-
Mei KODAMA
Session ID: R1-1
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Video distribution methods using caching had been studied to realize more highly efficient video contents delivery. However, in a video management scheme using hierarchical data, if a calculating method of priority order uses the frequency of use, there were some problems of decreasing the efficiency of video management method (VMM). This paper proposed a VMM in consideration of the condition of usage quality. It adds the frequency itself and the high frequency for the priority value of low quality data. A quality-independent VMM as the previous methods, and order-shift VMM as low-quality-weighted VCM method were used. Moreover, the improved methods were also defined to reduce transmission data for low and high quality data. We considered the effectively of the proposed method for the condition of usage quality.
View full abstract
-
Jun-ichi KIMURA, Naohisa KOMATSU, Jiro KATTO, Mei KODAMA
Session ID: R1-2
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper presents an accurate (reliable) motion estimation using block matching for sequences including spatial periodic patterns. The investigation shows that block matching distortion caused by motion vector sampling error (DVSE) causes the motion estimation errors. Based on the investigation, a new motion estimation algorithm which use maximum of DVSE is proposed. The experimental results confirm the superiority of proposed motion estimation algorithm.
View full abstract
-
Masataka TOZUKA, Kunihiko TAKANO, Makoto OHKI, Mitsuji MATSUMOTO, Koki ...
Session ID: R1-3
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
-
Kunio OHNO
Session ID: R2-1
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Recently technical writing skill to network applications has been changed based on such smart
devices as smart phones and tablet terminals. Applications are not intellectually organized by description
but developed through the sensational visual information of screen transitions. Then the explanation skill
to GUI for smart devices based on story-telling methodology and the application trend in network
virtualization technology should be considered.
View full abstract
-
Ken Tsutsuguchi, Shunichi Yonemura
Session ID: R2-2
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, in the use of materials with static media in sign language learning, we report the results of comparative experiments that to indicate the testers the teaching images by hand-drawn illustrations, and the teaching images synthesized from the video, and then to perform the sign operation in practice. From the experiment, we confirmed that the effectiveness of the use of teaching images synthesized from video in the representation of sentences with facial expressions.
View full abstract
-
Takaaki OHKAWAUCHI, Jun OHYA
Session ID: R2-3
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper proposes a recommendation system that fits to the process of writing papers. Researchers must survey literatures when they start to write papers. However, literature surveys cause large costs for novice researchers such as undergraduate students, who are not familiar with writing papers, because they might encounter some difficulties such as “they do not know keywords to be used for literature retrievals” and “they do not know what information they should use for the retrievals.” This paper’s recommendation method extracts key-words and related terms semi-automatically while a user is writing the outline of a paper or a note at that point, so that the system recommends literatures the user might need.
View full abstract
-
Akie NAITO, Yuki SHIGETO, Shigekazu SAKAI
Session ID: R2-4
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
-
Kunihiko TAKANO, Takuro ABE, Hiroshi KUWASAKI, Yukimune MATSUMOTO, Jyu ...
Session ID: R3-1
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we propose a projection of the Holographic 3D images on the mist screen enclosed in the thin soap bubble film. We show that a more stable projected 3D images are observed with full parallax, although there is a limit to the time the image can be displayed.
View full abstract
-
Kunihiko TAKANO, Hiroshi KUWASAKI, Koki SATO, Makoto OHKI, Kikuo ASAI
Session ID: R3-2
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we studied the method of the stabilization of projected image in the multi-planar volumetric mist screen. As this results, even if the winds hit the mist screen from the side, the fluctuation of the projected images become inconspicuous.
View full abstract
-
Kunihiko TAKANO, Yukimune MATSUMOTO, Jyunpei ANZAI, Koki SATO, Makoto ...
Session ID: R3-3
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we propose the holographic 3D display system using waterfall screen consist of viscous medium. We show that a more stable waterfall screen is produced by our method.
View full abstract
-
Makoto OHKI, Kunihiko TAKANO, Masataka TOZUKA, Koki SATO, Shogo KOZAKI
Session ID: R3-4
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this report, the electromagnetic diffracted field in the pyramid horn antenna is calculated under the Fresnel approximation. As a result, the complex hologram is made by diffracted electromagnetic wave on the computer. We considered that this is one of the computer generated hologram. Moreover, the wave source is numerically reconstructed as a result of Fourier transform and inverse Fourier transform on the horn aperture.
