using recessed-gate.
  • 湯ノ口 万友, 木場 爲広, 王 鋼, 吉田 宏, 玉利 陽三
    2002 年 122 巻 6 号 p. 916-921
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    It is well-known that magnetic stimulation is quite promising as a method for transcranial that the induced current parallel to the nerve bundle was most effective for excitation, as estimated from a simple model analysis using the activating function. It should be noted that the perpendicular alignment could also evoke neural responses under a certain condition. The next step will be investigated the characteristics of peripheral nerve excitation for each direction of current induced by magnetic stimulation. We performed the experiment to estimate a focality of magnetic stimulation by using the sciatic nerve removed a bullfrog. We noted how the nerve bundle is affected to the direction of induced current by magnetic stimulation. We confirmed the possibility of nerve excitation in the case that nerve bundle was placed perpendicular to the induced current by experiment using sciatic nerve of bullfrog.
  • 鵜根 弘行, 横山 隆志, 池坊 繁屋, 田中 昭信, 銭 飛
    2002 年 122 巻 6 号 p. 922-927
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    It is necessary to minimize average of transfer delay time of network. To achieve this, several adaptive routing algorithms are proposed. In this paper, authors described new routing algorithm using ants computing and reinforcement learning method, which is proofed to give optimum resolution in stationary network traffic. Authors show that how this algorithm controls flow of packets and gives good result by comparing other algorithms.
  • 本郷 保夫
    2002 年 122 巻 6 号 p. 928-935
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    Generally speaking, Japanese OCR cannot easily read Japanese documents that also contain alphanumeric data, bacause of the proportional pitch setting of alphanumeric characters displaced in the fixed pitch setting of the Japanese document.
    This paper describes how to extract character candidates from combinations of small patterns that may be components of separable Japanese characters or slim patterns as alphanumeric characters, and how to select true character patterns from character candidates. We propose a new segmentation and recognition method for alphanumeric-mixed documents based on pattern recognition information such as similarities, pattern sizes and character kinds.
    The method was tested on alphanumeric-mixed documents, which were 51 pages of technical journals and transactions containing 68, 867 characters. The resulting segmentation rate was 99.75% and the recognition rate was 99.05%, so we conclude that this method may be applied to Japanese OCR.
  • 後藤 邦博, 斉藤 文彦, 山本 和彦
    2002 年 122 巻 6 号 p. 936-942
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    This paper proposed a method to recognize the kind of an object and to determine the location of the object in a gray-scaled image. The kind, location and pose of an object are unknown under the general environment. In the object recognition, typical image features that can represent the characteristics of an object must be used. But it is generally difficult to determine effective image features for object recognition. Therefore the genetic algorithm is applied to determine the combination of effective image features and their weights automatically in the proposed method. The relative weights of effective image features are determined in the candidates of general image features. The weights of the candidates of image features are represented as an individual chromosome. An individual fitness is evaluated by the discriminant analysis using the total difference of image features between the area including an object and other areas. The experimental results show that the combination of the effective image features and their weights were determined automatically and the objects were recognized successfully using the combination.
  • 鶴沢 偉伸, 武田 和大, 中山 茂
    2002 年 122 巻 6 号 p. 943-949
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    In order to improve man machine interface, graphical user interface is adapted rather than character base interface, and two dimensional icons are employed as picture characters that are visually easy to understand. File directory structures are shown as two-dimensional tree structures. It is considered that three-dimensional demonstration is more desirable like the real world, and then man will be able to understand it intuitively by look-and-feel. However, such three-dimensional demonstration has not been popularly used. In this system, we have developed a building tool of three-dimensional icon, and a file viewer with the three-dimensional tree structure. The system indicates file configuration and file content more visually. We have made them transferable on the network by using object serialization, and then the three-dimensional file directory structure of a remote host computer on the network can be displayed. Furthermore, we have evaluated the sytem and then obatained a good evaluation result.
  • 藤原 伸行, 恩田 寿和, 五十嵐 貴, 庭川 誠, 組田 良則, 小田 博志, 越山 浩次
    2002 年 122 巻 6 号 p. 950-956
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    We have developed a measurement system of the shape of the land using a stereo vision. When we use a stereo vision system to measure points in the image, we need to know the parameters for each camera. Usually, the camera parameters are calculated from a camera calibration technique, but it is difficult to apply to outdoor scenes. Our system calculates the camera parameters using the surveying instrument and makes the surface data of the land using a stereo ranging technique. In this paper, we describe the stereo vision system, the system configuration, the process of the measurement system of the shape of the land and an experiment to measure an actual slope.
