All issues
Predecessor
23.8
Displaying 1-22 of 22 articles from this issue
- |<
- <
- 1
- >
- >|
-
Article type: Cover
Pages Cover1-
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSDownload PDF (17K) -
Article type: Index
Pages Toc1-
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSDownload PDF (122K) -
Yasuhiko Tsuge, Noboru OhnishiArticle type: Article
Session ID: IDY99-66/HIR99-1/MIP
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSSound Discrimination is basic for analyzing scene by hearing, and it's very important for us to decide what to do next. But the hearing impaired can't quickly respond to change of the surrounding environment because of their disability to sound discrimination. If the sounds are such alarm signals as crossingbell, emergencybell and so on, they will encounter a dangerous situation. We developed a system for discriminating alarm signals by using multivariate analysis. This system can discriminate five alarm signals(craction, crossingbell, emergencybell, alarmwhistle, siren). We conducted experiments with CD and real sounds, and could obtain high discrimination rate of 99%.View full abstractDownload PDF (643K) -
Mitsuru KAWAMOTO, Kiyotoshi MATSUOKA, Noboru OHNISHIArticle type: Article
Session ID: IDY99-67/HIR99-2/MIP
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSThis paper presents a method of blind separation which can separate original signals from their mixtures observed by the same number of sensors. The blind separation can be achieved without using strong assumption about source signals and any information about the mixing process, i.e., the transfer function between signal sources and sensors, except the fact that source signals are mutually statistically independent. Our proposed method can separate non-stationary signals(e.g., speech signal, music)from their mixtures. We apply our method to an experiment which extracts two source signals from their mixtures observed in a normal room.View full abstractDownload PDF (457K) -
Kazuhiro Hayakawa, Ryo Suzuki, Toshiharu Mukai, Noboru OhnishiArticle type: Article
Session ID: IDY99-68/HIR99-3/MIP
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSWhen we hear the footsteps of persons, we can visually identify the corresponding person making each footstep. Thus human being can find the correspondence of events between vision and audition. Though there are a variety of factors ranging from low to high levels to find the correspondence, we focus on the low level of physical law. In this paper, we propose a method to find the correspondence between vision and audition with a physical law of event synchronization in the situation where there exist plural moving objects, one of which makes a sound.View full abstractDownload PDF (834K) -
Ali MANSOURArticle type: Article
Session ID: IDY99-69/HIR99-4/MIP
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSIn the last two decades, High order statistics have been very important tools in signal processing field. In various studies(specially on blind identification or blind separation of sources), researchers propose algorithms or criteria based on the high order statistics(in most of the cases, the fourth order cumulants are used). Additionally, The stability or the performances results of the involved algorithms depend, in many situations, on the kurtosis of the used signals. In this paper, the relationship between the kurtosis of a signal and its probability density function(pdf)is investigated. In simple case, theoretical result proves that the value(or the sign)of the kurtosis can be predicted from the shape of the pdf. However, this result can not be generalized for more complex distributionsView full abstractDownload PDF (329K) -
Allan Kardec BarrosArticle type: Article
Session ID: IDY99-70/HIR99-5/MIP
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSOur study focuses on a new method of estimating the heart rate variability(HRV)which does not need the use of R-wave detection from the electrocardiogram(ECG). Contrary to this detection method which requires a sampling frequency higher than 100 Hz, the one proposed here could calculate the HRV from an ECG signal sampled around 5 Hz with a mean error of 0.03 beats^<-1>. The method also worked efficiently to extract the HRV from an ECG measured from a subject in exercise condition. The overall error was smaller as we low-pass filtered the heart rate variability.View full abstractDownload PDF (442K) -
Masaharu MURAKAMI, Hitoshi KIYA, Shogo MURAMATSU, Seiji KIMURA, Takahi ...Article type: Article
Session ID: IDY99-71/HIR99-6/MIP
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSIn this report, we discuss tiling wavelet tranforms in which an input image is partitioned into tiles and then the wavelet transform is executed on each tile. One of the conventional tiling methods, Spatially Segmented Wavelet Transform(SSWT), needs overlapping tiles to avoid block distortion along the boundaries of the tiles. Moreover, the overlap size of tiles must be decided by the number of stage S and the length of filters L, before executing the transformation. It approximately increases in proportion to 2^<S-1> and L. These limitations generate a lot of redundant calculations and disturb the adaptive transformations. In this report, we propose a new tiling method which gives the same wavelet coefficients as those of the SSWT, by using different approach. Thus, the proposed method does not generate the block distortion. Since the overlap size of tiles in the proposed method is independent of the number of stage S, it is possible to improve both problems of computational redundancy and adaptive transformation.View full abstractDownload PDF (490K) -