View full abstract
-
Yuguan XING, Hiroshi NAGAHASHI, Xiaolin ZHANG
Session ID: R4-1
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper presents a new method for estimating the external parameters of wide-area active surveillance cameras. In this method, a L-shaped tool is used as a gauge, and the optical axis of the camera is adjusted to each endpoint of the L-shaped tool. Then we can use only the information on the optical axis to calculate the external parameters, so that the accuracy is hard to be influenced by the distortion of lens or the error of image processing. This method is accurate.
View full abstract
-
Shuhei TARASHIMA, Go IRIE, Hiroyuki ARAI, Yukinobu TANIGUCHI
Session ID: R4-2
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
-
Hideki KOMAGATA, Yasuka OHNUKI, Masahiro ISHIKAWA, Naoki KOBAYASHI, Ta ...
Session ID: R4-3
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Cytodiagnosis is an important diagnostic to determine the presence or absence of lesions, and it is comprehensively diagnosed from such state of cell and nucleus by the pathologist. However, an experienced pathologist is required for an accurate diagnosis, and quantitative indicators are expected. In this study, we propose an eccentric quantitation approach for chromatin distribution using a concentration variance and a contour complexity variance of the nucleus divided into four parts. By using the microscope image of the cytological specimen of cervical cancer, we was examined whether our approach is possible to distinguish between five cell classification (normal, metaplastic, regenerative, CIS and SCC cells). As a result, it was confirmed that the discrimination rate is enhanced by the incorporation of our proposed method.
View full abstract
-
Natsuho YAMAMOTO, Anna NISHIBU, Sanghyun KIM, Hiroyuki MORIKAWA, Reiko ...
Session ID: R4-4
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this study, the authors focused on the partially converted stereoscopic images for advertising, and performed an experiment to examine the visual characteristics using lenticular printed materials.
View full abstract
-
Eisuke YAMAMURA, Sanghyun KIM , Hiroyuki MORIKAWA, Reiko MITSUYA, Taka ...
Session ID: R4-5
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this study, a subjective evaluation was carried out to examine the psychological effects caused by convergence operation in stereoscopic (3D) shooting.
View full abstract
-
Taira TOMITA, [in Japanese], [in Japanese], [in Japanese], Hirohisa SH ...
Session ID: R4-6
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this study, the authors developed enhancement techniques for depth sensations of 2D to stereoscopic (3D) converted image archives, and performed 3D content creation and evaluation in terms of the effectiveness of the techniques.
View full abstract
-
Shinji OKUMURA, Hiroto INOUE, Nobuji TETSUTANI
Session ID: S1-1
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Specific design patterns produce motion perception in spite of still images. The “Rotating snakes” is a representative example of these motion illusions and it is grouped into the optimized Fraser-Wilcox illusion. The purpose of this study is to clarify the mechanism of the “Rotating snakes.” This study examines a relationship among retinal eccentricity, contrast sensitivity function, and perception of the illusory motion.
View full abstract
-
Koki SASHINO, Shinji OKUMURA, Hiroto INOUE, Nobuji TETSUTANI
Session ID: S1-2
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
“Vection” and “induced motion” are illusion of motion. These illusions have common characteristics as following ; persons perceive the se illusions when something moves around a visual target . On the otherhand, eye movements are generally suppressed when persons gaze fixedly at something. This phenomenon suggests that visual fields are narrowed by degree of attention. Consequently, the authors have considered that the induction of vection and induced motion are suppressed when persons concentrate attention on objects in central visual fields. The purpose of this study is to clarify a relationship between degree of attention and these illusion of motion . In this paper, we assess some video conten ts subjectively and analyze impressions for vection.
View full abstract
-
Saya NAKABAYASHI, Nobuji TETSUTANI
Session ID: S1-3
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Subtitles of video content use assist of sounds, language translation and silent videos. Textual information such as subtitles generally displays the bottom of the monitor . However, subtitles that are displayed the bottom of the monitor are difficult to read because these are displayed in peripheral vision.Therefore, display method of subtitles need t o be optimized according to viewing condition such as when viewing on large monitor . In t his study, we study about monitor size and readability of subtitles, divide into medium monitor groups and large monitor groups.
View full abstract
-
Tadaki KAWAMURA, Yusuke IMAMURA, Mitsuji MATSUMOTO, Masao YANAGISAWA
Session ID: S2-1
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we consider the MAC layer protocol suitable for VLC LAN. Then we simulate and evaluate these system using CSMA / CD and PCF scheme.
