Recently it becomes reasonable to stream video contents according to the broadband popularization. In administrative organizations the importance of free access policy to self-governing body information, such as a broadcast of the assembly, enrich their WEB pages, etc., has been widely recognized. This paper presents a support system based on movie and sound processing for building streaming media contents about the assembly information. The content includes video contents of an assembly and superimposed dialogue of speech text. We show concepts of the support system through numerical examples.

抄録全体を表示

The Authors have proposed a method that detects inconsistencies between user interaction logs of a task and desired sequences for the task based on mouse click coordinate logs. The method models two successive clicks as a vector and thus a sequence of operation in a user/desired log as a sequence of vectors. To detect inconsistencies in user interactions and desired sequences, each vector from user logs is compared with each vector from desired logs. The Authors have developed a computer tool for logging and analyzing user interactions and desired sequences by the proposed method. The tool has applied to evaluation of "Sophia," a web-based system for browsing and searching assembly minutes. A total of 23 problems was found in three vers. of the system.

抄録全体を表示

There exists a strong need for better search engines that extract information that is minimal in quantity but maximal in usefulness. The conventional keyword-based search is advantageous in that it is basically domain-independent and computationally efficient. However, it is also often pointed out that "keywords" are not sufficient to specify the search condition and preference of the user. This paper presents a sentence-based search method where the search condition and preference are given by a sequence of sample sentences. This approach poses syntactic constraints on the relationship among keywords, and the contextual constraints on the sentences. We show an application of our search method to an assembly minutes system.

抄録全体を表示

Though various contents are provided through the internet recently, it is not easy to collect favorite contents among huge amounts of contents in terms of user's preference. In this paper, we focus on the collaborative filtering algorithm in the recommender system. We propose a modeling approach based on Modular fuzzy model for preference similarity model in collaborative filtering. In our approach, the model is constructed through optimization of MAE(Mean Absolute Error). The model decides the weights of preference similarity from the value of correlation coefficient, the number of items and variance values of evaluation ratings of each person. Through numerical experiments compared with conventional correlation coefficient based approach using Movie Lens data, we discuss validness of the model for collaborative filtering.

抄録全体を表示

In this paper, we introduce a new simple proposal for deletion of ellipse enclosing character in documentary images. In this proposal, ellipse enclosing character is labeled, parameters for the center point of an ellipse are found using circumscribing rectangle of the labeled ellipse, the parameters for half-lengths of the major axis and minor axis are determined by finding the farthest point and the nearest point on labeled ellipse from the center point, the ellipse orientation is determined by finding the angle between major axis and horizontal direction, then the ellipse is modeled using these parameters. Ellipse detection is conducted counting the number of pixels on the ellipse model. Finally, ellipse enclosing characters are distinguished from characters having an ellipse shape. Results of the experiments using appropriate images showed the effectiveness of our proposal. The sections that follow detail problems overcome and our projected work.

抄録全体を表示

In this research, we attempt to create a company information consulting system for job hunting students in Mie University, by developing a company information database which includes company profiles and job-opening information. Here, the paper based company information is converted to the digital documents by using image scanners and creating a database. The written data on the digitalized job-opening information sheets are recognized, and a searching system for the consulting system is developed using the recognized written data. We use OCR(Optical Character Reader) for this purpose. The lines affect the character recognition of OCR. The purpose of this research is to delete solid lines and dot/dash lines on the digitalized job-opening information sheets and improve the character recognition of the OCR. In this paper, we developed algorithms which are more effective than conventional methods for this purpose without affecting characters. Results of the experiments using appropriate images showed the effectiveness of proposals.

抄録全体を表示

This paper presents a novel algorithm, called Radial Sector Coding (RSC), for Translation, Rotation and Scale invariant character recognition. Translation invariance is obtained using Center of Mass (CoM). Scaling invariance is achieved by normalizing the features of characters. To obtain most challenging rotation invariance, RSC searches a rotation invariant Line of Reference (LoR) by exploiting the symmetry property for symmetric characters and Axis of Reference (AoR) for non-symmetric characters. RSC uses the LoR to generate invariant topological features for different characters. The topological features are then used as inputs for a multilayer feed-forward artificial neural network (ANN). We test the proposed approach on two widely used English fonts Arial and Tahoma and got 98.6\% recognition performance on average.