Hidetoshi YAGI, Katsuyoshi TANABE, Sakuichi OHTSUKA, Norimasa NAKANISH ...Article type: Article
Session ID: IDY99-72/HIR99-7/MIP
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSThe artery-vein diameter ratio on ocular fundus images is checked during medical check-ups, because it is an index for high blood pressure. Detection of the measurement points is one of the problems in automating this task. We propose a method of detecting the measurement points. In this method, the region most similar to that around the measurement points on the diagnosed fundus image for the same patient is searched. The search region is limited by rough registration using the optic disk, which can be automatically detected. The method was tested using 114 sample measurement points pointed out by medical doctors, and all were correctly detected.View full abstractDownload PDF (899K) -
Akira KAMEGAYA, Hirotsugu KINOSHITAArticle type: Article
Session ID: IDY99-73/HIR99-8/MIP
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSAt the present, some automatic indexing techniques are desired in the field of image retrieving. The cost of retrieving will be reduced and indexing is done uniformly using an automatic indexing. In this paper, we study a method of automatic recognition of the motion in moving images for the purpose of the image retrieving. we describes a method of the image retrieving using the moving object and the motion.View full abstractDownload PDF (780K) -
Masafumi SATO, Hirotsugu KINOSHITAArticle type: Article
Session ID: IDY99-74/HIR99-9/MIP
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSIn this paper, we propose a new digital signature system for images. This system generates the signature information based on the structured information of the image. It increases the tolerance resistance against processing and coding work. This characteristics is deployed as an image retrieval method which allows rough sketch matching. Furthermore, a data embedding method suitable for this system is also proposed. This characteristics is deployed as an image retrieval method which allows rough sketch matching.View full abstractDownload PDF (592K) -
Shoji Tanaka, Yuichi Iwadate, Seiji InokuchiArticle type: Article
Session ID: IDY99-75/HIR99-10/MI
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSDetecting interesting regions from images has become important in order to reduce the computational complexity associated with time-consuming processes such as object recognition. In this paper, we assume that the a figure-ground segmentation of a picture is necessary to extract attractive regions as the first process. On this assumption, a method for detecting figure regions from picture utilizing the feature of V4 cortex in the visual system is presented. The method detects the figure regions based on the level of contrast between a region and its surroundings. Contrast parameters are defined by the color difference and the texture difference between a region and its surroundings, as well as, the focus of the region. In our experiments, the proposed method could extract figure regions at the rate of 80% accuracy.View full abstractDownload PDF (966K) -
Kunio HATO, Katsunori YAMAOKA, Toshiyuki YOSHIDA, Yoshinori SAKAIArticle type: Article
Session ID: IDY99-76/HIR99-11/MI
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSCurrently we can get various kinds of information from the Internet. However, it is not easy to get required information within short time, especially image information. In this paper, we propose a new image retrieval system for multimedia information from the Internet and show the designed and manufactured results of a proposed system. The following two problems must be solved ; (1)Gathering images requires a large amount of data transmission, which increases network traffic. (2)To extract features from images requires a large amount of image processing capability. To avoid concentration of network traffic and image processing capability, the proposed system applies distributed configuration of system functions, which are databases, image gathering function and image filtering function. We propose the method to distribute these functions in the network and mechanism to gather feature information efficiently, considering network configuration.View full abstractDownload PDF (703K) -
Yasuko KAKIUCHI, Youichi MIZUNO, Toshiyuki YOSHIDA, Yoshinori SAKAIArticle type: Article
Session ID: IDY99-77/HIR99-12/MI
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSRecently, the demand for an image retrieval is increasing. In this paper, we propose image retrieval method based on the keyword which consists of natural language. The retrieval system consists of retrieving engine with dictionary. Dictionary has retrieval algorithm corresponding to the keyword. Retrieval algorithm consists of grouping switch, vector quantizer and N out of M(N/M) detector. The parameters of vector quantizer and N/M detector are optimized by Genetic Algorithm and Vector Quantization Algorithm. In this paper, we show the outline of the proposed retrieval system and the method of automatic generation of image retrieval algorithm. Further, we show the proposed method has the capability to generate image retrieval algorithm.View full abstractDownload PDF (588K) -
Keiichirou HOSOISHI, Ryuji HAMABEArticle type: Article
Session ID: IDY99-78/HIR99-13/MI
