Recently, many researchers have studied a visual perception. Focus is attended to studies of the visual perception phenomenon by using the grating stimulation images. The previous researches have suggested that a subset of retinal ganglion cells responds to motion in the receptive field center, but only if the wider surround moves with a different trajectory. We discuss the function of human retina, and measure and analysis EEG(electroencephalography) of a normal subject who looks on grating stimulation images. We confirmed the visual perception of human by EEG signal analysis. We also have obtained that a sinusoidal grating stimulation was given, asymmetry was observed the α wave element in EEG of the symmetric part in a left hemisphere and a right hemisphere of the brain. Therefore, it is presumed that projected image is even when the still picture is seen and the image projected onto retinas of right and left eyes is not even for the dynamic scene. It evaluated it by taking the envelope curve for the detected α wave, and using the average and standard deviation.

抄録全体を表示

This study uses one image with a echocardiography and detects arteriosclerosis on the common carotid artery by tissue characterization. The conventional methods are measurement of thickness on the vessel wall or stiffness on the common carotid artery. However, they need frames of several beats and great calculation times. But, we detect the arteriosclerosis with only one image. Our method estimate the grade of arteriosclerosis by AR model parameters fitted to the texture on the lumen-intima wall. Experimental results show the validity of the proposed method.

抄録全体を表示

This paper describes an automatic creation technique of background music track data for given video file. Our proposed system is based on a novel BGM synthesizer, called “Matrix Music Player”, which can produce 3125 kinds of high-quality BGM contents by dynamically mixing 5 audio files, which are freely selected from total 25 audio waveform files. In order to retrieve appropriate BGM mixing patterns, we have constructed an acoustic analysis database, which records acoustic features of total 3125 synthesized patterns. Developing a video analyzer which generates image parameters of given video data and converts them to acoustic parameters, we will access the acoustic analysis database and retrieve an appropriate synthesized BGM signal, which can be included in the audio track of the source video file. Based on our proposed method, we have tried BGM synthesis experiments using several around 20-second video clips. The automatically inserted BGM audio streams for all of our given video clips have been objectively acceptable. In this paper, we describe briefly our proposed BGM synthesized method and its experimental results.

抄録全体を表示

A simple method based on the EZW algorithm is presented for improving image compression performance. Recent success in wavelet image coding is mainly attributed to recognition of the importance of data organization and representation. There have been several very competitive wavelet coders developed, namely, Shapiro's EZW(Embedded Zerotree Wavelets)⁽¹⁾, Said and Pearlman's SPIHT(Set Partitioning In Hierarchical Trees)⁽²⁾, and Bing-Bing Chai's SLCCA(Significance-Linked Connected Component Analysis for Wavelet Image Coding)⁽³⁾. The EZW algorithm is based on five key concepts: (1) a DWT(Discrete Wavelet Transform) or hierarchical subband decomposition, (2) prediction of the absence of significant information across scales by exploiting self-similarity inherent in images, (3) entropy-coded successive-approximation quantization, (4) universal lossless data compression which is achieved via adaptive arithmetic coding. and (5) DWT coefficients' degeneration from high scale subbands to low scale subbands. In this paper, we have improved the self-similarity statistical characteristic in concept (5) and present a progressive image compression method.

抄録全体を表示

An improved algorithm for extraction of faces with spectacles is presented in this paper. To increase the recognition accuracy of human faces, the faces must be isolated precisely. In the conventional system, however, faces with spectacles were occasionally not exactly extracted because of the presence of spectacles. From this point of view, an improved algorithm for isolating spectacles and faces is presented. From the experimental results, correct extraction of faces with spectacles was obtained with high accuracy by the proposed technique.

抄録全体を表示

Volume intersection is one of the simplest techniques for reconstructing 3-D shape from 2-D silhouettes. 3D shapes can be reconstructed from multiple view images by back-projecting them from the corresponding viewpoints and intersecting the resulting solid cones. The camera position and orientation (extrinsic camera parameters) of each viewpoint with respect to the object are needed to accomplish reconstruction. However, even a little variation in the camera parameters makes the reconstructed 3-D shape smaller than that with the exact parameters. The problem of optimizing camera parameters deals with determining exact ones based on multiple silhouette images and approximate ones. This paper examines attempts to optimize camera parameters by reconstructing a 3-D shape via the concept of volume intersection and then maximizing the volume of the 3-D shape. We have tested the proposed method to optimize the camera parameters using a VRML model. In experiments we apply the downhill simplex method to optimize them. The results of experiments show that the maximized volume of the reconstructed 3-D shape is one of the criteria to optimize camera parameters in camera arrangement like this experiment.

