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
138 巻, 9 号
選択された号の論文の19件中1~19を表示しています
特集:知能メカトロニクス分野と連携する知覚情報技術
巻頭言
特集論文
<システム・計測・制御>
<知能,ロボティクス>
  • 飯塚 正樹, 秋月 秀一, 橋本 学
    2018 年 138 巻 9 号 p. 1088-1093
    発行日: 2018/09/01
    公開日: 2018/09/01
    ジャーナル フリー

    In this paper, we propose a method to recognize functional attributes of everyday objects for vision system of partner robots. On the related research, there is a method to optimize recognition result with dense (fully connected) CRF which use the estimation result of functional attribute for each pixels. However, since this method is optimized from RGB data, it isn't able to sufficiently consider the shape of object, which have a relationship with the function attribute. In the proposed method, the recognition accuracy of functional attributes is improved by considering the object shape with the dense CRF describing the three - dimensional positional relationship. As a result of the experiment, the recognition rate of the proposed method is 77.0 %, which is 3.8 % higher than the related method. In addition, we confirmed that the processing speed is high as a side effect by reducing processing cost by oversegmentation of input data and using high speed identification by Random Forests. The mean processing speed per an object was 109ms in the proposed method.

<メディア情報,ユーザ・インタフェース>
<音声画像処理・認識>
  • 道下 裕也, 東 嵩士, 柴田 雅聡, 村松 竜弥, 山田 英夫, 榎田 修一
    2018 年 138 巻 9 号 p. 1100-1107
    発行日: 2018/09/01
    公開日: 2018/09/01
    ジャーナル フリー

    Recently, Co-occurrence histograms of oriented gradients (CoHOG) describes image features to calculate the co-occurrence of pixels allocated at the local level and has attracted attention as an effective object detection method. However, this method has some problems. For feature descriptions that focus on individual pixels, calculation cost and the number of dimensions tend to increase exponentially with respect to the number of pixels. Multiresolution CoHOG (MRCoHOG) can suppress such exponential increases to linear increase without reducing the classification accuracy. This paper proposes a procedure in which a feature plane is divided using a Gaussian mixture model and a histogram is automatically divided to establish a less costly method for performing MRCoHOG. Experimental results demonstrate that the proposed procedure is more effective than conventional procedures.

<ソフトコンピューティング・学習>
  • 西崎 博光, レオ チー シャン, 牧野 浩二
    2018 年 138 巻 9 号 p. 1108-1115
    発行日: 2018/09/01
    公開日: 2018/09/01
    ジャーナル フリー

    Recently, deep learning technologies have been in the spotlight. Deep learning is one of a powerful technology to classify or recognize objects which captured by a camera. Such application has a high affinity with Internet-of-Things (IoT) devices. Therefore, it is considered that these technologies are used in embedded systems and IoT devices. In this paper, we verify deep learning applications like image classification can work well on a small computer such as Raspberry Pi. We develop three deep learning applications by using two types of deep learning frameworks (libraries). We prepare four types of small computers, and these applications are tested on the computers. In addition, we also investigate the relationship between the processing time, the memory consumption and the number of parameters of the deep neural network model. The verification experiments show that a program based on a deep learning library implemented by C++ language fast run and simple neural network models could work in real-time on small computers. Besides, the other experiment also clears that the more parameters increase the processing time and the consumption memory in proportion without depending on the deep learning libraries and small computers.

特集:国際会議ICESS 2017
特集論文
<情報処理・ソフトウェア>
  • Yoshiro Imai, Shinya Hara, Shunsuke Doi, Koji Kagawa, Kazuaki Ando, Te ...
    2018 年 138 巻 9 号 p. 1116-1122
    発行日: 2018/09/01
    公開日: 2018/09/01
    ジャーナル フリー

    This paper proposes a new approach to improve and support Information Security Education based on a visual CPU simulator. Our visual CPU simulator can provide suitable demonstration for learners to comprehend how buffer overflow causes some remarkable effects for security matter. At the first stage of Information Security Education, a normal program including subroutine call and its return has been graphically and correctly executed by the visual CPU simulator, secondly demonstration of buffer overflow phenomena can be performed by the simulator through stack overwriting in order for learners to understand typical software vulnerability. Learners will be able to recognize how such an overflow phenomena easily occur some violence of stack replacement and exordinary program execution through utilization of the CPU simulator. This approach to visualize Information Security Education with our CPU simulator has been evaluated through questionnaire by learners in classroom lectures. It is confirmed that approach to visualize Information Security Education has been effective for learners to study buffer overflow phenomena by means of qualitative and quantitative analysis of the result of questionnaire.