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSWe have already proposed Learning Motion Vector Quantization(LMVQ)that can eliminate the optimal codebook design procedure on Motion Vector Quantization(MVQ). In addition to motion estimation, this LMVQ method is simultaneously to do the codebook learning according to the nature of input image. In this paper, we will show a learning method that refers to the frequency counters of codewords at codebook learning and report the simulation results employing the Flower Garden video data. In performance evaluation of motion compensation, the SNR of this method was better than that of MVQ. In comparison with MPEG-1, although the SNR of LMVQ was about 0.4dB poorer, we could achieve the significant results such as reducing the computer amount of motion compensation to 1/11 and shortening the total processing time to 33%.View full abstractDownload PDF (874K) -
Jun FURUKAWA, Hitoshi KIYAArticle type: Article
Session ID: IDY99-79/HIR99-14/MI
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSIn MPEG4, motion vectors on both 8 × 8 blocks and 16 × 16 macro blocks can be used. In this paper, we propose a block matching algorithm which can find both of them with a small number of operations. It is shown that the number of operations for motion estimation can be reduced by expanding Succesive Elimination Algorithm(SEA)while maintaining the same accuracy of motion estimation as that of the full search algorithm. The proposed method is based on the idea that block matching algorithm is executed as the combination of smaller blocks and the results of computations in 8 × 8 blocks are reused to execute motion estimation in 16 × 16 macro blocks. In consequence of our simulation, the significance of the proposed method is verified.View full abstractDownload PDF (783K) -
Osamu ITO, Masato TSUKAHARA, Hideo HASHIMOTOArticle type: Article
Session ID: IDY99-80/HIR99-15/MI
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSVarious motion compensation methods using variable-size and variable-shape blocks have been proposed. In this paper, a new method called binary-tree method is proposed and compared with other methods under the same conditions. The proposed method has more types of blocks than the others and its block structure is described with binary tree. Computer simulation for various motion compensation has been performed, and shows that binary-tree method is the most reasonable among them in respect to prediction performance and information cost. In addition, performance is also evaluated when interframe difference error is used for block type decision criterion instead of motion compesation error.View full abstractDownload PDF (957K) -
Ichiro HAGINO, Shizuo NAKAURAArticle type: Article
Session ID: IDY99-81/HIR99-16/MI
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSIn landscape visual simulation, computer graphics and CAD are used to design buildings or bridges, and also image processing is used to modify background scenes taken from the real world. We developed the image processing techniques that remove unnecessary objects like electrio-light poles or miscellaneous signboards which make worse views, and that extract trees or other objects being used as textures. For those purposes we have developed color processing tools on the chromaticity diagrams, geometric transform with warping, texture mapping and color interpolation after unnecessary objects were removed.View full abstractDownload PDF (735K) -
Daisuke Honami, Toshimitsu Tanaka, Noboru OhnishArticle type: Article
Session ID: IDY99-82/HIR99-17/MI
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSColor variance by soil and dust is important for architectural simulation, so that their correct synthesis is necessary in order to improve reality of computer generated images. We thought soil and dust which generated by environment, frequency of use, and passage. in this paper, we presents an algorithm that generates textures of dirty floors caused by human traffic at indoor steps. We measured a distribution of walkers and calculated floor soil and dust. Then using their results, we generated textures with human traffic stain, and we were rewarded with good results.View full abstractDownload PDF (1354K) -
Yoshikatsu Date, Naoto Otsubo, Toshiyuki Sakamoto, Toru Nagamachi, Sei ...Article type: Article
Session ID: IDY99-83/HIR99-18/MI
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSNeeds of computer graphics(CG)production for TV programs has been on the increase in quantity, diversifying requirement and increasing its difficulty. Especially the need for high quality of CG image has been getting intense more and more. On the other hand, CG production for TV broadcast program has been struggling with time limitation and shortage of production costs everytime. Under the circumstances we produced more realistic CG image of an ancient temple successfully utilizing a unique surface shading method which takes 'dirt of surface' into considertion. In this paper we describe our experimental challenge and effort of generating more natural looking CG images during the production of a TV program focusing on simulating Horyu-ji temple.View full abstractDownload PDF (865K) -
Yoshihiro Yasumuro, Qian Chen, Kunihiro ChiharaArticle type: Article
Session ID: IDY99-84/HIR99-19/MI
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSThis paper describes a new approach for hand modeling. The hand model is composed of the surface part and the bone part. The surface part of the hand model is represented by polygonal surfaces which are generated by 3D data points obtained by measuring a real hand with a laser range finder. The bone part of the model is constructed based on the structure of the bone of human hand using the X-ray CT Images. The shape of the whole hand is controlled by a dynamics model which is able to evaluate the naturalness of the posture.View full abstractDownload PDF (972K) -
Article type: Appendix
Pages App1-
Published: February 02, 1999
Released on J-STAGE: June 23, 2017
CONFERENCE PROCEEDINGS FREE ACCESSDownload PDF (70K)
- |<
- <
- 1
- >
- >|