  • 原田 賢一, 小林 康秀, 沖田 豪
    2002 年 122 巻 6 号 p. 957-963
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    This paper proposes the system identification method for linear system with time delay and unknown order. We hypothesize some candidate models which have different order, because system order is unknown. In each model, we estimate the model's parameters based on maximum likelihood method by using non-linear optimization technique. Then both local optimization technique and global search method are used because the estimated parameters may fall into local minimum. After all candidate model's parameters are estimated, one model is selected among these models to estimate system order by using a posteriori probability based on Bayes' theorem. Finally, the validity of proposed method is shown by numerical example.
  • 小山 昌宏, 三宅 徳久, 藤井 健二郎, 薦田 憲久
    2002 年 122 巻 6 号 p. 964-971
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    A new programming method, named cell control language (CCL), for factory-automation systems has been developed. In CCL programming, an action sequence of each functional unit in each operation mode is described as a net module. This net module is a subclass of Petri nets, and this subclass is equivalent to a state transition diagram, which is a conventional design method. Moreover, several net modules are merged and translated into an internal code program, which is executed in a cell controller.
    An efficient method to describe a high-speed program, which needs a quick response to a short-cycle state change, and to execute the program by using an interruption process was developed. Applying this method to an actual cell controller, we confirmed that it can increase the ability to respond to the state change without reducing total control performance.
  • 清水 能理, 山中 雄一郎, 宮崎 道雄, 李 義頡, 李 相球, 秋月 影雄
    2002 年 122 巻 6 号 p. 972-979
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    In recent years, the chaos control as one of the prospering application of chaos research is to stabilize a state into the peculiar unstable periodic orbit(UPO) embedded in the chaos atractor. Delayed feedback control(DFC) applies inputs based on the difference between the current state and the τ-time delayed state, and stabilizes UPO. Where τ denotes the period of the stabilized orbits. This method needs correctly no information about orbits if the periods of UPO is known. However, there exists the limitation of the odd number property. As the method to compensate this fault, there is prediction-based feedback control(PFC) using the prediction value of the τ-time future state. However, PFC needs to calculate the prediction value analytically. Then, in this paper, we propose the hybrid type technique of the both methods to improve these faults. Namely we combine DFC and PFC by the parameter W, which indicates the balance of two control inputs based on each method, so that each advantage is not lost. PFC input and DFC input in our method have a common feedback gain matrix K. Therefore, we calculate the ranges of W and K using Jury's method and determine the optimal pair of (W, K) using evaluation function, and stabilize the state into the target τ periodic orbit.
  • Xiaohong Jiao, Tielong Shen, Katsutoshi Tamura
    2002 年 122 巻 6 号 p. 980-988
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    This paper deals with the problem of robust L2 disturbance attenuation for nonlinear systems with input dynamical uncertainty. The input dynamical uncertainty is restricted to be minimum-phase and relative degree zero. A sufficient condition is given such that the nonlinear system satisfies the L2 gain performance and input-to-state stable (ISS) property. Using this condition, a design approach is given for smooth state feedback control law that solves the robust L2 disturbance attenuation problem, and the approach is extended to more general case where the nominal system has higher relative degree. Finally, a numerical example is given to demonstrate the proposed approach.
  • 遠藤 真一郎, 小西 正躬, 今林 弘資, 杉山 早美
    2002 年 122 巻 6 号 p. 989-994
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    Small Scale Hydroelectric Power Generation is made using the Water drop caused by the gradient difference between that of the river and that of the water guide channel. The water flow in this system apt to vary with time. Especially, at the time of the flood or that of the dry, the flow in the water stream varies largely. To realize the stable generation, it is necessary to maintain the water level of the head tank located in the mid way of the stream channel at a constant level. In this paper, variation of water level is forecasted using two tank models for the water channel and head tank and the water level control algorithm is proposed by finite settling time control. The effectiveness of the algorithm is evaluated beforehand using the simulation program of the water environment for the hydroelectric generation system.
  • Stephen Karungaru, Minoru Fukumi, Norio Akamatsu
    2002 年 122 巻 6 号 p. 995-1000
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    This paper presents a neural network based face detection system. Our objective is to design a system that can detect human faces in visual scenes at high searching speed and accuracy. We used a neural network with a simple structure but trained using face and non-face samples preprocessed by several methods (position normalization, histogram equalization etc) to attain high accuracy, then pruned the size of the neural network so that it could run faster and reduced the total search area of a target visual scene using skin color detector. Skin color detection assumes that faces reside only in skin color regions. The system designed, is made up of two parts: the face detecting system (FDS) that detects the faces (made up of the face locator, the down sampler, and the merger), and the searching speed improving system (SIS). Speed improvement is achieved by reduction of the size of the face locator (FL) network using structural learning with knowledge and by reducing the face search area using skin color detection system (SCD). Faster training of the neural networks was also achieved using variable step sizes.
