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, 10 号
選択された号の論文の22件中1~22を表示しています
  • 長谷川 まどか, 加藤 茂夫
    2002 年 122 巻 10 号 p. 1699-1704
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    Digital image information plays an important role in multimedia systems. In this article, we describe recent technologies for high quality still image coding and communication. A growing number of new applications require additional functionalities from a compression standard, which either are not offered by JPEG, or if offered, not in the desired from. The new compression standard JPEG 2000 is designed to satisfy such requirements.
    We also describe the research on “extra high quality audio-visual system for creation of future V-A works and tow-way presence communication.” This system aims to present extra high quality audio-visual information which can evoke our deep impression.
  • 横顔像と等濃線を用いた研究
    岡本 教佳, 中村 納
    2002 年 122 巻 10 号 p. 1705-1712
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    This paper proposes a new algorithm of the automatic personal identification system. The contents are included in over three sections. Section 2. makes a survey of the conventional algorithms developed thus far. Section 3. explains the identification systems based on human face profiles. In this system, however, some deformation of the profiles, not only exact profiles, are investigated in detail. Section 4. describes a personal identification algorithm based on isodensity maps. The isodensity maps is a set of contour lines extracted by dividing a gray level histogram of the facial area into eight areas. Feature quantities of a face are well reflected in the isodensity maps. Experimental results for identification are also explained in each section.
  • 村瀬 曜子, 田中 敏幸, 岡 輝明
    2002 年 122 巻 10 号 p. 1713-1720
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    Sarcoma is a kind of tumor. There are many types of sarcomas. So it is very difficult for physicians to classify types of sarcomas and benign tumor or malignant tumor precisely, except for physicians who specialize in sarcomas. For example, dermatofibroma (benign tumor) and dermatofibrosarcoma protuberans (malignant tumor) are non-epithelial skin tumors which need to be distinguished. In this research, first of all, segmentation of cell nuclear regions whose features are general points of pathologic diagnosis is considered. And from some features computed from the segmented nucleus regions, an automated classification of benign tumor or malignant tumor is finally operated.
  • 青山 督, 山内 紀克
    2002 年 122 巻 10 号 p. 1721-1726
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    Whispering gallery modes in dielectric microdisk resonators are high-Q hybrid modes that propagate just inside the edge of microdisks. We derived 6 differential equations of the modes in dielectric disks using 3 dimensional model and cylindrical coordinates. We calculated resonant frequency and axial propagation constant kz using the iteration method between radial and axial boundary conditions. The resonant frequency was calculated as complex numbers. Imaginary part of this resonant frequency means radiation loss to radial direction. We could analyze effect on Q factor caused by the radiation.
  • 森 秀人, 名雪 琢弥, 曹 念文, 福地 哲生, 藤井 隆, 根本 孝七, 竹内 延夫
    2002 年 122 巻 10 号 p. 1727-1733
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    A differential absorption lidar (DIAL) system for measuring vertical concentration profiles of atmospheric NO2 was developed. The laser transmitter consists of a Nd: YAG pumped dye laser with LDS765 dye and a sum frequency generation (SFG) system consisting of a pair of KD*P crystals, and can emit laser pulses of wavelength 446.8 nm and 448.1 nm on alternate pulses, which are used as the “off” and “on” wavelengths in measurement of atmospheric NO2. This system is better suited for long time measurement of NO2, since the output stability of the SFG system was measured to be about 56 times higher than that of a conventional dye laser system using Coumarin 445 dye. Using the DIAL system with SFG, NO2 concentration profiles of about 20-40 ppb were measured for vertical range 500-1000m. These values are in agreement with ground-based measurement results.
  • Sajiiad Mohsin, Yukihisa Kurimoto, Yukinori Suzuki, Junji Maeda
    2002 年 122 巻 10 号 p. 1734-1741
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    We propose a method to extract the QRS wave in an electrocardiogram (ECG). It is extracted by detecting two characteristic points (CPs), the Q and S points. There are two main problems that make detection of CPs difficult: 1) noise contaminating the ECG and 2) individual variation of waves and complexes in the ECG. We use DP matching for overcoming the problem of noise contamination and a neural network of ART2 for overcoming the problem of individual pattern variation. These two methods are fused using a multichannel ART-based neural network (MART) for reliable detection of CPs. The method was evaluated using an MIT/BIH arrhythmia database. It was found that the rates of accuracy within 6 ms error were 99.6% for S point detection and 96.4% for Q point detection, indicating that the present method has good potential for detection of CPs on an ECG.