抄録全体を表示

When charged bodies such as a human body and a tool wagon etc. move near an electronic equipment of a partially opened metal box, an electrostatically induced voltage appears in the metal box. The induced voltage causes a breakdown of a device or a malfunction of the equipment. In this study, the author measured the induced voltage in the box using a sphere gap and an electromagnetic wave sensor. This paper presents that the author measures the induced voltage with changing the area of two metal foils in the box. As the results, the author clears that the induced voltage increases up to 1.6 times when the area of induction electrode is 10 times (100cm²) larger than that of 10cm², and the induced voltage doesn't change so far as the area of an ungrounded metal foil is not changed. The fruit will be useful for designing the printed circuit board as a basic study for preventing the malfunction or the failure of the electronic equipment of the partially opened metal box.

抄録全体を表示

Cooperated multi robots system has much dominance in comparison with single robot system. Multi robots system is able to adapt to various circumstances and has a flexibility for variation of tasks. Robots are necessary that build a cooperative relations and acts as an organization to attain a purpose in multi robots system. Then, group behavior of insects which doesn't have advanced ability is observed. For example, ants called a sociality insect emerge systematic activities by the interaction with using a very simple way. Though ants make a communication with chemical matter, a human plans a communication by words and gestures. In this paper, we paid attention to the interaction based on psychological viewpoint. And a human's emotion model was used for the parameter which became a base of the motion planning of robots. These robots were made to do both-way action in test field with obstacle. As a result, a burden sharing like guide or carrier was seen even though those had a simple setup.

抄録全体を表示

This paper proposes an algorithm for decomposing a multi-class classification problem into a set of two-class classification problems. The algorithm divides a set of input pattern vectors corresponding to each class into subsets according to the distribution of the selected input pattern vectors. The distribution is represented by Gaussian mixture models which are estimated by EM algorithm with MDL criterion. In this paper, the algorithm applied for constructing a modular neural network. Experimental results showed that the algorithm simplifies multi-class classification problems efficiently.

抄録全体を表示

Japanese electric power companies currently utilize existing equipments completely and maintain facilities effectively. Human experts presently judge various hardwares whether they are be reusable or not to utilize equipments completely. Especially, this paper considers about crossarm reuse judgement. This judgement is based on rust, which attaches on crossarms, by human experts. However, this judgement depends on human expertise and it is difficult to keep constant judgement accuracy. Electric power companies want to take constant and good judgement accuracy. Therefore, we develop a crossarm reuse judgement system based on rust images using machine learning techniques. The system consists of commercial microscope and standard note PC to keep the cost. And we estimate the judgement accuracy of various pattern classification methods without the special image processing such as extracting features. The results show that support vector machine is the most suitable method for this judgement system.

抄録全体を表示

Genetic Network Programming (GNP) has been proposed as a new method of evolutionary computations and its basic characteristics are studied. GNP is constructed using graph structures whose gene consists of directed graphs, so it is possible to search solutions effectively especially for the dynamic real world problems due to the implicit memory function of its structure. Also GNP can easily implement a priori knowledge into its structure as functional nodes. There has been no example where GNP is applied to real world systems until now. In this paper, Elevator Group Supervisory Control System (EGSCS) which is a typical real world system is studied using GNP. From the simulations, the availability of GNP to EGSCS is confirmed and it is clarified that the proposed method can show better performances than other conventional methods.

抄録全体を表示

It has been pointed out that if the social configuration of the three relations (market, communal and obligatory relations) is not balanced, a market based society as a total system fails. Using multi-agent simulations, this paper shows that a sustainable society is formed when all three relations are integrated and function respectively.
When agent trades are based on the market mechanism (i.e., agents act in their own interest and thus only market relations exist), weak agents who cannot perform transactions die. If a compulsory tax is imposed to enable all weak agents to survive (i.e., obligatory relations exist), then the fiscal deficit increases.
On the other hand, if agents who have excess income undertake the unselfish action of distributing their surplus to the weak agents (i.e., communal relations exist), then trade volume increases. It is shown that the existence of unselfish agents is necessary for the realization of a sustainable society.
However, the survival of all agents is difficult in a communal society. In an artificial society, for all agents survive and fiscal balance to be maintained, all three social relations need to be fully integrated. These results show that adjusting the balance of the three social relations well lead to the realization of a sustainable society.