  • Shinya Hara, Yoshiro Imai
    2018 年 138 巻 9 号 p. 1123-1130
    発行日: 2018/09/01
    公開日: 2018/09/01
    ジャーナル フリー

    This paper reports an educational tool and its evaluation in order for learners to study Computer Architecture, especially in the higher education of Science and Technology field. This tool, which is called Visual CPU Simulator, can provide graphical simulation of assembly program code (instead of machine language) and demonstration of Register-transfer-level micro-operations inside of CPU, namely precise detail of structure and behavior of inner CPU. This educational tool for CPU simulation has been designed and implemented in Javascript language as Web service. Its users select simulation modes by micro step, by machine cycle and by automatic repetition of such cycles. So they can learn how a computer works graphically, recognize inner structure of CPU easily and understand micro-operation based behavior of CPU effectively. This Simulator has been also evaluated through some kinds of questionnaires by learners in many classroom lectures. It is significantly confirmed that the simulator has been very useful and effective to learn Computer Architecture and organization/performance of CPU through its simulating facilities.

特集研究開発レター
<音声画像処理・認識>
  • 川上 裕介, 服部 哲郎, 今井 慈郎, 安藤 一秋, 堀川 洋, R. P. C. Janaka Rajapakse
    2018 年 138 巻 9 号 p. 1131-1132
    発行日: 2018/09/01
    公開日: 2018/09/01
    ジャーナル フリー

    In our previous papers, we have reported that automated arrangement of color image by Histogram Matching (HM) based on Gaussian Distribution (HMGD) gives good feeling impression if the brightness histogram of the original image has single peak. However, if there are multi-peaks in the histogram, the arrangement by HMGD processing does not always bring good results. In this paper, we propose an HM method using moving averaged brightness histogram of original image as the reference histogram instead of using Gaussian distribution. Also we present the experimental results by applying the both processing to the same image.

論文
<生体医工学・福祉工学>
  • 松野 茂, 沼田 崇志, 岩橋 利英, 高増 潔, 小谷 潔, 神保 泰彦
    2018 年 138 巻 9 号 p. 1133-1140
    発行日: 2018/09/01
    公開日: 2018/09/01
    ジャーナル フリー

    Ultrasonic diagnosis enables us to evaluate lifestyle diseases. However, operating probe precisely for several ultrasonic diagnoses is difficult for general people in clinic or house. In this study, we developed a reassembled ultrasonic diagnosis robot with parallel link mechanism which has six degrees of freedom by three trays. The robot was designed by genetic algorithm to be small and storable. In addition, we proposed a calibration method for assembly errors of directions among trays by driving each tray and evaluated the precision of the proposed method.

  • 立山 宗径, 木下 郁英, 赤尾 旭彦, 白坂 将, 沼田 崇志, 佐藤 大樹, 神保 泰彦, 小谷 潔
    2018 年 138 巻 9 号 p. 1141-1147
    発行日: 2018/09/01
    公開日: 2018/09/01
    ジャーナル フリー

    The autonomic nervous system (ANS) activity has been considered as an effective measure of the mental load increasing in modern information society. The heart rate variability (HRV) are commonly used to evaluate the balance between the sympathetic and parasympathetic activities. However, HRV can also be influenced by non-autonomic contributions such as blood flow. Moreover, the nonlinear interaction between the sympathetic and parasympathetic activities affect the HRV and thus they might not be clearly distinguished solely by the heart rate measurements. Therefore, more accurate evaluation requires investigation on co-variation across the indices of the ANS activities not only the HRV but also the ones which are not influenced by the change in blood flow and which are innervated by the sympathetic and parasympathetic branches in distinct ways. In this study, under the stimulation of postural change, three indices of the ANS activities; cardiovascular activity, sweat, and pupil diameter are measured simultaneously and evaluated. This experiment reveals high correlations between posture angle and each indices, and there is significant difference between correlations of RRI (R-R interval) and pupil diameter and between changes in RRI and sweat in supine position. These difference suggests difference between effects of sympathetic and parasympathetic activities.