  • 紙本 伸明, 王 丙辰, 大松 繁
    2002 年 122 巻 10 号 p. 1742-1747
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    This paper is concerned with an automatic classification of interphones into good and bad clusters based on the spectral data using two kinds of neural networks. One is the self-organizing map (SOM) by Kohonen and the other is the layered neural network (LNN) using the error back-propagation method. The SOM is used to find the representative teaching data for each cluster in order to achieve the fast convergence of learning of the LNN and reduction of the network size. The LNN is used to classify the input data into good and bad clusters. From the real data classification of interphones, we can see that the proposed method using two kinds of NNs could classify the data more precisely compared with the case using only a conventional LNN.
  • Zhigang Chi, Toshifumi Kimura, Kenji Yamauchi, Kennichi Hatakeyama
    2002 年 122 巻 10 号 p. 1748-1754
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    In this paper, we propose a method to segment images into suitable number of contents by motion information, and merge the contents with the same or almost the same motion to one to decrease the overhead and increase the compression ratio. For content with the global motion and the local motion, we use adaptive methods respectively to deal with both kinds of the motion compensation. In our experiment, we use Cubic Curve Model to compensate the local motion of the lip part and use extended affine transform to compensate the content global motion.
  • 斉藤 文彦
    2002 年 122 巻 10 号 p. 1755-1761
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    This paper proposes a method to generate a high-resolution document image from plural low-resolution images with almost same view areas that are captured by a handy camera. The relations between the locations of pixels in the high-resolution image and the types of low-resolution images are determined by the genetic algorithm. The individual chromosome is represented by the two-dimensional array of genes that include integers to distinguish low-resolution images. When the high-resolution image is supposed to have fine definition and sharpness, a higher fitness is given to the individual. The experimental results show that the high-resolution images with good quality were generated by the proposed method because the visual readability of the characters in the image was improved.
  • 安藤 慎吾, 今井 聡子, 中島 真人
    2002 年 122 巻 10 号 p. 1762-1771
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    When acquired image is outputted, information on the original image is often lost because of narrower dynamic range of output device such as printer or CRT display than input device such as digital camera. Accordingly, we proposed a method which corrects local contrast and brightness of the input color image automatically and dosen't make the image information lose. In this method, the input image is divided into some region by LOG filter. After that, optimal conversion curve of each region is constructed toward the intensity ingredient based on Adaptive Histogram Equalization, and then too bright or dark regions are improved automatically. As a result of applying this method to the multiple image, remarkable effect was confirmed for visibility enhancement.
  • 寺田 賢治, 星川 京子, 水上 幸子, 山口 順一
    2002 年 122 巻 10 号 p. 1772-1778
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    About 60, 000 fires have occurred every year in Japan. To be most important to the fires is an early period fire fighting. At present, the automatic devices of detectiong fires is needed. The alarms which can detect smoke and heat are utilized to house fires. However, such the alarms is not useful for the outside of house such as the incendiary or woodland fire. Therefore, in this paper, we propose the method that detecting the fire in the wide area by using camera images This method is able to detect such a flame that becomes a fire is the early period. First, the region of the flame in the images obtained from the observation camera is detected. Next, the characteristic quantity that expresses the flame is extracted. This characteristic is not influenced to the light such as the flashlight and street lamp. In other words, the only flame can be detected, from the range which looks like the flame in the image.
  • 中村 貴弘, 川村 俊明, 持木 幸一, 道川 太一
    2002 年 122 巻 10 号 p. 1779-1785
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    A handy type plastic discriminator has been developed. It consists of a near infrared light source, a bundle of optical fibers, a spectroscope, an electric cooled InGaAs linear image sensor, and a CPU board for discrimination processing. Six kinds of plastic can be identified using the Soft Independent Modeling of Class Analogy.