抄録全体を表示

We propose a method for quantification of liver tumors using statistical characteristics of ultrasonic RF signals. Parameters of K distribution show the characteristics of a soft tissue. The proposed method of chi-square test(CST) can estimate the most appropriate parameters of K distribution, which are fitted to an observed amplitude distribution. Method of moment(MOM), method combining maximum likelihood and MOM(ML/MOM), and CST are compared on simulated RF signals. The CST is applied to RF signals of liver tumors including 38 hepatocellular carcinomas(malignancy) and 12 hepatic hemangiomas(benignancy). The accuracy of discriminant analysis are 78% and 50% for malignancy and benignancy, respectively.

抄録全体を表示

In recent years, the implementation of a virtual laboratory on the Web server over the Internet has attracted a great deal of public attention. The proposed “Streaming System for Experiment of Single Pendulum" uses the CGI(Common Gateway Interface) and the streaming technology in order to control the laboratory equipments and broadcast the streaming live video through the Internet.

抄録全体を表示

In an acoustic echo canceller, it is necessary to consider the nonlinearity of an echo path in order to attenuate an echo signal conventionally, and the various methods using adaptive Volterra filters are proposed. Many of those methods have not resulted in the decisive solution about the double talk states in a nonlinear acoustic echo canceller. In this paper, we propose the echo canceller algorithm for adaptive Volterra filters using the idea of the OECLMS (Orthogonal Expanded Correlation Least Mean Squares) algorithm in consideration of the double talk problem. The convergence speed of the proposed method is improved by using Gram-Schmidt orthogonalization for the input signal.

抄録全体を表示

Interpolation of non-uniform samples is required for various cases of signal processing. In such a case, we often use sampling functions to interpolate signals. We show one example of sampling functions to interpolate signals from non-uniform samples, inheriting the properties of the Fluency sampling functions introduced by Toraichi et al. The proposed sampling function is locally supported and composed with piecewise polynomial functions of degree 2 as the Fluency sampling function. In this paper, we extend the Fluency sampling function and derive the non-uniform sampling function. By using this sampling function, we get smooth and small undulate signal from arbitrary samples.

抄録全体を表示

In this paper, the distinctive tones of Thai in running speech are studied. We present rules to synthesize F₀ contours of Thai tones in running speech by using the generative model of F₀ contours. Along with our method, the pitch contours of Thai polysyllabic words, both disyllabic and trisyllabic words, were analyzed. The coarticulation effect of Thai tones in running speech were found. Based on the analysis of the polysyllabic words using this model, rules are derived and applied to synthesize Thai polysyllabic tone sequences. We performed listening tests to evaluate intelligibility of the rules for Thai tones generation. The average intelligibility scores became 98.8%, and 96.6% for disyllabic and trisyllabic words, respectively. From these result, the rule of the tones' generation was shown to be effective. Furthermore, we constructed the connecting rules to synthesize suprasegmental F₀ contours using the trisyllable training rules' parameters. The parameters of the first, the third, and the second syllables were selected and assigned to the initial, the ending, and the remaining syllables in a sentence, respectively. Even such a simple rule, the synthesized phrases/senetences were completely identified in listening tests. The MOSs (Mean Opinion Score) was 3.50 while the original and analysis/synthesis samples were 4.82 and 3.59, respectively.

抄録全体を表示

The X-ray scene image of a brain blood vessel of a rat can be shot by using SPring-8 that is a high-performance X-ray sensing system. However, the contrast of the scene image is not enough visually. Even if the contrast of a scene image is poor, a single image that has good contrast can be generated by accumulating all frames in the scene image. However, objects in the scene image must be fixed to use the method. This paper proposes a method to generate a single image with good contrast from X-ray scene images of brain blood vessel of a rat. The partial optimal frames that are useful for generating an output image with good contrast are selected in all frames and the selected frames are accumulated after optimal two-dimensional moving. The genetic algorithm is applied to optimize the selection and moving of frames. The experimental results show that the generated output image by the proposed method had good contrast, clearness and fewer noises in comparison with the generated image by the conventional method.