抄録全体を表示

This paper describes a machine learning approach for visual pattern recognition framework that is capable of processing images rapidly while achieving high recognition rates. The key contribution of this study is a tractable model suitable for object recognition, that takes advantage of two complementary components: (a) a bag of covariance matrices object descriptor based on feature selection, allowing to combine the advantages of histogram and appearance model with the ability of scalability, (b) a model with incremental learning based on on-line variant of Random Forests (RF) learning algorithm. We evaluate the potential of the proposed procedure with empirical studies in the domain of the GRAZ02 dataset. The system yields in object recognition performance comparable to the benchmark standard RF, AdaBoost, and SVM classifiers. Experimental results demonstrate significant performance improvement.

抄録全体を表示

This paper proposes a computational scheme for fuzzy similarity analysis and classification of images. First, a special growing unsupervised learning algorithm is introduced for Information Granulation of the original "raw data" (the RGB pixels) of the image with much smaller number of information granules (neurons). Then two features are extracted from each image, as follows: the Center-of-Gravity of the image and the Image Volume (as number of neurons). These features are used as inputs of a special fuzzy inference procedure that computes numerically the similarity degree between a given pair of images. Finally, a sorting procedure with a predefined threshold is used to obtain the final classification results for all available images. The proposed classification scheme is illustrated on the example of 18 images of flowers. It is also shown in the paper that by appropriate tuning of the parameters of the Fuzzy Inference procedure, high plausibility of the classification can be achieved.

抄録全体を表示

The objective of this talk is to introduce the concept of BioInformatics or Computational Biology, and various kinds of intelligent techniques used in the area of BioInformatics. First of all, we review the definition of BioInformatics and it's necessity in the era of mass production of bio data. After summarizing the basic principle in biological system, i.e. the central dogma of molecular biology, the necessity of intelligent techniques in the research on bioinformatics is discussed. We have defined the problems which are investigated by the biologists and computer scientists on the information flow in the central dogma of molecular biology. We will also show how the intelligent techniques are used in the above bioinformatics problems.

抄録全体を表示

Sequence similarity algorithms using dynamic programming (DP) are useful for sequence comparing of DNA, RNA, protein, and even program source code. Those algorithms adopted a matrix manipulation from the sequence alignment which produces a similarity degree proportional to the length of the sequences. So, the existing algorithms are unsuitable for sequences with various lengths. In this paper, we propose S2 algorithm which is a sequence similarity algorithm irrelevant to the sequence length. This algorithm suggests a normalized similarity for two sequences. We also perform the evaluation of the S2 algorithm with the source code similarity comparing.

抄録全体を表示

As Adleman proposed before, the deoxyribonucleic acid (DNA) strands employed towards calculating solution to The NP-complete Hamiltonian Path Problem (HPP). No longer is the non solution for any NP problem. Lipton also demonstrated that Adleman's techniques could be used to solve the satisfiability (SAT) problem. In this paper, we demonstrates how the DNA operations in the Adleman-Lipton model are used to construct bio-logic molecular parallel subtractor and divider.

抄録全体を表示

Recently, our life style has two kinds of communities, one is a real face-to-face communication and the other is virtual communication. In school SNS, counseling system for university students has been developed. When the system receives a request for counseling, the scheduling function assigns the counseling appointment to an appropriate counselor by the adjustment of schedules. Simultaneously, the system analyzes the contents of counseling requests from the students to classify them into three annoying patterns {living, study, and job} by morphological properties. This method analyzes the distribution of morphemes in the input text by using Self Organizing Map. The system can show the analyzing result of annoying pattern of the counseling request and its emotional degree for each category by Emotion Generation Calculation.

抄録全体を表示

Vocational decision making is one of the most important developmental tasks in adolescence and even in middle-age. We have been constructing an interactive career support system for undergraduates and middle-aged people. We report how to scale vocational indecision and rigidity and to construct an interactive support system for career decision making system using Boolean analysis and rough sets, and the effectiveness of the system is discussed.

