-
Article type: Cover
Pages
Cover1-
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
-
Article type: Index
Pages
Toc1-
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
-
Ryosuke SUZUKI, Atsushi YAMASHITA, Toru KANEKO, Hajime Asama
Article type: Article
Session ID: ME2012-1
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Print media like books is an important source of information. However, it is very difficult for blind people to obtain the information from print media. This research aims at making a system to read out tables and graphs which cannot be read out by existing products. In addition, there is a problem that scanners yield distortion at the seam of books. In order to solve this problem, we correct the image distortion by using stereo camera. Experimental results showed the effectiveness of the proposed method.
View full abstract
-
Takuya Oe, Atsushi Yamashita, Toru Kaneko, Susumu Ishihara, Hajime Asa ...
Article type: Article
Session ID: ME2012-2
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Currently, drive recorders and in-vehicle cameras have become popular with general automobiles. Hence, a system using images taken by such devices is required. To satisfy the demand, retrieval of the shared data in the system is important. In this paper, we propose a method for generating arbitrary viewpoint images for users based on the 3-D map which is constructed by estimating the position and pose of the camera for each viewpoint. Our method uses Bundler and Pach-based Multi-view Stereo of the image analysis software, and estimates the camera poses by Structure from Motion and generates the 3-D map by dense point groups. Experimental results showed the effectiveness of the proposed method.
View full abstract
-
Satoshi ASAI, Atsushi YAMASHITA, Toru KANEKO, Hajime ASAMA
Article type: Article
Session ID: ME2012-3
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
An autonomous robot needs functions of environment recognition and motion planning when moving to a destination. In this paper, we propose a method to realize these functions for a biped robot which cannot accomplish self locating by odometry because of unstableness of its foot movement. In our method, the robot executes visual SLAM using images acquired with stereo camera, generates a 3-D grid map, and plans appropriate path and motions to the destination. Effectiveness of the proposed method was shown by experiment.
View full abstract
-
Shingo ANDO, Junki KUBOTA, Shoji YAMAMOTO
Article type: Article
Session ID: ME2012-4
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
We have developed the manufacturing support system which can measure the feature size by using web camera. Several conventional systems, such as stereo camera, light section method, and structure light method were introduced into the actual manufacturing factory. However, these systems need the complicate calibration and strict setting in exchange for an accuracy of measurement. Therefore, we investigated the easy-to-use feature measurement which has priority to simplicity rather than accuracy. In this paper, two simple methods with ultrasonic sensor or estimation of scale factor were proposed, and we evaluated the usability and feasibility of these methods in this research.
View full abstract
-
Shinichi SAKAMOTO, Shoji YAMAMOTO
Article type: Article
Session ID: ME2012-5
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Recently, a media board and digital textbook with information technology have been introduced into the educational field. These tools are expected to give wide knowledge to student. Moreover, it is necessary to conduct profound consideration which is derived from writing exercise and representing their idea diagrammatically. Therefore, we developed an electronic notebook system with pen interface for a creative education. Our system can make the "My digital note" with several functions, which are able to access the web from important phrase and link to related notebook or textbook by using the internal search engine.
View full abstract
-
Yasuhiro KAWABE, Kazuhiro MATSUSHIMA, Mie SATO, Masao KASUGA, Naoki HA ...
Article type: Article
Session ID: ME2012-6
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Surrounding image projection gives a viewer an immersive feeling by covering him with screens. However, devices for surrounding image projection are costly and large-scale for an individual. In this paper, we aim to develop a small-scale and reasonable device that can present surrounding images. In our image projection, we place fish-eye lenses in front of projectors to expand the projected planes, and use surrounding walls and the floor as a screen. We project images on the front, back, left, right, and bottom as a surrounding image.
View full abstract
-
Toshiaki TANAKA, Michimi INOUE, Mie SATO, Masao KASUGA, Tomoharu ISHIK ...
Article type: Article
Session ID: ME2012-7
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Recently, HDR (High Dynamic Range) display has been actively studied. HDR display has a more wide dynamic range of luminance and more gradations. Therefore, response characteristics that are suitable for a HDR display and a general display are not necessarily the same. In addition, since there is a minimum luminance-difference which can be recognized in human eyes, we cannot say that the number of gradations has increased if the luminance-difference is not recognized. Our study measures the numbers of recognizable gradations along two types of response characteristic, and analyzes their features in order to examine what kind of response characteristic is suitable for HDR display.