View full abstract
-
Minglei LIU, Mitsuji MATSUMOTO
Session ID: S2-2
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we tested and evaluated the performance of the usual QR code system which has been widely used today, and then replaced it by using a LED array for data transmission. The results shows that the recognition rate of the system by using LED array is greater than 98%, under indoor condition, with communication distance of 4 meters. We also showed that the LED array system is expected to be the most appropriate method for indoor wireless optical communication, as the ulter-high-speed data transfer rate can be achieved by adjusting the flicker frequency of the LED.
View full abstract
-
Yusuke KIKUCHI, Saya NAKABAYASHI, Nobuji TETSUTANI
Session ID: S3-1
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
The authors are developing a system such as watched users can adjust protection level of privacy. In this paper, we propose protect privacy to adjust the focus ring on the camera, and describe “presence system” that application is expected. We propose the issue of the basic algorithm to decide the number of people from the blurred image, and assess the improved algorithm.
View full abstract
-
Masaaki KAJIHARA, Masahiro TAKATSUKA, Yutaka HIRAKAWA, Kazuo OHZEKI
Session ID: S3-2
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
When we run systematic block diffusion, the image of brightness value will overflow. So we have a problem that we can’t embed digital watermarking.Previous studies adjust brightness value for embedding the watermark.In this paper we perform to embed watermark with systematic block diffusion without using the adjustment.
View full abstract
-
Masahiro TAKATSUKA, Masaaki KAJIHARA, Yutaka HIRAKAWA, Kazuo OHZEKI
Session ID: S3-3
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Digital watermarking attract the attention for protection method of digital content. In this work, We propose a new image watermarking which have robustness for compression and clipping. Proposed embedding method using DCT and quantization achieves robustness for compression. The proposed method divides an input image into small blocks and then embeds watermark into all the blocks, which brings robustness for a clipping attack.
View full abstract
-
Hikaru HAGIHARA, Kiyoaki ITOI, Yukio KOBAYASHI
Session ID: S3-4
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Some danger for the password. For example spoofing or loss. W e combined the biometric technology to signature. In order to reliably determine the authenticity of the signature in the machine.
View full abstract
-
Kamei KO, Youngha CHANG, Nobuhiko MUKAI
Session ID: S4-1
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
We report the result of the experiments that how many faces of cats, which are one of the normal pets in Japan, can be detected by using Haar-like features.
View full abstract
-
Zhao WANG, Ye LI, Jing YAN , Jun OHYA
Session ID: S4-2
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Fingering is one of the most important aspects in guitar learning. As guitar is a string instrument and the same pitch notes could be at different places on fretboard, it is incomprehensive to check the fingering information of player only through audio signal, visual information is also needed to the analysis.This paper mainly proposed an algorithm of detecting and tracking guitar neck within the research of guitar teaching system based on Kineck [1]: With the information extracted from this method, we can further our research which could finally recognize whether player of guitar pressed the right chord or not automatically.
View full abstract
-
Yusuke NATSUI, Tomoharu NAGAO
Session ID: S4-3
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Super-resolution to generate a high-resolution image from a low-resolution image is required. Single frame super-resolution is of wide application, but the computational cost and image quality has become a trade-off. We proposed a method for constructing super-resolution using an evolutionary computation, and shows good results for this trade-off. However, using all the pixels in the learning image when constructing, it is considered similar parts is in the image. We consider efficient construction method by selecting the appropriate pixels. As a result, it was possible to learning efficiently.
View full abstract
-
Michihiro FUKABORI, Tomoharu NAGAO
Session ID: S4-4
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In recent years, the needs of region extraction is increasing in various fields. However, these methods depend on the specific object and the construction of these methods take much time and labor. Therefore, automatic construction of region extraction method is needed. From the viewpoint of automatic construction of image processing, Automatic Construction of Tree-structural Image Transformations (ACTIT) has been proposed. In this study, we propose a method of automatic construction of region extraction using ACTIT with AdaBoost.
View full abstract
-
Ryota SUETSUGU, Yoshiki MIZUKAMI , Katsumi TADAMURA
Session ID: S5-1
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we propose a system for supporting automatically locating flames and smokes which are used to express a house fire in the virtual space. The evacuation drill using the virtual space is acknowledged that it is useful to the disaster reduction because it allows us to repeatedly have trainings of the decision making under the similar situation of the actual disaster. Here, in order to construct an outdoor large-scale-fire evacuation drill simulator, it is necessary to efficiently represent many house fires which occur simultaneously. We propose a simple method for generating video textures consisting of both flame and smoke corresponding to various combustion stages as the solution for this problem.