  • Tomohiro Murase, Kosuke Oiwa, Akio Nozawa
    2018 年 138 巻 9 号 p. 1148-1153
    発行日: 2018/09/01
    公開日: 2018/09/01
    ジャーナル フリー

    In automated driving at Level 3, drivers constantly need to allocate attention to the driving environment to react immediately to a take-over request. However, the amount of attention of drivers to the driving environment has not been quantified. The objective of this study was a quantitative evaluation of the amount of attention to the driving environment, and the psychophysiological state of the driver, during automated driving at Level 3. Attention and driver state at Level 3 were evaluated using ERP and psychophysiological indices respectively with compared to Level 2. The dual task method was used, in which twelve subjects performed driving tasks (Level 2 and Level 3) on a driving simulation system while in parallel performing an auditory oddball task. The data showed that the amount of attention to the driving environment at Level 3 decreased in comparison to Level 2, as indirectly measured by a 19.2% increase in attention to the auditory oddball task. Sympathetic nervous system activity during Level 3 automated driving decreased as compared to Level 2. In the psychological state, comfortable feeling and arousal level decreased at Level 3.

<システム・計測・制御>
  • 今泉 淳, 三浦 元毅, 森戸 晋
    2018 年 138 巻 9 号 p. 1154-1162
    発行日: 2018/09/01
    公開日: 2018/09/01
    ジャーナル フリー

    As railway rolling stock is expensive, it is necessary to utilize them as efficiently as possible. Making rolling stock schedule (RSS) is one of processes in the planning phase. RSS is a sequence of daily schedules. A daily schedule is a sequence of train services for a specific train-set. RSS is to make daily schedules and a sequence of them so as to satisfy constraints for daily and monthly inspections which are performed within a certain period of time or a certain mileage, respectively. Each train-set is equally used regarding on cumulative mileage by following the RSS. However, it is often necessary to assign train-set(s) to daily schedules for some reasons irrespective of RSS. Rolling stock assignment (RSA) is to find an assignment of train-sets to daily schedules provided in RSS. In this paper, we represent a problem of rolling stock assignment by a network in which sequences of daily schedule for train-sets correspond to paths and formulate it into a set covering type problem. Furthermore, we propose a solution method based on column generation and rolling schedule.

  • 杉本 幸代, 朝日 一平, 椎名 達雄
    2018 年 138 巻 9 号 p. 1163-1168
    発行日: 2018/09/01
    公開日: 2018/09/01
    ジャーナル フリー

    When change over time of hydrogen gas concentration in a certain point is measured, a gas sensor is fixed at the measurement point, or a syringe is utilized to sample the gas, which is analyzed by gas chromatography. However, the gas sensor and the syringe have an influence on gas flow and its measurement. Non-contact measurement technology with high temporal resolution and high spatial resolution is necessary. Therefore, in this study, for the purpose of non-contact measurement of the hydrogen gas concentration in the small area under the gas flow, a specialized experimental device was developed. A laser beam (wavelength 355nm, pulse energy 120µJ, PRF 1kHz) with the beam diameter 1mmϕ at the measurement spot was irradiated to the hydrogen gas. Raman scattering light at the measurement point of 750mm ahead was focused by Fresnel lens, which is arranged at the orthogonal direction against the transmitting laser beam. As a result, it was proved that Raman signal intensity had good linear correlation with hydrogen density of more than 100ppm. The gas concentration in a small area could be quantitatively estimated without interference of the gas flow.

  • 井上 昭, 佐藤 孝雄, 矢納 陽, 鄧 明聡
    2018 年 138 巻 9 号 p. 1169-1175
    発行日: 2018/09/01
    公開日: 2018/09/01
    ジャーナル フリー

    This paper proposes an optimal control law to suppress ripples in inter-sample of transient state of multi-rate sampled-control systems. The law is obtained by deriving equations to calculate output ripples as solutions of Lyapunov equations. So far, to reduce the ripples in transient-state, there exists only trial and error method and does not exist a systematic method. In this paper, first a reference model is defined which generates a reference output with no ripples and has the same outputs with multi-rate control systems at sampled points. Then the ripples in transient state is measured by the integrals of squared output errors. And equations to calculate the integrals are given by solving of Lyapunov equations. Finally the optimal control law is derived by differentiating the equations of several sampling intervals by the control inputs. The sampling intervals are called by horizon.

部門記事
 
feedback
Top