View full abstract
-
Koichiro SUZUKI, Naoki HASHIMOTO
Article type: Article
Session ID: ME2012-8
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
A lack of recognition with blind spots of driver's view is one of causes of traffic accidents. However most of the efforts of ITS to solve the problems of road traffics are still in the research phase. Furthermore, the number of the accidents is not drastically decreased, and further efforts to improve safety of the road traffics are strongly required. Therefore, in this research, we propose a visual assistance system which can decrease the blind spots by using shared vehicle camera images with computer vision techniques.
View full abstract
-
Tomoki SUNAGA, Naoki HASHIMOTO
Article type: Article
Session ID: ME2012-9
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
In recent 3D modeling, we often make a rough shape of targets, and refine it as post processing. In this style, rich functions are not useful and usability is much more important for flexible expression. Therefore, in this research, we propose an intuitive 3D modeling system for expressing our images as 3D shapes roughly and quickly. In order to archive interactive 3D modeling, this system scans real clay shaped by a user, and converts it into a polygon model in real-time. Additionally we implement a painting function using 3D interface in real time. These functions are composed of only a depth sensor and a camera.
View full abstract
-
Shoi HAYAMA, Masaaki FUJIYOSHI, Hitoshi KIYA
Article type: Article
Session ID: ME2012-10
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
This paper proposes an embedding parameter decision for speeding up the reversible data hiding method. The reversible data hiding method focused in this paper spends up to 70% of the embedded processing time to determine an embedding parameter. The proposed method determines the embedding parameter up to 250 times faster than the conventional method, so the proposed method speeds up faster reversible data hiding.
View full abstract
-
Ryosuke NAKAMURA, Kento ISHIZUKA, Toshiyuki GOTOH, Naoyoshi TAMURA
Article type: Article
Session ID: ME2012-11
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
This paper presents the integration between spatio-temporal visual and auditory informations, in order to realize the virtual orchestra systems which synchronizes an image and music. First, we describe a developed visual and auditory integration system which enables us to display both music and CG model with any given timing. Then, we also discuss the experiment which evaluates the relationship between sensory integration and the presentation timing of vision and music, using this system.
View full abstract
-
Masaki ICHIKAWA, Toshiyuki GOTOH, Seiichiro KAGEI, Tae IWASAWA
Article type: Article
Session ID: ME2012-12
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Lung has two types of arteries, pulmonary and bronchial, which supply deoxygenated and oxygenated blood. Since the bronchial arteries are much smaller than pulmonary arteries, the blood flow in pulmonary arteries is only visible in general using contrast enhanced MRI. However, cancer cells induce capillary growth (angiogenesis), making them to supply aortal blood via bronchial arteries. In this paper, we first describe a method of cancer region evaluation using the fitting error when one-source blood flow model is applied into time-sequence contrast enhanced MRI. Then, we discuss the effects of a multiple-source blood flow model to discriminate the pulmonary and aortal blood flow.
View full abstract
-
Dai Sato, Yuma Iwao, Toshiyuki Gotoh, Hideyuki Wakai
Article type: Article
Session ID: ME2012-13
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
The large-sized vehicles which is moving in mining environments have a wide dead angle from the driver, and have potential dangers of derailment, a fall, and a collision due to the viewing situation change in the influence of the weather and/or the night. If the three-dimensional map of the haul road is generated, a dangerous position is able to be judged in advance. In this paper, we describe a method for modeling three dimensional map of haul roads using distance image camera, which can estimate three dimensional camera position and integrate the information of the successive frames.
View full abstract
-
Ryutaro NIHO, Yasuyuki SAITO
Article type: Article
Session ID: ME2012-14
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Back Ground Music (BGM) is used to enhance the impression of the image. However, we have no method of automatic recommendation in currently. In this paper, we propose an automatic recommendation system of music matching input image. To recommend a music, the system extracts some features of both the input image and the music in database. And the system learns the correspondence of the image features to the music features using Canonical Correlation Analysis. We designed the system structure and will verify the effectiveness of the proposed method.
View full abstract
-
Takahiro IIJIMA, Yasuyuki SAITO
Article type: Article
Session ID: ME2012-15
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
This study proposes a system which semi-automatically arranges an orchestra score for a duo score. To arrange an orchestra score for a duo score, we need to divide an orchestra score into solo and piano accompaniment score. However, it is too difficult to arrange an orchestra score for piano score. An orchestra score consists of several instrumental parts, therefore it needs to be arranged for piano accompaniment to play. This system arranges considered problems for playing piano and generates a piano accompaniment score. It combines a piano accompaniment score with solo part score manually extracted in advance, it generates a duo score.