View full abstract
-
Makoto Enomoto, Yuzhen Wu, Jun Ohya
Session ID: S5-2
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper explores conditions for accurate generation of portraits and how and whether noses should be represented in the portraits. ASM is applied to the original face image so that feature points in the face are detected. Using training portraits drawn by illustrators and their original faces, neural networks are trained. An unknown person’ face’s feature points are inputted to the trained neural networks so that that person’s portrait is automatically generated. As a result of experiments, multiple neural networks assigned to facial components tend to give smaller errors between the generated and manually drawn portraits than a single neural network assigned to the entire face. And the result of the compare test between portraits with nose and portraits without nose show that it is necessary to generate nose in portraits.
View full abstract
-
Lin Zhang, Jun OHYA
Session ID: S5-3
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
-
Shintaro MASUI, Chamidu ATUPELAGE, Kota AOKI, Hiroshi NAGAHASHI
Session ID: S5-4
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
-
Tatsuya MATSUSHITA, Zdeněk PROCHÁZKA
Session ID: S6-1
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Recent car safety systems exploit not only internal information about state of engine or breaks, but also external information about surroundings of the vehicle. This study deals with vision based estimation of traffic lane center line. By proper estimation of the center line we can gather various information usable for car safety system. A method described in this study is based on analysis of road region within image, and doesn’t necessary require surface marking. A two estimation methods, Kalman filter and particle filter respectively, are applied to three types of center line model, and results of estimation for each approach are discussed.
View full abstract
-
Shota ORIHASHI, Ryoki TAKADA, Jiro KATTO
Session ID: S6-2
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we consider image mosaicing using images by on-vehicle cameras. When we match feature points in image mosaicing, we limited matching range based on camera position. We also limited matching range to the foreground area. Consequently, we improved success count of mosaicing.
View full abstract
-
Ryoki TAKADA, Shota ORIHASHI, Jiro KATTO
Session ID: S6-3
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Traffic sign recognition contributes the safety of driver and people. There is a problem that the images taken by vehicle camera are blurred and distorted. In this paper, we propose the method that transforms to perfect circle from ellipse and does template matching with distorted template. As a result, we extracted 90.83% circular traffic sign and recognized 60.00%.
View full abstract
-
Yuri HIROKANE, Tomoharu NAGAO
Session ID: S6-4
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper , I propose to perform classification of pedestrian attribute . To determine the attribute, local features are needed. But we can’t successfully extract features if feature extraction region is fixed. Therefore, I determine the appropriate division of each attribute by optimizing feature extraction region using genetic algorithm.
View full abstract
-
Ryo YOKOYAMA, Kota AOKI, Hiroshi NAGAHASHI
Session ID: S7-1
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
We propose a dense motion analysis method of ultrasound image. The motion analysis is implemented by tracking the lattice points. In this study, two novel processing are performed to ultrasound image. One is to track the lattice points by optical flow intr oduced multiple spring models. The other is to detect the lattice points based on texture information with confidence coefficient, the detector corrects the tracking errors. We evaluated our method using a simulated ultrasound image of 5 min. The mean error and maximum error of our proposed method achieved the best performance in the conventional methods.
View full abstract
-
Junichi MORI, Tomoharu NAGAO
Session ID: S7-2
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Automated anomaly detection is needed in areas such as defect inspection of industrial products, Analysis of security camera, and automatic diagnosis in the medical field. We propose normal learning method for anomaly detection like human’s habituation and incompatibility in order to reduce the cost to redesign algorithm and deal with various anomalies.
View full abstract
-
Yuichi HIRANO, Tomoharu NAGAO
Session ID: S7-3
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Automatic image classification process of using the pattern recognition technology has been working extensively in various fields such as industrial and medical. However, selection of the image features and appropriate pre-processing for the input image is required. Then, We propose Automatic Classification System (ACSYS) by optimizing structure of image processing filters and image features using genetic algorithm.
View full abstract
-
Ken SHIMAZAKI, Tomoharu NAGAO
Session ID: S7-4
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper presents a method of semantic segmentation considering location and co-occurrence in natural outdoor scene. Before recognizing objects in image, we classify image in terms of scene using Support Vector Machine (SVM), and recognize objects using Semantic Texton Forests (STF). In addition, we consider the location and co-occurrence of object in scene. Experimental results show that our proposed system is effective in several categories.
View full abstract
-
Koji KAINOU, Hotaka TAKIZAWA, Mayumi AOYAGI, Nobuo EZAKI, Shinji MIZUN ...