抄録全体を表示

Self-organizing map (SOM) is a type of artificial neural network. SOM is trained using unsupervised learning to produce low dimensional representation of the training samples while preserving the topological information of the input space. There are three problems when applying SOM for clustering: map initialization, computational cost, and limited capabilities for the representation. Hierarchical SOM and tree structured SOM have been previously proposed to solve the problems of map initialization or computational cost. In this paper we propose an adaptive tree structured clustering method using SOM in order to improve the classification capability. In our proposed method, separate SOMs are arranged to correspond to nodes of a binary tree structure. The binary tree structure is generated by recursive child node creation that is determined by the classification results of the corresponding parent node SOM. The proposed method utilizes the competitive learning feature of SOM, and the relationships in the data set are shown as the generated tree structure.

抄録全体を表示

In some optimization problems where user's preference should be considered, it is difficult to build the preference into the evaluation function quantitatively. We adopt the following approach for solving the issue: An intelligent system presents two or more candidate solutions which have good evaluation values in the viewpoints except for user's preference, and the user chooses a favorite one out of the set of the solutions. In this research, to construct such a system for the combinatorial optimization problems, we aim to propose Ant Colony Optimization (ACO) method which can search two or more high-quality diverse solutions. ACO is one of the methods to solve the Traveling Salesman Problems (TSPs). This method basically outputs only one discovered best solution by a single run. In this paper, we introduce tabu search mechanism into ACO for discovering diverse solutions in a single colony. Ants register the discovered solutions in a tabu list as ants search for the optimal solution. When the best found solution is not updated for a certain period, the search is regarded as converged. Then, ants search different solutions from ones registered in the tabu list. We applied the proposed methods to some TSPs, and confirmed that the acquired diverse solutions are useful for users' decision makings.

抄録全体を表示

Recently, computer systems are popularly composed of two or more computer servers connected via LAN, where each computer server individually serves a specific function. The computer system provides the required service by the combination of individual server operations. The computer system records many events into a system log file. The log file is recorded the messages generated by specified software operation records. However, the detailed notations are not strictly defined. Therefore, actual descriptions as well as the definition of logs include many discrepancies in the log files. Such problem like syntax analysis is more serious in evolutionary search. If the elements of operands are automatically extracted from the log file with regular expression, GP search results in the explosive size of tree including some partial same structure recursively. In order to avoid such environment, we propose a similarity detection technique in the target words.

抄録全体を表示

This paper introduces two applications of an ontology-based fuzzy system to marine system : a ship's steering control system and a collision avoidance system for ship. In the steering control system, an ontology-based fuzzy support agent consists of linguistic instruction ontology and a fuzzy support agent on which a ship's steering controller is based. In the second system, ontology is applied to identify a new case in a fuzzy CBR for ship collision avoidance support system.

抄録全体を表示

In this article is given to know the technique of adaptive control FRMLC which is applied to diffuse controllers in order to adjust its base of rules so that the system behaves according to a reference model. This technique is applied to the control of speed of an automobile, which is subject to interferences that affect its dynamics.

抄録全体を表示

This paper deals with the optimization of Petri net modeled concurrent service systems, subject to the customer satisfaction constraint. Service systems are inherently concurrent systems with multiple asynchronous activities. Petri nets are ideal tools to model concurrent systems. The objective function consists of service costs and waiting costs. Service cost is the cost of hiring service-providing professionals, while waiting cost is the estimate of the loss to business as some customers might not be willing to wait for the service and may decide to go to the competing organizations. Particle Swarm Optimization (PSO) is selected as the meta-heuristic for optimization, owing to its simplicity in programming and its rapid convergence. The optimization procedure is illustrated with the example of a practical service system. Our experiments show optimum results for the operation of the concurrent service systems.