  • Hansheng Wu
    2002 年 122 巻 10 号 p. 1786-1793
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    The problem of decentralized stabilization is considered for a class of large scale linear time-varying systems with delayed state perturbations in the interconnections. In the paper, the upper bounds of the perturbations are assumed to be unknown, and a class of adaptation laws is introduced to estimate such unknown bounds. By employing the updated values of these unknown bounds, a class of decentralized memoryless state feedback controllers is also proposed. Based on Lyapunov stability theory and Lyapunov-Krasovskii functional, it is shown that by employing the proposed decentralized controllers, the solutions of the resulting adaptive closed-loop large scale time-delay system can be guaranteed to be uniformly bounded, and their states can converge uniformly asymptotically to zero. Finally, a numerical example is given to demonstrate the validity of the results.
  • 和久屋 寛, 信太 克規
    2002 年 122 巻 10 号 p. 1794-1802
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    Numerous studies on time series prediction have been undertaken by a lot of researchers. Most of them typically used uni-directional computation flow, i.e., present signals are applied to the model as an input and predicted future signals are derived from the model as an output. On the contrary, bi-directional computation style is proposed recently and applied to prediction tasks. A bi-directional neural network model consists of two mutually connected subnetworks and performs direct and inverse transformations bi-directionally. To apply this model to time series prediction tasks, one subnetwork is trained a conventional future prediction task and the other is trained an additional task for past prediction. Since the coupling effects between the future and past prediction subsystems promote the model's signal processing ability, bi-directionalization of the computing architecture makes it possible to improve its performance. Furthermore, in order to investigate the acquired signal transformation, two kinds of chaotic time series, i.e., the Mackey-Glass time series and the “Data Set A”, are adopted in this paper. As a result of computer simulations, it has been found experimentally that the direct and inverse transformations developed independently and their information integration give the bi-directional model an advantage over the uni-directional one.
  • 井上 武士, 古橋 武, 前田 宏, 高羽 実
    2002 年 122 巻 10 号 p. 1803-1811
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    Nurse Scheduling Problem (NSP) is a combinatorial optimization problem. There is a growing demand for practical automatic nurse scheduling systems. This paper presents an interactive nurse scheduling support system. This system generates new schedules using Bacterial Evolutionary Algorithm (BEA). The nurse-in-chief can modify/fix parts of candidate schedules. The system reschedules the parts that need improvement. The proposed system is required to produce high quality schedules with a short computation time for a good human-machine interaction. This paper clarifies effective BEA operations and parameters, and studies hardware implementation of BEA on Field Programmable Gate Array (FPGA). It is estimated that the hardware implementation of a practical NSP can be solved within 16 seconds.
  • Qingyu XIONG, Kotaro HIRASAWA, Jinglu HU, Junichi MURATA, Dazi LI
    2002 年 122 巻 10 号 p. 1812-1820
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    In this paper, Universal Learning Networks with Branch Control (ULNs with BC) is proposed. The point of the paper is to adjust the outputs of the intermediate nodes of the basic network using an additional branch control network. The adjustment means to multiply the nodes outputs by the coefficients ranging from zero to one, which is obtained from the branch control network. Therefore, the following are done in ULNs with BC, (1) the branch is cut when the coefficient of its branch is zero, and (2) multiplication is carried out in the nodes outputs adjustment when the coefficient takes a nonzero value. ULNs with BC is applied to a function approximation problem and a two-spirals problem. The simulation results show that ULNs with BC exhibits better performances than the conventional neural networks with comparable complexity.
  • 村橋 善光, 道木 慎二, 大熊 繁
    2002 年 122 巻 10 号 p. 1821-1829
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    One of the methods to solve a connection area problem on the hardware implementation of neural networks is to use pulsed neuron (PN) model. Conventional PN models are dependent upon the biological observatoins and the trial-and-error method. Hence it appears that little attention has been paid to the design methodology such as how to decide time constant, how to generate pulses, how to process the pulse coded signals. Thus, a novel Pulsed Neural Networks based on Delta-Sigma Modulation (DSM-PNN) is proposed. Delta-Sigma modulation is an attractive technique in the field of Digital Signal Processing. DSM-PNN can transfer information with only 1-line connection between two neurons, therefore its small circuit scale is effective for hardware implementation. All the more, it is possible to transmit the signal faithfully by noise-shaping effect, and multi-input summation and weight multiplication can be operated precisely. Our proposed method by a neural network is evaluated with Generalized Hebbian Algorithm (GHA) which is a learning rule of Principal Component Analysis (PCA). Simulation results show that the proposed system has same accuracy to those with floating-point unit.
