The research focuses on approach for vision system. We propose a method to construct Visual Attention System Acquiring Environmental Information using moving camera according to Previous-Proposed Visual Attention System. We try to find a way that can rapid create blinking feature map ang moving feature map, and change the interesting map according to the camera mmoving using space information map.

View full abstract

This paper proposes modified VSM based on Meta keyword such as adjectives and adverbs, called M2VSM(Meta keyword-based Modified VSM). It is shown that M2VSM is superior to VSM in terms of ability for classification that reflects detailed contents and ability to adjust granularity of clusters. Furthermore, towards the construction of text mining system using M2VSM, structure of the system and an implementation of M2VSM with data mining tool MUSASHI are shown.

View full abstract

Because of the rapid increase in web pages recently, it has become very difficult to discover suitable web pages we really want using common web mining techniques based on only text information. In this paper, we present an algorithm for discovering suitable web pages using the notion of the web community computed from the analysis of the link structure of the web and the set of sample pages given as users input. The experiments show that our algorithm discovers more suitable web pages closely related to the input than the related method for many sample topics. We also discuss how to apply the PageRank measure to improve our algorithm so that it can appropriately take care of unwanted advertisement pages in the output.

View full abstract

Incident report system is widely employed to prevent medical accidents in hospitals. An incident report is the document described by nurse for an occurrence that might lead to medical accident during working time. It is expected to prevent medical accidents by analyzing incident reports. However, the analysis for incident reports has been statistically done by only using their metadata, e.g. occurred time, category of occurrence, skill of staff and so on. Though this statistical analysis gives us the tendency or classification, it has lost the most important information written in the text parts. This paper proposes a new knowledge extraction method from the text parts of incident reports using metadata and co-occurrence information in the text data. The proposed method can generate a keyword graph with multi-connected hierarchical architecture using text data and metadata, and one of the features of this method is that we can actively analyze what we want to know. This paper applies the proposed method to actual incident reports, and it shows the effectiveness of this method.

View full abstract

Recently, various digital information is shared by the Internet. The search engine that retrieves necessary information from huge database such as the Internet has become a most important application. However, it is difficult to select the best keyword to look for information that is necessary by the search engine because knowledge and the evaluation of the keyword are different from the user in the search engine. In this paper, we propose the information retrieval system based on intention inference for solving the above problem.

View full abstract

This paper deals with a cooperative game with stochastic payoffs and assumes that the preference of each player is based on a stochastic dominance. A new core based on second degree stochastic dominance, called SSD-core, is defined. Moreover, after a new balanced game, called SSD-balanced game, is defined, it is shown that if the game is SSD-balanced, then the SSD-core of the game is nonempty.

View full abstract

Dempster's rule of combination is the standard way of combining multiple pieces of evidence given by independent sources of information. However, it aroused great controversy about the validity and many alternative combinations were proposed. The paper reconsiders the model of combination which Dempster discussed in the original paper, and points out that it confuses the independence among information sources with the independence among bodies of evidence. Then, it proposes a new model of combination which introduces belief spaces, as well as a new combination called combination by compromise. The comparison with the other methods shows that the proposed is one of the most desirable combinations.

View full abstract

In this paper, we focus on two-level nonlinear programming problems, in which there is not coordination between the decision maker at the upper level (the leader) and the decision maker at the lower level (the follower), and propose a computational method using Particle Swarm Optimization (PSO) for obtaining Stackelberg solutions to two-level nonlinear programming problems. In our computational method, as we consider the above mentioned property of two-level programming problems, we introduce PSO to a PSO operating on each level by distinguishing a PSO operating on decisions of the leader (the upper level PSO) from a PSO operating on decisions of the follower (the lower level PSO). We carry out numerical experiments in order to demonstrate feasibility and effectiveness of the proposed method.

View full abstract

In this paper, we review several existing similarity measures of intuitionistic fuzzy sets (IFS) and then some numerical examples are applied to show that some existing similarity measures may not always be effective and reasonable in some cases. We propose new similarity measures of IFS which are able to overcome several drawbacks of some existing measures. Several examples will show the comparison of the proposed similarity measures with some existing measures. We also present two propositions concerning the relationship between similarity and dissimilarity (distance) measures of IFS, and further propositions in regard to the fact that a suitable comb- ination of similarity measures may be a similarity measure.