View full abstract
-
Ayano Tango, Yutaka Nakano, Toshiyuki Yoshida
Article type: Article
Session ID: ME2012-16
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Printed documents with a fixed format are often modified very partially, resulting in very similar documents after modification. The authors have proposed an automatic comparison and change detection technique by using an image matching technique and an extraction of the longest common sequence. However, an application of the technique was strictly limited to documents with single column and without tables and figures. This paper thus extends the technique in order that it can be applied to multi-column documents with tables. The experimental results in this paper illustrate that the proposed technique works properly as expected.
View full abstract
-
Hisatoshi OGATA, Masaki MIZUNO, Yutaka NAKANO, Toshiyuki YOSHIDA
Article type: Article
Session ID: ME2012-17
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
We have proposed a cell tracking technique for cell image sequences in order to assist research in the area of epigenetics, which is based on cell region extraction and cell tracking including parent-child identification. However,, the tracking accuracy was not enough to treat with huge number of cells over images. This paper thus improves the technique by taking cell motion and division into account. Experimental results show that the proposed technique can keep track of more cells precisely than the previous technique.
View full abstract
-
Nobuyuki KITADE, Yutaka NAKANO, Toshiyuki YOSHIDA
Article type: Article
Session ID: ME2012-18
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
We have proposed a comparison technique for mutually similar printed documents using image matching and extraction of the longest common sequences (LCSs). In order to cope with a difference of the bounding boxes for a character, the technique employs a bounding-box concatenation technique and a new LCS extraction algorithm. However, these techniques have not been realized actually in our previous reports. This paper thus tries to implement the techniques completely and revise the matching process. Experimental results show that such an extension is useful to improve the detection accuracy with our method.
View full abstract
-
Naoki YAMADA, Hiroto INOUE, Masato MOTEGI, Katsumi ANZAI, Mie SATO, Ma ...
Article type: Article
Session ID: ME2012-19
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Recently, Japanese corporations are starting to adopt not only "lifetime employment system" but also "pay-per-performance system". However, the relationship between the employment systems and business performance is unclear and it has not been studied enough. This paper was based on a standpoint of the labor liquidity and analyzed the relationship. As a result of correlation analysis, it was shown that growth corporations were utilizing non-permanent workers to raise the labor liquidity. Also, managerial characteristics in each business field were different.
View full abstract
-
Satoru HASEZAKA, Yuuri NIIDATE, Katsumi ANZAI, Mie SATO, Masao KASUGA
Article type: Article
Session ID: ME2012-20
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Recently, financial institutions intellectual property as an investment criterion when they finance the company. However, it is difficult to evaluate corporate growth effectively with low cost. In this paper, we focused on technology that is included in the intellectual property. And we considered a simple quantitative assessment method for corporate growth based on the intellectual property.
View full abstract
-
Yuuri NIIDATE, Satoru HASEZAKA, Katsumi ANZAI, Mie SATO, Masao KASUGA
Article type: Article
Session ID: ME2012-21
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Recently, business environment of SME (Small and Medium-sized Enterprise) is very harsh. In this background, it has been rising corporate bankrupcies. A bankruptcy damages affect not only the field of business but also our livelihood. Therefore, it is exceedingly important to estimate possibility of bankruptcy and discover factors of bankruptcy. We aimed to propose a new method for estimate the bankruptcy. In this paper, we listed problems of existence method and developed new indexes that are able to estimate financial conditions more easily and correctly.
View full abstract
-
Mirai TAKAHASHI, Hiroto INOUE, Mie SATO, Masao KASUGA
Article type: Article
Session ID: ME2012-22
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Recently, design of characters has become diversified with the spread of the computer. When designing the characters, it needs to choose a font and to consider depth expression of characters. There are many types of depth representation for lettering. We pay our attention to the shade expression that is thought the strong interaction with font features. As a result of our subjective evaluation experiment, this study reports relations between characters with shade expression and their impressions.
View full abstract
-
Haruto SUZUKI, Hisae SHIMAMURA, Shota MORI, Mie SATO, Masao KASUGA, Na ...
Article type: Article
Session ID: ME2012-23
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Recently, improvement of 3D display technology has enabled us to enjoy 3D contents not only a on liquid crystal television but on a mobile display like a smart phone. When view the same 3D contents on screens of different sizes, it is thought that our impression of 3D contents is changed. In this research, an impression evaluation experiment of stereoscopic images is conducted with a smart phone and a liquid crystal television, and we examine adjustment effects of stereoscopic image suitable for mobile displays like a smart phone.
