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
141 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
特集論文
<生体医工学・福祉工学>
  • 小野 蓮太郎, 松浦 康之, 中根 滉稀, 高田 宗樹
    2021 年 141 巻 9 号 p. 940-947
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    In accordance with developments in the image technology over recent years, it has become easier to watch 3D video clips. As the viewing of 3D video has become more commonplace, symptoms such as visually induced motion sickness (VIMS) and eye fatigue have been reported. Although various theories regarding the underlying causes of the VIMS have been proposed, these causes have yet to be fully investigated. In a previous study, the background of an image was shown to influence the equilibrium function of a subject viewing a 3D video clip. In this study, the authors investigated the influence of image elements such as the size of the visual field especially in the backgrounds of the projected image and velocity of the visual target speed on the body of a subject viewing a 3D video clips by measuring their body sway, the electrocardiogram signals, and the cerebral blood flow using functional near-infrared spectroscopy (fNIRS). Our results has indicated that the value of the total locus length is significantly higher while viewing a video clip whose visual fields of backgrounds is not narrowed than that with narrower backgrounds. This suggests that stability in the system to control their posture can be enhanced by reducing the amount of visual information.

<システム・計測・制御>
<メディア情報,ユーザ・インタフェース>
  • 青木 広宙
    2021 年 141 巻 9 号 p. 956-961
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    In this study, the creation of pseudo-haptics during swimming in a virtual reality (VR) space was examined. For this purpose, the user can swim in the VR space, to visualize spheres floating around them in the water. The spheres move from the front of the user to the rear as the user perform a breaststroke. Perception of the movement of these spheres can create the sensation of swimming against the flow of water. Thus, the developed system presents pseudo-haptics by controlling the amount of movement of these spheres. Four types of presentation methods were experimentally examined and compared, and their effects were verified by a psychophysical method. The results suggested that the sensation during pseudo-haptics can be finely separated into different levels by generating a constant fluid force against the user.

<音声画像処理・認識>
  • 相澤 宏旭, 村瀬 王哉, 加藤 邦人
    2021 年 141 巻 9 号 p. 962-968
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    Learning disentangled representations of emotions and identities from facial images is important for understanding human faces. This is because the human face consists of emotions and identities, and the emotions affect the face as expressions. However, because these factors interact with each other, it is difficult to learn to decompose them from facial images. Such entangled representations lead to failure in facial recognition and facial image synthesis tasks. Therefore, we aim to disentangle emotions and identities from the facial image. To achieve this goal, we propose a novel method for learning disentangled representations of emotions and identities. In the experiment, we confirmed that our method achieves to learn disentangled representation of emotions and identities, and allows us to perform an emotion-controllable image synthesis by swapping two facial images.

  • 山本 雅也, 大橋 剛介
    2021 年 141 巻 9 号 p. 969-976
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    In this study, a large amount of pseudo images for pre-training is created to achieve high accuracy even with the limited amount of ground truths in nighttime traffic sign detection. First, a large amount of pseudo images is created which does not require manual process for acquiring ground truths by randomly changing parameters of traffic sign template images with conditions based on the analysis of existing ground truths. Next, the model initialized by ImageNet weights is pre-trained with the created large amount of pseudo images, and finally the pre-trained model is fine-tuned with real images. By using a large amount of pseudo images for pre-training, suitable weights for pseudo nighttime traffic sign detection are obtained. Finally, the pre-trained weights can be adjusted for real nighttime traffic sign detection even with the limited amount of data by fine-tuning the model with real images. Experimental results show that as the detection accuracy of proposed fine-tuned model is improved compare with the model without pre-training by pseudo images, pseudo images can be used for effective training of a model.