抄録全体を表示

In many practical situations, the system is modeled as a continuous-time fuzzy system, while the control input is the zero-order hold, which can be represented as a piecewise-continuous delay. Such a system is called a sampled-data system. In this paper, we consider sampled-data stabilization for fuzzy systems with multiple time-delays is discussed. We take an input delay system approach to the sampled-data control problem. The closed-loop system with the zero-order state feedback controller becomes a system with multiple time-varying delays. First, stability conditions for the closed-loop system are given in terms of linear matrix inequalities(LMIs). Such conditions are derived by using Leibniz-Newton formula and free weighting matrix method for fuzzy time-delay systems under the assumption of time-varying sampling. Then, a design method of sampled-data stabilizing controller for fuzzy time-delay systems is proposed. Numerical examples are given to illustrate our sampled-data control design.

抄録全体を表示

To generate virtual images of a soccer game from free view point, we have proposed a method of pose recognition from silhouette of the soccer player image by applying the parametric eigenspace method. However, there are some cases that positions of the both hands are replaced each other and it makes a mistake in the direction of body because of using only silhouette information. This paper propose a new method of excluding the candidate pose where right and left limbs switched positions. It is assumed that the change of the pose (especially the positions of the four limbs) comes to the continuous transition between sequential frames. As a result, the positions of four limbs in the space and image can be estimated and the search range can be restricted in the eigenspace. Further, to process with the continuous frames, we describe a method of setting an initial state and handling error, in which a correct pose image exists out of the restricted range.

抄録全体を表示

This paper proposes the virtual scissors system using virtual hands. Users can cut virtual paper with their own hands. The purpose of this system is to make it easy to implement different tools by only changing parameters in software. Other purpose is to make system with small-scale input and output devices for general applications, for example, only using thin haptic information and force feedback. In this VR (Virtual Reality) scissors system, various feedbacks were introduced in order to cover any impressions such as the interference feeling for interface problem while using the virtual scissors and cutting the virtual paper. Using this system, users can manipulate and cut virtual scissors through a data-glove as similar to real ones.

抄録全体を表示

This paper presents a fuzzy filter for document image improvement. For document image (specially for historical and handwritten document image) binarization, Noise reduction, smoothing of background and text areas, contrast enhancement, and edge gradation for stain areas are necessary preprocessing operations. Usually, a tradeoff between noise reduction and contrast enhancement is necessary: for noise reduction, it needs to decrease high frequency elements, but such process will bring down contrast of images. Address this problem, we present a fuzzy method to realize noise reduction and contrast enhancement simultaneously. We define membership funcitons to divide different cases automatically by local analysis. We also discuss how to tune the membership functions, and use experimental results to compare our filter with the median filter and the low-pass Wiener filter.

抄録全体を表示

In this paper, a model to realize an interactive manipulation of virtual water using a virtual container is described first. Then, this model is extended to represent the wave surface in the container. The mesh model are introduced for the wave representation so that the waves are propageted. Proposed system with this integrated model makes it possible to swing the water surface and to spill and/or stir the water.

抄録全体を表示

This paper proposes a self-calibration and neural network based photometric stereo for virtual image generation, and purposes an improvement in precision of generating image. This method can generate virtual images without an calibaration object which has the same reflectance function as the test object under the same light source direction. However, the accuracy for generated virtual image for improvement requires a large number of images in the previous neural netwok based rendering. The proposed method uses the angle between a unit surface normal vector and an equally divided vector between each light source direction and viewing direction, and the approach can increase efficiency for neural network learning. Better virtual image can be generated from a few observed images. Generated images are evaluated using Phong model and Torrance-Sparrow model in comparison with the previous approach.

抄録全体を表示

This paper proposes a polyarticular tendon-driven system to simulate compliance properties of human finger. The compliance of a human finnger contributes to ensure stability of a grasping object and dexterous of human finger and then an exoskeleton assistive hand should have the same property of human finger compliance for safe and agreeable support. The compliance of the finger in orthogonal direction to a grasping force decreases as the grasping force increases, when human pinches an ob ject with his index finger and thumb. Besides, the compliance is not sensitive to direction of a fingertip displacement when an external force affects to the fingertip. Through numerical simulations and experiments, it is confirmed that our proposed polyarticular tendon drive system has the similar compliance properties to the compliance of human finger.