  • Gaofeng Xiong, Tomonori Hashiyama, Shigeru Okuma
    2002 年 122 巻 10 号 p. 1830-1836
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    The daily bidding strategy in a day-ahead electricity auction market is studied from a supplier's point of view. An improved evolution strategy is developed to evolve the bidding strategy and to maximize the supplier's profit in a long run. A competitive day-ahead electricity auction market, where no supplier possesses the market power and all suppliers winning the market are paid based on their own bid prices, is assumed here. The dynamics and the incomplete information of the market are emphasized. A market clearing system is also included in the implementation. An agent-based simulation method is presented in this paper. The simulation results show the feasibility of the proposed bidding strategy.
  • 田村 宏樹, 唐 政, 若田 航太, 石井 雅博
    2002 年 122 巻 10 号 p. 1837-1845
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    We propose a Hill-Climbing Learning Metiod (HCLM) for the flowshop problems. The hill-climbing learning methd is a technique which uses the local searching like method (we call it the energy converging method (ECM)) to decrease the energy function and gets a solution to the flowshop problem. If the solution is a local minimum, the HCLM attempts to drive the solution out of the local minimum by changing the parameter of the energy function and gets to an optimal or better solution. Simulations are performed to many flowshop problems using this proposed method, and the validity is shown.
  • 阿閉 祐介, 小磯 貴史, 仲谷 美江, 西田 正吾
    2002 年 122 巻 10 号 p. 1846-1855
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    Information filtering is a function to select useful or interesting information for the user among a large amount of information. This function has become very important as network technology develops rapidly. This paper focuses on information filtering for emergency management. When a large-scale disaster happens, the problem of information flood can be very serious because a great deal of information occur in a short time and they are sent to a person or an organization that is responsible for managing the situation.
    In this paper, we first consider features of information in emergency situations, and propose an information filtering system for emergency management. The proposed system is composed of three parts, that is, message structuring part, causal analysis part and information classification part. A prototype system is developed and its effectiveness is evaluated by experiments.
  • 浪岡 保男, 薦田 憲久, 松平 隆之
    2002 年 122 巻 10 号 p. 1856-1865
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    An ontology-based, automated generation method for state space models is presented. The models are used to derive sequence control program specifications writen in SFC (Sequencial Function Chart). The state space models feature (i) dividing the plant state space into the machine state space and the equipment state space, (ii) allowing the generation of the machine-level state space models before searches, and (iii) reducing both the search space for planning, and the number of planning steps. It is currently difficult to make practical state space models, however, because of the large size of such models and the difficulty of expressing the details of operations. To overcome this problem, we propose a method of automatically generating the machine-level state space models from plant ontology in short time without errors. The plant ontology includes actions, side effects, constraints, and hierarchical states that are used to represent other parts of the ontology. To investigate the effectiveness of the method, we have applyed it to a planning-based VIsual Programming System for Sequence control program SpeciFiCation Design (VIPS/S-SFCD).
  • Sajjad Mohsin, Yukihisa Kurimoto, Yukinori Suzuki, Junji Maeda
    2002 年 122 巻 10 号 p. 1866-1867
    発行日: 2002/10/01
    公開日: 2008/12/19
    ジャーナル フリー
    We propose a method for detecting characteristic points (CPs), the Q and S point, in an electrocardiogram. The method is a fusion of two methods, DP-matching and ART2 neural network. A multichannel ART neural network (MART) was used for the fusion. The method was tested using an MIT/BIH arrhythmia database. The rate of accuracy within 6 ms error was 99.6% for S point detection and 96.4% for Q point detection, indicating that the method is reliable.
feedback
Top