論文
<情報通信工学>
  • 森村 俊
    2021 年 141 巻 9 号 p. 977-981
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    OPGW (OPtical fiber composite overhead Ground Wire), which is used as communication lines for electric power companies, freezes and causes communication faults in winter season because of rainwater immersion into aluminum pipe (OP unit) which envelops optical fibers. Thus, effective detection methods of water immersion are needed. Hydrogen gas is generated when corrosion reaction of water and aluminum occurs, and causes optical transmission loss in wavelengths of 1.24/1.625 µm. This article argues about a water immersion detection method using 1.24/1.625 µm-OTDR (Optical Time Domain Reflectometer) and some experiments demonstrating increasing of hydrogen gas volume and optical transmission loss in real OP units.

<生体医工学・福祉工学>
  • Mitsuru Ueda, Masafumi Kondo, Isao Kayano, Yoshihiro Sejima, Tomoyuki ...
    2021 年 141 巻 9 号 p. 982-991
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    Although the tube diameter and stenosis rate are important observations in the diagnosis using medical images of digestive organ, such as digestive systems, the diagnosis is based on the physician's subjective evaluation. On the other hand, automatic measurement of tube diameters by image processing has been used for blood vessels, which are tubular tissues similar to digestive systems, and it is also possible to visualize the vessel diameter by assigning pseudo-colors according to the length of the vessel. However, it is difficult to apply the existing diameter visualization methods directly to digestive system medical images due to the processing time and accuracy problems because the tubes are large and tortuous, unlike blood vessels. In this paper, we propose a method to measure the tube diameter quickly by dynamically calculating the next featured pixel using the history of the midpoints of the tube diameter line. Furthermore, we derive an approximate curve corresponding to the centerline of the tube from the history of the featured pixels and measure the tube diameter with high accuracy. We applied the proposed method to various types of digestive system medical images and confirmed that the proposed method could accurately measure and visualize the tube diameter along the shape of the tube and reduce the processing time by up to 99%.

  • 小杉 亘, スマリ ブライアン, 三村 將, 満倉 靖恵
    2021 年 141 巻 9 号 p. 992-998
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    In this research, we aim to extract features of menstrual cycle and mood using features of electroencephalogram (EEG). The menstrual cycle is classified into four categories: luteal phase, menstrual phase, follicular phase, and ovulation phase. We extract characteristic EEG of these periods. In addition, this study examines the extent of the difference in emotional state at the four stages of the menstrual cycle and quantitatively shows the period during which mental disorders are likely to occur.

<システム・計測・制御>
  • 松井 義弘, 綾野 秀樹, 増田 士朗, 中野 和司
    2021 年 141 巻 9 号 p. 999-1007
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    This paper gives a data-driven PID gain update method based on model matching with a reference model using closed-loop step reference data obtained from a poorly-tuned closed-loop system. The proposed method has the following features: 1) the method can suppress the influence of approximation of model matching error on the PID gain update; 2) the method can predict the responses of the closed-loop system after updating the controller; and 3) the response predictions can be useful in the evaluation of the suitability of the reference model for model matching. The effectiveness of the proposed method is shown through experiments.

  • 西野 智也, 石井 光治
    2021 年 141 巻 9 号 p. 1008-1015
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    This paper focuses on a CSMA/CA based consensus problem in which multiple agents are autonomously controlled based on the information received only from the neighbor agents. To boost the control performance, wireless communication technique should be customized considering the features of consensus problem. The aim of this work is to efficiently suppress the interference caused by multiple-access. Different from the conventional communication channel where multiple agents try to access the common destination (access point), individual agents, in the focused consensus problem, try to transmit its own information to as many neighbor agents as possible. To meet above demands, this paper proposes to adaptively change the backoff period considering the number of neighbors' neighbor agents as well as the number of its own neighbor agents. Simulation results show that the proposed CSMA/CA based consensus problem can enhance the convergence performance.