抄録全体を表示

In this study, we aim to develop an intelligent power assist system which can detect object and guide the operator to the object for helping the operator in long hour of monotonous and repetitive work with keeping the maneuverability. We propose a visual guidance power assist system. The visual guidance system generates a guidance force field based on an target object position. The guidance force field is obtained by an expected force based on Hill's low, and it helps to the operator to approach the object in spite of the operation velocity. The guidance force obtained by this potential field makes it the operator to perform repetitive works. Experimental results will show the proposed method is effective for improving the maneuverability of the power assist system.

抄録全体を表示

This paper proposes a method of 3D (3 dimensional) perception-assist for an upper-limb power-assist intelligent exoskeleton with visual information in order to help daily activities of physically weak persons. In daily activities, a user moves his/her arm in 3D space, sometimes to touch or grab objects. In the propose method, the exoskeleton assists the perception of user's motion in 3D space when the user is going to interact with the environment. By using the attached sensors, the exoskeleton can identify the required object for the user when the user is going to interact with such an object in the environment. If the user is moving his/her hand in a correct trajectory with respect to the position of the object, the exoskeleton performs only the ordinal power-assist. In the case when the trajectory of the user's hand is different from that of the estimated 3D trajectory required to properly interact with the object, motion modification is carried out in 3D by the exoskeleton by introducing an additional 3D force vectors at the end-effecter of the exoskeleton to correct the trajectory of the user's hand.

抄録全体を表示

We show the unified dynamics computation employing Jacobian that relates the center of gravity (COG Jacobian) to joints of link mechanisms. COG Jacobian is used for the behavior planning and the control of humanoids. Also, it usually expresses the relationship between the joints and COG of a robot's whole body. However, this scheme that is suggested by us calculated regarding each link and not the robot's whole-body. Moreover, we can obtain the torques of the active joints in mechanisms by relating COG Jacobian to the applied forces to COG and by using principle of virtual work. The COG Jacobian is obtained via the process calculating the accelerations at COG of links and the equation of motion expressed matrix forms by using COG Jacobian. Therefore, we can compute the equation of motion of a mechanism effectively, and the forward dynamics which can be calculated by the obtained equation. We can calculate the inverse dynamics regardless of open- and closed-link mechanisms.

抄録全体を表示

In this paper, we present the color recognition algorithm based on color constancy algorithm of using Self-Organizing Map (SOM) for our robot vision system. SOM is an unsupervised learning algorithm that performs topology-preserving transformation from higher-dimensional vector data spaces to low map spaces. The SOM has become a powerful tool in many areas such as data mining, data analysis, data classification, and data visualization. Our robot vision system is based on YUV and HSV color map spaces. Both color maps have different vector spaces, then, some objects are recognized by using threshold to use logical addition to YUV and HSV thresholds. To realize a robust robot vision system against light changing environment, in our new robot vision algorithm, recognition of the light color environment and the threshold parameters are, both, estimated by using SOM.

抄録全体を表示

Due to the steady progress in technology and the fact that the number of computing resources is increasing, today parallel computing on computer grids or on multi-core systems can significantly reduce the computation time for highly complex modeling, simulation, and optimization problems. Particularly optimizing multi-objective problems is more challenging as there are several objectives which have to be optimized simultaneously. Here, we investigate parallelization of multi-objective algorithms on a set of heterogeneous parallel computers such as a grid. The solution of multi-objective optimization problems is usually a set of solutions represented as an optimal front i.e., none of these solutions can be improved in one objective without getting worse with respect to some other objective. In order to perform the multi-objective optimization on a grid we must parallelize the algorithms. Here, we study two different parallelization paradigms: The Master-Slave and the Island models. In the proposed algorithms for the Island model, every computing resource is responsible for solving one part of the optimal front so that in a collective way all of them obtain the entire optimal solutions. Besides the collective property, we allow the cooperative aspect in the heterogeneous system such that the computing resources indirectly exchange the best found solutions. For the both paradigms, the grid must be represented as a unified resource to the user, who gives the objectives to the system and waits for the optimal solutions. On the other hand, in a heterogeneous system we must make use of all of the resources from very slow to very fast ones. Although the solutions in a multi-objective problem are dependent to each other by the domination relation, the fast processors have to continue working as soon as they are done with their tasks without wasting time in waiting for the slow processors.