  • 佐野 泰仁, 喜多 一
    2002 年 122 巻 6 号 p. 1001-1008
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    This paper discusses optimization of functions with uncertainty by means of Genetic Algorithms (GA). For such problems, there have been proposed methods of sampling fitness values several times and taking average of them for evaluation of each individual. However, in important applications having uncertain fitness functions such as online adaptation of real systems and optimization through complicated computer simulation using random variables, possible number of fitness evaluation is quite limited. Hence, methods achieving optimization with less number of fitness evaluation is needed. In the present paper, the authors propose a GA utilizing history of search (Memory-based Fitness Evaluation GA: MFEGA) so as to reduce the number of fitness evaluation for such applications of GA. In the MFEGA, value of fitness function at a novel search point is estimated not only by the sampled fitness value at that point but also by utilizing the fitness values of individuals stored in the history of search. Numerical experiments show that the proposed method outperforms the conventional GA of sampling fitness values several times at each search point in noisy environment.
  • 櫻井 茂明, 市村 由美, 酢山 明弘, 折原 良平
    2002 年 122 巻 6 号 p. 1009-1015
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    A text mining method using domain-dependent dictionaries can classify text data with various viewpoints. The method uses a key concept dictionary, which stores important words and phrases for domains. Also, the method uses a concept relation dictionary, which is a rule set consisted of their combination. In the method, the knowledge dictionaries are very important and give a strong influence to classification results. However, we have to generate the dictionaries through trial and error. It is difficult to apply the method to many tasks. In this paper, we try to learn a concept relation dictionary automatically. The method extracts key concepts using lexical analysis from text data, generates training examples from the concepts and their classes given by a human expert, and applies the examples to a fuzzy inductive learning algorithm, IDF. Also, the paper shows the method acquires an appropriate rule set by numerical experiments based on 10-fold cross validation and using more than 1, 000 daily business reports.
  • 横山 隆志, 銭 飛, 平田 廣則
    2002 年 122 巻 6 号 p. 1016-1022
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    In network computing systems, load balancing is the problem of job distribution among multiple processors. In the area of job distribution, Round Robin method and random job division are the simplest. The demerit of these approaches is to be unable to adapt to the change of network environment. Therefore, a job scheduling mechanism is required to make reasonable job assignment for efficient use of the network. This mechanism should be able to not only adapt dynamically to the destination but also allocate jobs even without advance prediction of the changes of traffic, individual computer process speed, or the length of waiting queue.
    We propose learning automaton based job scheduling method for the load distribution problem in parallel distributed computing systems. This method can dynamically respond to job addresses and make proper job assignment. We show the effectiveness of our method by computer simulation.
  • 川中 普晴, 山本 康高, 吉川 大弘, 篠木 剛, 鶴岡 信治
    2002 年 122 巻 6 号 p. 1023-1032
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    Nurse Scheduling Problem (NSP) is the problem that allocating shifts (day and night shifts, holidays, and so on) for nurses under various constraints. Generally, NSP has a lot of constraints. As a result, it needs a lot of knowledge and experience to make the scheduling table with its constraints, and it has been made by the head nurse or the authority in the hospitals. Some researches for NSP using Genetic Algorithm (GA) have been reported. The conventional methods take the constraints into the fitness function. However, if it reduces the fitness value a lot to the parts of solution against the constraints, it causes useless search. Because most of chromosomes are selected in the initial population or as the change by the genetic operations. And if it doesn't reduce the fitness value so much, the final solution has some parts against the constraints. Some of them are established by the Labor Standards Act or the Labor Union Act, so the solution has to be modified. As a result, it is difficult to acquire an effective scheduling table automatically. We study the method of the coding and the genetic operations with their constraints for NSP. In this paper, we propose a new coding method and genetic operations considering the constraints. We apply this method to the NSP using actual shifts and constraints being used in a hospital. It shows that an effective scheduling table satisfying the constraints is acquired by this method.
  • 権田 英功, 宮田 仁志, 大北 正昭
    2002 年 122 巻 6 号 p. 1033-1042
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    In this paper, we propose a new type of membership functions (MSFs) and its efficient use to improve an optimization of fuzzy reasoning using a steepest descent method. In self-tuning of fuzzy rules using the steepest descent method, an algorithm to avoid suboptimal solutions by modifying learning coefficients has been proposed, where piecewise linear MSFs were introduced. In such an algorithm, when learning data has a radically changing distribution, it is impossible to avoid suboptimal solutions.
    To improve such a difficulty, we propose to apply double right-angled triangles's MSFs to the self-tuning of fuzzy reasoning. By using the MSFs, a radically changing grade can be represented easily. Besides, through a technique of the simulated annealing (SA) we propose to move peak positions of MSFs according to the progress of the learning, in order to arrange the MSFs on some positions where learning data is changing radically. Compared with the algorithm by the piecewise linear MSFs, this new algorithm can avoid the suboptimal solution more effectively. Advantages of this new technique are shown by numerical examples of some function approximations.