View full abstract
-
kyosuke CHIBA, Teruaki HIRANO
Article type: Article
Session ID: ME2012-24
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
An virtual placement supporting system of furniture using AR (Augmented Reality) and an algorithm of using other makers for the AR at the same time are presented. The AR is technique for emphasising the video image which was taken with web camera, by putting the CG on a video image. The furniture as 3DCG is placed in a real space by the AR, through the web camera. By using this proposed system, it is more effectively imagined that arrangement of the furniture, than imagining in a head. The furniture which is heavy like a bed, can be easier move as the 3DCG in the real space by only moving a paper printed the marker corresponding to the bed.
View full abstract
-
Kousuke FUKUYOSHI, Naoki HASHIMOTO
Article type: Article
Session ID: ME2012-25
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
In recent years, many researchers have been tried to provide immersive virtual environments with surrounding images. However, large equipment and complicated calibration of them are troublesome problem for general users. Therefore we propose a wide-field image projection system by using smartphones. This system provides compact and simple installation with the combination of a smartphone and a projector. And geometrical collection with multiple projection units is simply achieved by using smartphone's camera and CPU. Wide-field image projection is also supported by cascade calibration algorithm with the multiple smartphones.
View full abstract
-
Tasuku ISHIDA, Naoki HASHIMOTO
Article type: Article
Session ID: ME2012-26
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Recently, many kinds of devices with an interactive touch panel are often used in our daily life. These are typically flat displays. In this research, we propose an interactive balloon display. Image projection with a fish-eye lens covers the whole surface of the spherical balloon. And the interaction with our hands as touching, sliding, and slightly and hardly pushing are detected with an infrared camera, installed inside of the balloon, and becomes a trigger of some image effects. And then, everyone can intuitively enjoy the interaction with the unique softness of the balloon.
View full abstract
-
Junichi SAKURAI, Naoki HASHIMOTO
Article type: Article
Session ID: ME2012-27
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
We propose an interactive projection mapping which provides ultra-realistic effects by mixing the real and a virtual world. It is achieved by making the video presentation by the augmented reality of projection mapping in common spaces, working in sync with its audience's behavior. Sensing of the audience is carried out with a camera, and video effects are with high-quality layered videos for making various patterns. Thus, dynamic change of the real space is virtually achieved, and we can feel it as is a part of the real world.
View full abstract
-
Yuya Yamazaki, Toshiyuki Yoshida
Article type: Article
Session ID: ME2012-28
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
To obtain highest possible video quality on a band-limited network, the authors have proposed a video coding technique that maximizes the estimated spatio-temporal Mean Opinion Score(MOS), and a video-quality equalization technique based on the estimated MOS. The authors have also applied the techniques to a scalable video coding(SVC) to show its effectiveness. This paper improves the bit allocation technique in each layer together with the MOS estimation technique by considering variation of the coding efficiency in SVC. The experimental results given in this paper indicates the effective ness of the proposed technique.
View full abstract
-
Tatsuhiro YAMAGUCHI, Yutaka NAKANO, Toshiyuki YOSHIDA
Article type: Article
Session ID: ME2012-29
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
"Shape from focus/defocus" has been known as one of the 3-D shape reconstruction techniques, which estimates depth information from a set of multi-focus images. However, the conventional technique gives poor depth estimation for the surface texture with non-uniform frequency components such as one comprising two distinct frequency components. To overcome the problem, this paper proposes a new technique based on frequency band decomposition by using a set of 2-dimensional band-pass filters, The experimental results show that the proposed technique is promising.
View full abstract
-
Naoki TAKEMURA, Kazuki NISHI, Naoya SHIRAI, Yasutake KAWASHIMA
Article type: Article
Session ID: ME2012-30
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Image scanning-type barcode readers are useful tools for acquiring the object information from camera images. The built-in liquid lens system makes the focusing control easier, but has the disadvantage that the response time is slow. In this paper, the 'over-drive' technique for speeding up the response time and the 'range mapping' technique for decreasing repetitions of the read operation are proposed.
View full abstract
-
Saya MORISHITA, Makoto OKABE, Rikio ONAI
Article type: Article
Session ID: ME2012-31
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Recently, there are a lot of animations and video works with the CG. When you create these videos using existing methods, it takes a long time to add sound effects to videos. So we created a supporting system for adding sound effects to videos. This system decide on the timing of sound, volume, and sound playback time. The user study, we confirmed this system reduce the creation time.