抄録全体を表示

In this paper, we examine the effect of using a non-geometric binary crossover operator in genetic algorithms through computational experiments on some test problems including a GECCO 2007 worst one-max solver competition task. Whereas standard binary crossover generates an offspring in the segment between its two parents under the Hamming distance in the genotype space (i.e., the sum of the distances from the generated offspring to its two parents is equal to the distance between the two parents), non-geometric crossover generates an offspring outside the segment between its two parents. We demonstrate that non-geometric crossover can slow down the evolution toward the optimal solution through computational experiments on the worst one-max solver competition task. Its usefulness as a diversity maintenance mechanism is also shown for a knapsack problem and a function optimization problem of Schwefel function.

抄録全体を表示

An innovative neural fuzzy system is considered for cognitive design using a neural tree structure with nodes of neuronal type, where Gaussian function plays the role of membership function. The total tree structure effectively works as a fuzzy logic system. The structure of the tree is determined by domain knowledge and each node represents an entity of the domain of concern. The states of these entities are dependent on the stimuli at the input and the relationships between the stimuli and the states form the core for cognitive design. Namely, for each stimulus the status of the system is known and interpreted not only at the output but also at the granulated level concerning the system sub-domains. The research is described in detail and demonstrative applications are reported.

抄録全体を表示

Genetic Algorithm (GA) is the probabilistic search technique that derived from evolution of living thing. Many of GAs are inspired by Darwin's theory of evolution, but there are various opinions in the theory of evolution, and there is the virus theory of evolution in one of them. Virus Evolutionary Genetic Algorithm (VE-GA) is the technique that is based on virus theory of evolution. VE-GA is performed a global search and a local search by co-evolution of two kinds of population. On the other hand, in late years Estimation of Distribution Algorithm (EDA) is studied actively. It is a technique to generate children individual using a joint probability distribution of each locus. EDA has the advantage in the place where the effective gene destruction doesn't happen by genetic operations such as crossover and mutation. However, in general EDA, there is no mechanism that searches for the local solution. Therefore, we propose a co-evolution EDA which performs local search using two kinds of population. In addition, we try the application of the proposal system to Multi-Objective Problems.

抄録全体を表示

Grasp recognition is a part of Programming-by- Demonstration (PbD) for a five-fingered robotic hand. The robot receives instructions from a human operator to perform different grasps to be used for a robot task. For this purpose the finger joint angle trajectories are recorded by a dataglove and modeled by fuzzy clustering and Takagi-Sugeno modeling. The resulting grasp models use time as input and the joint angles as outputs. Given a test grasp by the human operator the robot control system recognizes the grasp and generates the type of grasp shown before. Three methods of grasp recognition are presented and compared with each other. In the first method a test grasp is compared with model grasps using the difference between the model outputs. In the second method qualitative fuzzy models are used for recognition and classification. The third method uses Hidden-Markov-Models (HMM) for recognition.

抄録全体を表示

For late few years, research area of intelligent robot was one of emerging technologies. Many intelligent robots have been commercialized. For examples, vacuum cleaning robots, house keeping robots, and mower robots are already on the market. Also, many vehicular technologies have been developed. For instance, driver assistance system, navigation system, etc. are demonstrated in the exhibition. Nowadays, some researchers are making a new trial to fuse robotic intelligence and vehicle technology. A new terminology should be defined to present the technology of the new trial. So, in this paper, the definition, technologies, and applications are presented.

抄録全体を表示

This paper proposes an intelligent player in a fuzzy coach player system for learning and controlling an industrial robot manipulator. The function of the intelligent player is implemented by a visual evaluation process based on visual coordination. It is performed by a supervised self organizing map (SSOM) and fuzzy voice commands are used to learn the visual evaluation. The learned system is capable of positioning the robot manipulator to a point in 3D working space as instructed by the voice commands. The proposed idea is demonstrated with a PA-10 industrial manipulator.