<知能,ロボティクス>
  • 張 斌, 田中 直人, 林 憲玉
    2021 年 141 巻 9 号 p. 1016-1022
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    A collision force suppression system for a human friendly robot, which consists of two arms, a trunk, a waist, an omnidirectional mobile base, is proposed. In this paper, the proposed passive collision force suppression system is realized by the collision force suppression mechanisms and the reaction system. When the collision occurs between the robot and an object, the joints of the arm and the waist are disconnected by the collision force suppression mechanism depending on the magnitude of collision forces. The reaction system is based on the strength and existence time of the collision force, and the joints are disconnected by the order of arm and waist, and the finally the mobile base will move if the reactions are not enough to decrease the passive force. The effectiveness of the proposed system is proven by experiments.

<音声画像処理・認識>
  • 伊藤 浩朗, 吉岡 理文
    2021 年 141 巻 9 号 p. 1023-1031
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    The H.264/AVC standard has better coding efficiency comparing with former video coding standards. However, it still includes temporal artifacts, known as intra flicker.

    In this paper, we propose intra flicker reduction method, which characteristic is to compensate average signal level of I picture in terms of macro block depending on the difference of average signal level of macro block between I picture and that of previous P picture. In detail, we propose 2 approaches to realize our idea. The first one is to implement our idea as post filter of decoder, which complies with the H.264/AVC standard. The second approach is to implement it in in-loop filter of both encoder and decoder, which requires a little change from the H.264/AVC standard. As our evaluation result, we confirmed that both approaches have efficiency to reduce intra flicker, 14.1[%] with the first approach, and 16.8[%] with the second approach, both as the maximum effect.

  • 上田 美凪, 松田 彩伽, 安川 生馬, 今泉 祥子
    2021 年 141 巻 9 号 p. 1032-1038
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    An RGB-D image has extra information, i.e., depth information, which represents the distance between the camera and the objects. We extend a PEE-HS (prediction-error expansion with histogram shifting) based reversible data hiding method for RGB-D images to embed the depth information into the RGB layer without visible artifacts. The proposed method achieves the compatibility between the RGB-D and normal RGB images using a single marked image containing the depth information. Our experimental results show that the marked images have high quality without serious distortion.

<ソフトコンピューティング・学習>
  • 寺田 昌弘, 皆川 拓海, 安齋 博人, 今村 誠
    2021 年 141 巻 9 号 p. 1039-1047
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    In order to automatically analyze human motion data, it is important to extract basic actions such as “stand up” and “walking”. The most conventional methods are several issues: (1) they need labeled data, (2) they only classify data but not extract the structure of data, (3) they require a window size of time series in advance. This paper proposes a novel unsupervised pattern discovery method which extracts a series of repetitive actions by recognizing basic actions in time series. The proposed method consists of a basic action enumeration process and a grammatical inference process. The former discovers the subsequence of basic actions using the Motif discovery method. The latter discovers grammatical patterns constructed with the sequence of the basic actions. Furthermore, we evaluate the accuracy of extracted basic actions and the duration time of a series of actions for packing operations and screw-tightening operations.

  • 但馬 慶行, 濱上 知樹
    2021 年 141 巻 9 号 p. 1048-1054
    発行日: 2021/09/01
    公開日: 2021/09/01
    ジャーナル 認証あり

    In this study, we propose deep learning model that not only has a high accuracy, but also a small distance between training and prediction labels for unknown data in the neighbor of the training data. We define a knowledge distribution as probability distribution for each labels. The relationship between the knowledge distributions corresponds to the distance between the labels. In the proposed method, we firstly train a model to minimize distributional distance between internal features and knowledge distribution, then train the model to classify labels correctry. We visually confirmed that the proposed method can predict labels with small distance between labels for unknown data near the training data. In addition, we compared the proposed method with a existing method that uses a typical convolutional neural network. Experiments on MNIST show that the proposed method achieved the same or better accuracy than existing methods. The proposed method also achieved mean absolute error of 0.32, while the existing method achieved 0.374.

研究開発レター
<電気回路・電子回路>
<システム・計測・制御>
部門記事
 
feedback
Top