View full abstract
-
Satoshi YOSHIDA, Masao IZUMI, Hiroshi TSUJI
Article type: Article
Session ID: ME2012-32
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
To research on the ability of a cheap device, Kinect, we try to discriminate people who are walking. We defined the movement of walking, and made feature vectors using information of 5 persons from Kinect. These feature vectors show each movements of walking. Then we made a feature space using Principle Component Analysis, and discriminated test patterns using centers and variances of each classes on the feature space. As a result, it was found that Kinect could discriminate 5 persons who are walking in about 80 percent. Then we considered the result and the way to study.
View full abstract
-
Misato MOCHIZUKI, Hiroki TAKAHASHI
Article type: Article
Session ID: ME2012-33
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Although many recipe search engines are available, few engines recommend user preferable recipes. This paper proposes a various recipe recommendation system based on (1) preferable ingredients, (2) surplus ingredients and (3) cooking categories. Preferable ingredients, which are not used recently, are strongly recommended in (1). Older ingredients are used in (2). Various cooking categories are recommended in (3). Recipes are recommended based on the non-linear combination of them. By using the non-linear combination, we can recommend user preferable recipes.
View full abstract
-
Takashi Umemura, Keita Takahashi, Shigeyuki Sakazawa, Masahide Kaneko
Article type: Article
Session ID: ME2012-34
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
In this paper, we present a 3-D caricature generation system that can emphasize individual facial features using eigen-space based analyses. We also propose a characterization method for 3-D caricatures by imitating baby scheming techniques of a professional caricature drawer.
View full abstract
-
Rin KYO, Keita TAKAHASHI, Masahide KANEKO
Article type: Article
Session ID: ME2012-35
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
In this paper, we propose a method for automatically drawing a profile caricature from a photographed profile. Conventional methods for profile caricature generation were based on extrapolations of the facial edges, resulting in limited flexibility and expressiveness. The authors' group developed a more flexible caricature generation scheme in which shapes and locations of the facial components can be controlled independently by using principle component analysis (PCA). However, the scheme was limited to frontal faces. In this study, we extended the PCA-based scheme to photographed profiles to generate profile caricatures in which individual features can be emphasized with sufficient flexibilities. We also investigated automated aesthetic assessment for profile faces by evaluating esthetic lines and nasolabial angles.
View full abstract
-
Takayuki ARAI, Keita TAKAHASHI, Masahide KANEKO
Article type: Article
Session ID: ME2012-36
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
Robots that interact with humans are required to determine appropriate timing for interruption so as not to disturb humans. In this paper, we present a delivery robot that can work with conditions where multiple people are talking; the robot should determine the appropriate timing to approach to the receiver. The processing flow can be divided into two stages. The first stage is the speaking person estimation, in which person identification using a color camera and sound source localization using a microphone array are combined to estimate who is speaking now. Then, the result of the estimation is used to update the state of the talking-state model we developed, which was formulated as a Markov model. Our model can successfully handle the situation where the receiver him/her-self is not speaking now, but may be listening to other person. Furthermore, social relations between the receiver and other people there can be reflected in the model to differentiate the importance of the conversation according to the person he/she is talking with.
View full abstract
-
Yuki Iwakami, Daishi WATABE, Hideyasu SAI, [in Japanese], Osamu NAKAMU ...
Article type: Article
Session ID: ME2012-37
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
An algorithm that improves the robustness of an ear biometrics system is proposed and targeted for use in surveillance systems based on ear biometrics. To deal with pose variation when ears are rotated in depth, Gabor jets of different poses are estimated and used as training data for a discriminant-analysis-based classifier. Experimental evaluations show the effectiveness of the proposed algorism and the possibility of improving the robustness of a single-view-based ear surveillance-system.
View full abstract
-
Emi ISHIGO, Eriko AIBA, Yusuke NAKAMURA, Atsushi ISHIDA, Noriko NAGATA
Article type: Article
Session ID: ME2012-38
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
The need for rendering woven fabrics arises frequently in computer graphics. The motion of woven cloth is generated based on physically-based dynamic simulation, where the setup of each parameter is decided ad hoc by users. In this study, we present a method for estimating the dynamics of woven cloth from the optic characteristics using the subjective impressions concerning the relationship between the optics and the motion properties.
View full abstract
-
Article type: Appendix
Pages
App1-
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
-
Article type: Appendix
Pages
App2-
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
-
Article type: Appendix
Pages
App3-
Published: February 11, 2012
Released on J-STAGE: September 21, 2017
CONFERENCE PROCEEDINGS
FREE ACCESS