抄録全体を表示

This paper proposes a control scheme for a robot system via a knowledge database using visual image. Generally speaking, 3D calibration for the position information acquisition is necessary, when we use visual images by cameras. However, the 3D calibration is a complicated procedure. In addition, especially for hand-eye systems, the calibration is required for every movement of the robot. Thus the situation is more difficult. In this work, we built calibration-free robot system via a knowledge database which gives 3D quantity of deviation for the robot arm from 2D image data by two cameras. Moreover in order to establish and update the knowledge base, we used the simultaneous perturbation method. We made a system in which the knowledge database is effectively established for every movement with SP method. As a result, the hand-eye robot arm system can move quickly without 3D calibration.

抄録全体を表示

A wheeled mobile mechanism with a passive and/or active linkage mechanism for rough terrain environment is developed and evaluated. In our previous research, we developed wheeled mobile robot with six wheels and a linkage mechanism, and its maneuverability was experimentally verified. And also, we developed a switching controller system for wheeled mobile robots in rough terrain. This system consists of two sub-systems: an environment recognition system using Self-Organizing Map (SOM) and an adjustable control system using neural network. In this paper, we propose the PID-like controller for wheeled mobile robots. The PID-like controller consists of same inputs of adjusted controller using neural network. The PID-like controller parameter tuned based on the adjusted neural network gradient. Our proposed controller shows almost same performance of adjusted neural network controller in the experiments of going up the stairs. And also, that controller shows better performance than well tuned PID controller.

抄録全体を表示

Gesture recognition is a strong tool for communication between robots and humans. Many researchers have worked on a vision-based system that recognizes the gestures/postures of humans for robots serving human society. Most research regarding markerless gesture recognition has been related to developing a user-friendly interface. However, these approaches have constraints, in the environment must be a simple background and the humans interacting with the robot must wear tight clothes. In daily-life, however, it is preferable that the robust markerless gesture recognition utilizing a parallel stereo camera. To evaluate the validity of our method, we show the results for that gesture recognition with any clothes and a cluttered background.

抄録全体を表示

This paper propose the classification by the spherical SOM using learning vector quantization (LVQ) in order to specify this boundary automatically, and try to improve the classification accuracy. As a result of the numerical experiment, a clustering method using the LVQ yielded 98.7 % accuracy. Though it was only data of Iris and Wine, the area where a boundary line exists by using LVQ could be confirmed. But it wasn't possible to specify a boundary line actually. We could improve classificaton accuracy by considering a dendrogram and a spherical map interactively.

抄録全体を表示

In this paper, we propose a method for determining class borders. The method approxi- mates the borders on a SOM from nearest neighbor data set of different classes. The method decides the border on the SOM using the characteristics of the SOM which can map a high dimensional data set to a low dimensional map which is 2 or 3 dimensions. Using iris data set and the wine data set, it is shown that the proposed method can improve the accuracy of cluster analysis. The method allows the class borders to be visualized accurately and com- prehensible. The advantages were successfully dem- onstrated using two examples which contain the class borders that are really difficult to depict.

抄録全体を表示

We propose the use of a high speed spherical self-organizing map (HSS-SOM) to visualize climate variability as a complementary and alternative method to empirical orthogonal function (EOF) analysis. In the fields of meteorology and climatology, the EOF analysis, which is the same as the principal component analysis, is often used to extract leading patterns of climate variability. However, since EOF analysis obtains a linear mapping only, sometimes we can not obtain any meaningful result. On the other hand, because of computational limitation, it is difficult to apply conventional self-organizing map to huge climate datasets. Recently, one of the authors has developed HSS-SOM with dynamically growing neurons to reduce computational time. First, we validate our method using observational climate data, and show the effectiveness of the HSS-SOM as a complementary and alternative way to the EOF. Next, we extract dominant atmospheric circulation patterns from huge climate data in the general circulation model at the first time, in which both present climatology and future climate are reproduced. These patterns correspond to those obtained in the previous studies, which suggest that the HSS-SOM is usable for climate research.