抄録全体を表示

In this paper, an algorithm for solving the algebraic Riccati equation (ARE) that has an indefinite sign quadratic term related to weakly-coupled large scale systems is investigated. A novel contribution is that a new iterative algorithm is derived by combining Newton's method and the fixed point algorithm. As a result, for sufficiently small ε, we can obtain the ARE solution with a quadratic convergence rate. Moreover, it is possible to calculate the ARE solution by the same dimension of each subsystem. As another important feature, the algorithm for solving the filtering ARE is also discussed. Finally, in order to demonstrate the efficiency of the proposed algorithm, the numerical example is given.

抄録全体を表示

This paper describes the calculation of deformation of a flexible contact sensor. Expressions to calculate the deformation are derived using Castigliano's theorem. The deformation is calculated in real-time with the use of measured contact forces. Accuracy of the calculated deformation is confirmed experimentally. The experimental results show that the calculated deformation agrees approximately with the actual deformation.
With the aim of applying this calculated deformation to certain robot tasks, the sensor is attached on a robot hand and the mechanical impedance parameters of an object which is grasped by the robot hand is estimated with the use of a Kalman filter. This experimenta demonstrates that these parameters are stably estimated exactly.

抄録全体を表示

This paper describes a novel trajectory control scheme for power assisted wheelchair. Human input torque patterns are always intermittent in power assisted wheelchairs, therefore, the suitable trajectories must be generated also after the human decreases his/her input torque. This paper tries to solve this significant problem based on minimum jerk model minimizing the changing rate of acceleration. The proposed control system based on minimum jerk trajectory is expected to improve the ride quality, stability and safety. Some experiments show the effectiveness of the proposed method.

抄録全体を表示

In this paper, a reinforcement learning method is proposed that optimizes passenger service in elevator group systems. Task-oriented reinforcement learning using multiple agents is applied in the control system in allocating immediate landing calls to the elevators and operating them intelligently in attaining better service in this stochastic dynamic domain. The proposed system shows better adaptive performance in different traffic profiles with faster convergence compared to the other learning elevator group control system.

抄録全体を表示

In this paper we describe an information filtering system using the Kullback-Leibler divergence. To cope with information flood, many information filtering systems have been proposed up to now. Since almost all information filtering systems are developed with techniques of information retrieval, machine learning, and pattern recognition, they often use a linear function as the discriminant function. To classify information in the field of document classification more precisely, the systems have been reported which use a non-linear function as the discriminant function. The proposed method is to use the Kullback-Leibler divergence as the discriminat function which denotes to user's interest in the information filtering system. To identify the optimal discriminat function with documents which a user evaluates, we decide the optimal function using the genetic algorithm. We compare the present method with the other one using a linear discriminant function and confirm the effectiveness of the proposed method.

抄録全体を表示

This paper presents a method for atypical opinion extraction from ungrammatical answers to open-ended questions supplied through cellular phones. The proposed system excludes typical opinions and extracts only atypical opinions. To cope with incomplete syntax of texts due to the input by cellular phones, the system treats the opinions as the sets of keywords. The combinations of words are established beforehand in a typical word database. Based on the ratio of typical word combinations in sentences of an opinion, the system classifies the opinion typical or atypical. When typical word combinations are sought in an opinion, the system considers the word order and the distance of difference between the positions of words to exclude unnecessary combinations. Furthermore, when an opinion includes meanings the system divides the opinion into phrases at each typical word combination. By applying questionnaire data supplied by users of a mobile game content when they cancel their account, the extraction accuracy of the proposed system was confirmed.

抄録全体を表示

Wrong:Intelligent Systems Research Instutute, AIST
Right:Intelligent Systems Research Institute, AIST

抄録全体を表示

Wrong:The School of Information Environment, Tokyo Denki University
Right:School of Information Environment, Tokyo Denki University

抄録全体を表示

Wrong:The Department of Electrical Engineering, Tokyo Denki University
Right:Department of Electrical Engineering, Tokyo Denki University

抄録全体を表示

Wrong:Shikoku Reserch Institute INC.
Right:Shikoku Research Institute INC.

抄録全体を表示