View full abstract

In recent years, social capital came around the old and rotten, such as concrete structures including the bridge built in the high growth era. It is becoming a subject to perform the maintenance management efficiently with degradation of these structures. However, it evaluation of the soundness in maintenance management is depends on experiential intuition of advanced specialists and engineering judgment until now. While the necessity for maintenance repair of a structure increases, the present condition is that specialists are insufficient. The objective of this research is perform diagnosis on a computer for degradation diagnosis of a concrete structure. First, it was inspected from this technology. This technique for a crack is extracted using evolutional image processing technology from the digital image which photograph the concrete structure.

View full abstract

In this paper, we propose a sound creator of hitting a golf ball by using Interactive Genetic Algorithm (IGA). IGA is a useful method for Kansei designing because it does not require any quantitative functions for optimizing systems. User only evaluates a sound which the system created and can get a golf sound which he hopes. But the user has a problem with being distressed by fatigue. So we need to quickly optimize a golf sound which composed of any parameter. We tried to reduce parameter and we are considered improving optimization performance by getting wave which we use the Discrete Wavelet Transform. As a result, changes in the optimization performance when parameter is reduced can be quantitatively compared.

View full abstract

We had proposed a rule-extraction method from computer log files using Automatically Defined Groups. In this method, agents search multiple rules cooperatively based on the difference between normal state log files and abnormal state log files. The more frequently used and the more reliable rule is, the more agents are allotted to the rules. Therefore, the number of agents for each rule can represent the importance of the rule. However, when the rule-extraction method is applied to the large-scale log files, which may contain a number of rules, a problematical situation on the number of agents could be observed. In the previous proposed method, the number of agents is fixed, so the number of agents may be lack for evaluating the each rule's importance minutely. In this paper, an improved method is proposed, where the number of agents is adaptively increased depending on the number of discovered rules. As a result, by increasing the number of agents, the importance of respective rules could be evaluated minutely. In addition, the proposed method could acquire more rules than those by the method with the fixed number of agents.

View full abstract

The present paper discusses a new approach to depicting an evolutionary optimization process. The approach depicts an evolutionary optimization process as change in distribution of degree of nodes in a network that emerges in the process. An evolutionary optimization process involves identification of interactions between variables, and ability in identifying the interactions influences the search efficiency. Therefore, network structures formed by using the indentified interactions information in an evolutionary process draw how an evolutionary algorithm solves a given optimization process. Futhermore, such a way of depicting an evolutionary optimization process could be utilized for classifying evolutionary algorithms. As an example, a mutation-based evolutionary algorithm that evolves developmental timings is used, and it is shown that power-law-like network topologies emerge in its evolutionary optimization process even for uniformly-scaled problems.

View full abstract

Personal computers are widely used today and we have expected paperless office, but paper still have great advantages that cannot be replaced by the digital devices yet. For example, looking over and viewing widely sense easily are two of the most important properties of paper and affording writing capability is also the important characteristics. To enhance or augment these characteristics of the paper, some pioneer work has been proposed but it needs large scale equipment and hard to bring to outside of the office. In this paper we discussed on the use of paper as the interface for the digital devices. Some typical situations have been investigated and the advantages using paper are also noted.

View full abstract

This paper proposes to apply association rules to user preference mining for TV program recommendation. Transition to digital terrestrial television broadcasting (DTTB) will bring us difficulty in finding TV programs worth watching from a number of TV channels. Therefore, a method for recommending TV programs will be important in near future. This paper proposes to recommend TV programs and related information based on user's TV watching log and social bookmarks. The method represents user's TV watching log with the form of bookmark, which is the same as social bookmarks. Association rules are extracted from the set of bookmarks including TV watching logs and social bookmarks, and extracted rules are used for TV program recommendation. The method is going to be used as one of the modules for realizing Web-based intelligent environment for human-robot casual communication.

View full abstract

This paper proposes Information Recommendation Module for providing humans with useful information under TV Watching Environment. The HuRoC(Human Robot Communication supporting system), which is proposed for supporting human-robot communication in a living room, estimates the user's interests from his/her TV watching logs and utterances under TV watching situation, and recommends relevant information gathered from the Web. The proposed module uses a profile written in bookmark format as input, and determines the timing and way of presenting information to be recommended based on the importance of the information and user's situation. The devices used for information recommendation are implemented as RT modules, so that the alternative presentation devices can be added easily into the system. The system aims to realize advanced human-robot relationship by supporting humans' information acquisition without explicit instruction to the robot or using PC.

View full abstract

In order to master any fundamental English sentences, we thought that it would be a good idea to memorize each sentence itself. Therefore, we developed an e-Learning SYSTEM with emphasis on memorization of sentences. When a learner makes a mistake, the correct answer and explanation of grammatically important points are shown, and then, the learner has to type the whole sentence correctly. Unless the task is cleared, the learner cannot proceed to the next question. Thus, we aimed to memorize the sentences by typing a correct answer repeatedly.