抄録全体を表示

Abstract- In this study, we are concerned with the concept, properties and algorithms of differentiation of the Choquet integral. The differentiation of the Choquet integral of a nonnegative measurable function is studied in the setting of sensitivity analysis. It is shown that the differentiation of Choquet integral reflects a way in which the aggregation is sensitive to the function being aggregated. The aggregation function using fuzzy integral could be viewed central to data-mining, the business process mining, web-marketing, e-commerce and the other pattern-matching problems in economics.

抄録全体を表示

A 3D (three-dimensional) pseudo-reconstruction method from a single image is presented as a novel approach reconstructing a 3D model with no prior internal knowledge of outdoors image. In the proposed method, an image is represented as a collection of sky layer, ground layer, and object layer. A visual radical coordinate system with vanishing viewpoint is established to accommodate the extracted 3D data from images. Learning method is done via the layers database. The experiment results show that the visually acceptable 3D model can be extracted less than one minute. That means a higher resolution in much shorter time, compared to conventional methods. This method can be applied to computer games, industrial measurement, archeology, architecture and visual realities.

抄録全体を表示

Classification method for traffic sign using block-partitioned matching method is proposed. So it describes a new solution according to the idea of simplifying the problem, using color information and shape information. The proposed method uses HSV image in the preprocessing step. HSV is convenient to match the color and to determine one color is similar to another color. The first step is to divide by group. The second step is used Histogram analysis method for color information and projection method for shape information as feature of traffic sign. Binary image uses for the projection. Finally block partition matching uses in the matching step for classification. This method does not need many sample images of traffic sings in the training process, only the standard traffic sign uses in this method as a reference image. So this method is able to reduce the time for the traffic sign recognition and to overcome the occlusion and illumination problem. The experimental results on circular traffic signs show the effectiveness of recognition system.

抄録全体を表示

A text-dependent speaker recognition method using trapezoid fuzzy similarity function to decide how similarity between voice features of a test user with those of the registered speaker with nearest distance is proposed. The trapezoid fuzzy similarity function is constructed based on three-time data recorded during enrolment to present personal identification voice (PIV) and a statistical data of a person recorded many times in a long period to absorb the intra-variation. A set of acoustic voice features is also introduced to present some general speaker and text dependent characteristics that are effective for modeling PIV, thus can capture the inter- variation from one speaker to another. The experimental results on 24 speakers recorded in 4 different sessions show that, without false accepting, the proposed system can decrease 23.15% of false rejecting cases, compared to traditional nearest neighbor approach. The focus of this work is on applications which require fast processing and few burdens for users.

抄録全体を表示

Healing music and Musicotherapy attract attentions recently. They are beginning to be used in the medical treatment. On the other hand there are little practical utilization to relax with video images compared with music. Visual information is said to occupy about the whole of 80% humans receive from five senses. Therefore it has some potential that visual information affects human's mental and physical state as well as aural information does. This paper attempts to investigate the relation between characteristics of video images and mental relaxation. The one of characteristics we focus is motion speed of video images. It is considered that it might correspond to tempo of music. This paper investigates the relaxation effect by using the index calculated by heart rate variability analysis. The experimental results and analysis confirmed that viewing video images might have higher effect of reducing stress than eyes closed state and the relaxation effect might depend on motion speed.

抄録全体を表示

This study is aimed to relax the people, who are under some stress, by using visual stimuli. Visual information is said to occupy about the whole of 80% humans receive from five senses. But little is known about the relation between visual information and mental relaxation (mental stress reduction). In this paper we will attempt to study and develop a relaxation method using visual stimuli, and this study is the first step for multi use of several media (video, music and so on). We prepared five color stimuli (red, yellow, blue, green and purple) to display on a computer screen. The experiment intend to compare the relaxation effects of each color stimulus by using TPA. The analysis indicated "yellow" was the most effective and "purple" was the least for mental relaxation and stress reduction.

抄録全体を表示