Session ID: S8-1
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This report proposes a network-based robot that can perform the interactive ADL training for the visually impaired. The robot has a pan-tilt head equipped with a camera, a microphone and a parametric speaker that can generate ultrasonic sounds. The ultrasonic sounds are converted to audible sounds on object surfaces. The visually impaired can hear the sounds as if they are generated by the objects. By use of this mechanism, the robot can tell the visually impaired the object positions. This robot system can be applied to the ADL training where visually impaired individuals learn to use various objects arranged in living environments. In this report, the experimental results of user study are shown.
View full abstract
-
Yusuke KURAMOCHI, Hotaka TAKIZAWA, Mayumi AOYAGI, Nobuo EZAKI, Shinji ...
Session ID: S8-2
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
We've been developing the Kinect cane system for helping the visually impaired find an object that he/she wishes to utilize. In our previous works, it was reported that the system could recognize chairs, upstairs and downstairs. However, it is necessary to be able to recognize other objects. In this paper, we propose new methods of recognizing elevators and handrails. The new methods are applied to actual scenes, and the experimental results are shown.
View full abstract
-
Kazuma YADA, Kiyoaki ITOI, Yukio KOBAYASHI
Session ID: S8-3
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In recent years, tends to increase the visually impaired. In addition, products with electronic devices of
the visually impaired for auxiliary equipment such as white cane has come out. The wearer the distance to
the obstacle walls and by focusing on the study of Konstanz University In this study, we also detected to
direction not only to measure the distance to the obstacle or wall, to change the vibration time of the
vibrator I consider to manufacture a device for informing.
View full abstract
-
Takahiro FUKAYA , Natsuki TAKAYAMA, Hiroki TAKAHASHI
Session ID: S8-4
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Gaze input has a problem which is called Midas Touch Problem. It is a problem which the system is unable to discriminate the intention of the user from eye-gaze. Authors propose a method which uses the screen button and saccade in order to solve the Midas Touch Problem. This paper evaluates the accuracy of the method when a user selects icons. Two sizes of an icon are used in the paper to measure an error rate. One size is decided by the accuracy of a gaze measuring device. Other size is decided from the icon which is used in handheld devices. Moreover, This paper evaluates the saccade undetected rate when a saccade threshold is changed.
View full abstract
-
Tsubasa Ishii , Shunichi Yonemura
Session ID: S8-5
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
There are many person having difficulty in construction of human relations, because they cannot
take the communication in the first meeting well. The causes include a point that skills of the conversation are not
enough. In this study, it is intended that I build the system which supports smooth communication in the first
meeting by solving a cause ,“topic switch does not go well” ,“a topic does not develop” .Therefore I clarified what
kind of effect you had on a conversation at the time of the first meeting by the offer of the common topic. As a
result, silence in case of the topic switch decreased by a topic offer.
View full abstract
-
Ayaka NAKAJIMA, Ryosuke ENDO, Shunsuke NAGAYA, Yoshinori INABA, Shiger ...
Session ID: P-1
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this study, we constructed a system of real-time recognition of finger shapes through range images obtained by Kinect sensor , and evaluated accuracy of this system. We used 8 finger shapes as dictionary, and discriminated by k-nearest neighbor algorithm. As a result, the accuracy improved by adjusting number of k.
View full abstract
-
Shunsuke NAGATA , Yudai ARAI, Yoshinori INABA, Shigeru AKAM ATSU
Session ID: P-2
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this study, we generated dynamic 3D facial expressions that were obtained by a motion capture system that used actual facial motions. The low- dimensional parameters of the facial motions and the 3D shapes of faces were obtained by applying Principal Component Analysis. We calculated an estimation matrix that transformed the facial motion parameters to the 3D shapes of the face parameters is calculated by equalizing both of them with a least squares method. Using this estimation matrix, we transformed facial motions from neutral into the generation of dynamic 3D facial expressions.
View full abstract
-
Selection of age sensitive features based on the ratio of between-class to within-class variances
Hirofumi ASAMIZU, Yoshinori INABA, Shigeru AKAMATSU
Session ID: P-3
Published: 2014
Released on J-STAGE: August 31, 2021
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this study, we developed an estimation method for a face’s age that uses the Gabor features extracted from
face images. W e selected more a ge sensitive ones to form a feature vector of the reduced dimensions. In this feature
selection process, we introduced a criterion based on the ratio of between class to within class variances.
View full abstract