View full abstract

Moodle is one of the typical educational managed software used for e-Learning. It is open source software which has many functions. In this paper, it pays attention to the function to make tests in that, and we proposes the tool that supports the teacher who uses it. Concretely, the neural networks are implemented on Moodle, it consider the questions of the next tests based on the result, and the tendency of making the questions.

View full abstract

The recent years, the Internet is rapidly spread, information gathering on various fields become easily. A lot of people use web browser to gather information, they want to visit again the pages that they visited in the past. Because the title and URL on the page which it looks for are not necessarily memorized, a target page may be not searchable well only by the browsing history of a browser. In such case, it is helpful to remember the page's visual image that the screen on each page can be thumbnail displayed. In this paper, we proposed that the web browsing history is made visible by a thumbnail display on each web page. Using our system, user can do easily in the step where the retrieval of the web page visited once is shorter. We constructed it as an enhanced feature of Mozilla Firefox. When the web page is read by a browser, the flow for which the history of making to visible is made with this system preserves the title, URL, and the visit date in the XML file, and makes the thumbnail. The history with the thumbnail can be displayed by opening the XML file by a browser.

View full abstract

In this paper, we review Bayesian Filtering. Bayesian Filtering has the problem that probability of spam is almost the neighborhood of 0 or 1. We propose a new Mail-Filtering which is using Gray-Zone. For one spam-mail judged as a non spam-mail by existing Bayesian, the proposed Mail Filtering extracts the spam-mail from non spam-mail box and divides it Gray-Zone box.

View full abstract

Recently, many information systems provide the service to regular users only, and these information systems are protected by authentication system form spoofing. However, much damage by the spoofing is reported. In this paper, we explain the problem of the conventional authentication system is not only in authentication algorithm but also in an access control model using risk matrix. In this paper, we propose a multi-stage access control model for building a safer information system. The multi-stage access control model can adjust the level of the authority to access granted to a user using continual authorization. Therefore, user can control damage from the spoofing based on security policy of user self.

View full abstract

Abstract A lot of automatic composition systems exist now. However, many of they have the fault of becoming a composition of the computer leaving and not reflecting man's completion easily. Then, it proposes the composition support system that uses the conception support method in the creation study in this research. Moreover, "Loop sequence" is used for the input system to compose even if there is no musical knowledge, and MAX/MSP that specializes in the music software production is used for making the system.

View full abstract

In this paper, we put ecological driving knowledge of human in computer by developing intelligent controller. The intelligent controller consists of two parts, the judging situation part and the control part. The judging situation part judges situation from the velocity of the car, the gear number of the car, or target velocity. The judging part outputs a controller selection signal. The control part selects controller according to the signal, and selected controller outputs accelerator and brake. The judging situation part is built using fuzzy logic reasoning. The developed intelligent controller is simulated using MATLAB, and is compared with PID controller.

View full abstract

Manufacturing process is alive because the factors of the processes changes continuously and suddenly. In response to the change problem, sophisticated knowledge of skilled human experts has been applied to solve them flexibly. However, this kind of solution is subjective, inconsistent, and dependent on certain experts. Deskilling is a key to realize stable manufacturing with reasonable quality, cost, and delivery. Many types of automated systems have been deployed at the manufacturing process for realizing the deskilling. However, there still remain problems in flexibility against the changes. In this article, an automated sensory inspection system is employed to discuss the problems and the flexible solution against the changes. Based on the discussion, two types of solutions are proposed. The first one is fusion model and the second, human-machine collaborative decision making. Both solutions are considering the data distribution maturity in the manufacturing line.

View full abstract

We propose a novel classification method, which utilizes not only the feature vectors but also the neighborhood information on data observation points. The proposed method achieves an effective classification even if the boundary of each class is overlapped in the feature space. The practical effectiveness of the proposed method is verified by applying it to the classification problem of the intravascular ultrasound radiofrequency data.

View full abstract

A novel membership function makes class memberships of outliers less clear-cut and thus resolve the problem of classification based on normal populations or normal mixtures. The membership function is used for classification problems in a manner called as post supervised. We apply the fuzzy classifier to the combustion images obtained from a video camera focusing at the combustion and melting zone of a melting furnace. The algorithm takes about 30 seconds for a 10-fold cross validation on 2,000 images with 20 PCA basis vectors. The classifier is order of magnitude faster than the k nearest neighbor (k-NN) algorithm.