  • 奥谷 巖, 高瀬 達夫
    2002 年 122 巻 6 号 p. 1043-1051
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    In this paper three evolutionary algorithms, i.e., genetic algorithm(GA), niche method for evolutionary programming(NM) and immune algorithm(IA), are applied to solve a so-called man scheduling problem where it is required to reduce the project duration time with a constraint on the total availability of certain key resource and also to level everyday resource demand. Evaluation of the proposed algorithms is extensively conducted using ten randomly created networks. Algorithm efficiency is evaluated by maximum fitness designated as Fmax, mean fitness of ten best schedules, Fmean, and computation time. Model testing shows that GA and NM are comparable to each other based on the Fmax and Fmean criteria although NM can be said to be more efficient than GA as far as computation time is concerned when network size gets big with nodes over 60 while IA almost consistently and substantially outperforms GA and NM with respect to fitness as well as computation time.
  • 出木原 裕順, 中村 泰明
    2002 年 122 巻 6 号 p. 1052-1059
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    A spatial-temporal data structure, called the PMD-tree (Persistent Multi-Dimensional tree), has been proposed for managing the live intervals and locations of spatial objects. In the paper, novel concepts of time space bounding box (TSBB) and motion list are introduced to the PMD-tree to manage moving spatial objects efficiently. A TSBB is an extended bounding box for a moving object that covers the trajectory of the object. As an object moves, a TSBB corresponding to the object is enlarged to enclose the trajectory of the object. A TSBB is divided, when it becomes greater than a limit. An object and corresponding TSBBs are managed by a doubly connected linked list, called motion list. TSBBs are also managed by the PMD-tree. Introducing the concept of TSBB and motion list to the PMD-tree, moving objects can be efficiently managed and be quickly found for spatial-temporal queries. By the series of simulation tests, the storage requirements and search performances are evaluated for several types of moving objects. As a result, the proposed method is superior to the conventional methods.
  • 富井 規雄, 周 利剣
    2002 年 122 巻 6 号 p. 1060-1067
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    An algorithm developed to make daily station shunting schedules is introduced. To make daily station shunting schedules involves adjustment of arrival/departure times of regular trains and change of their arrival/departure tracks, which is quite difficult to perform by computers. We introduce an algorithm to store daily station shunting schedules that were performed in the past and to make daily station shunting schedules using the stored past schedules as precedents. In this paper, we propose a couple of ideas such as to retrieve the case base not only by the train numbers that are naturally considered as the key of the case but by several attributes of the related trains and to modify the retrieved case by using the algorithm we have proposed for the fundamental shunting schedules. Using actual train schedule data, we have confirmed that about 90% of the necessary changes to the fundamental schedules are successfully made by our algorithm.
  • 上平 員丈
    2002 年 122 巻 6 号 p. 1068-1070
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    Hyper realistic displays, such as large area displays and 3-D displays, have been interesting people because of the sense of immerse and presence. Recently, their application has been spreading rapidly as the virtual reality, cyber space technologies, etc. have advanced, although it used to be limited to a few fields such as amusement. Consequently, they are highly expected to create new picture world. This paper reviews the trends of hyper realistic displays and views the future prospect of their application.
  • 松井 敏明
    2002 年 122 巻 6 号 p. 1071-1073
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    Wireless distribution services using K/Ka bands have started in Japan for subscriber lines, paving the way to the use of millimeter-wave spectrum for broadband communications. Last year also saw the establishment of consecutive technical standards for 60-GHz indoor, wireless communications systems, thus accelerating possible deployment of these systems. Meanwhile, ongoing national projects such as the Stratospheric Wireless Platform Program are expected to influence the current terrestrial wireless systems as a whole in the near future. This paper reports on the trends in millimeter-wave communication systems along with the key technologies that are essential for them to be created. Emphasis is placed on antenna-circuit integration and high-speed devices from a viewpoint of improving system performance.
  • 関 康和
    2002 年 122 巻 6 号 p. 1074-1076
    発行日: 2002/06/01
    公開日: 2008/12/19
    ジャーナル フリー
    IGBTs have become to occupy wide area of the power device applications by replacement of BJT, GTO and another power devices. The characteristics of IGBT are developed year by year with using the technologies of new device design concept. After conventional IGBT with epitaxial wafer, new design concept IGBT with using FZ wafer are appeared. Furthermore, final ideal IGBT will be realized in terms of FS (field stop) IGBT. The new technologies for fabricating FS-IGBT are required of thin wafer process.
feedback
Top