View full abstract

This article presents the new Estimation of Distribution Algorithms, EDABs, which apply the bootstrap to the process of estimation of distribution. EDABs choose the bootstrap samples from the population and construct the bootstrap distribution of the fitnesses. Then EDABs select the globally optimal individual from the distribution and estimate the distribution of the next population based on it. We evaluate the performance of EDABs using order-3 deceptive problem. We find population size significantly influences the performance, but bootstrap replication does not. The interesting result is that the ratio of the globally optimal individual in the bootstrap distribution and the population decreases as population size increases. We must find out why it happens and how the ratio of the globally optimal individual can be increased.

View full abstract

There are probabilistic restrictions on traditional Fuzzy c-Means methods which identify membership functions with the sum of membership values at each element as one. On the other hand possibilistic clustering methods identify the membership functions without such a constraint, but the shapes of membership functions are independent from the clusters estimated through the possibilistic methods. In this paper, we proposed FCM based on evidenctial theory. Using our method, we can obtain non-additive membership functions, and their shapes depend on the data distribution, which means that they differ from each other. To show the feasibility of the proposed method we have carried out some numerical experiments.

View full abstract

This paper discusses fuzzy discriminant analysis for diabetes data by using a simplified distance-type reasoning method, which is a simplified version of distance-type fuzzy reasoning method. Moreover, learning algorithm is proposed for the distance-type reasoning method and applied to a diabetes diagnostic system, showing the satisfactory learning results.

View full abstract

This paper shows a distributed fuzzy classification system for the purpose of speeding-up the process. Two processes are distributed over multiple computing elements in this paper. One is a fuzzy rule-generation process where the consequent class and the certainty factor are determined from training patterns. The other is a fuzzy inference process where an input pattern is classified using the generated fuzzy If-Then rules. In order to identify the bottle-neck process in fuzzy classification systems, first we investigate the computational effort in the process of fuzzy classification process. Then we introduce the idea of distributing the two processes over multiple computers that are connected to each other. In the computational experiments of this paper we show an example of the application of the distributed fuzzy classification system for high-dimensional pattern classification problems.

View full abstract

In many pattern classification literatures an error rate is often used to evaluate the performance of classification systems because a misclassification cost is the same for all patterns. However, it is necessary to consider the existence of important and unimportant patterns. That is, important patterns have a high misclassification cost while misclassification cost is low for unimportant patterns. We propose a fuzzy rule-generation method based on the misclassification cost. We show the effectiveness of the proposed method through computational experiments.

View full abstract

Flying objects are observed by the radar to be automatically classified into multi-classes. The problem is solved by the pairwise classifications and their aggregation. The summation of the memberships of the pairwise judgements, instead of their minimum, are employed, in order to reconfigure the classes after the classification. For the summation aggregation, the consistency between final- and pairwise-classifications requires the constraints, e.g. negative votes, in the membership functions of the pairwise judgements.

View full abstract

A fuzzy regression model has the objective to illustrate the possibility that a focal system has. On the case of an interval fuzzy regression model, the possibility is illustrated by the interval that the model expressed. Therefore, it is hard to find any important facts. If a center of model coincided with a center of system, it could have been acceptable subjectively. Therefore, we propose the fuzzy regression model that coincided with the center of possibility system with the center of our model in this paper.

View full abstract

The literary property has been used as an important management property for the enterprise. The most is invented in the research and development section in the enterprise. A long period and many funds are spent till result is reduced to its elements in a company. However, it may not achieve profit by all result. Seed of research and development needs to be determined based on needs or a company trend. In this research, I consider the analysis which used fuzzy theory for existing analysis of intellectual property.

View full abstract

We have many kinds of data of time series such as stock prices. We understand them via their natural language expressions. We, therefore, proposed a method to express a global trend of time series in a natural language. In this paper, we propose a method to extract outstanding local features to express largely varying time series. We generate data of time series for the global trend via fuzzy rule expressions of the global trend and extract local features as the position of large difference between the original data and the global trend. We apply the method to the data of Workshop on Multimodal Summarization for Trend Information (MuST).

View full abstract

Our main purpose is to extract useful rules from log files on computers, to detect various levels of errors, and to automatically inform these errors or configuration mistakes to system administrators in order to easily manage them without knowledge. For this purpose, we performed an extraction experiment using Automatically Defined Groups (ADG), which is based on Genetic Programming. Moreover, we focused on "System State Pattern" related to the difference between normal daily state and abnormal state that some errors occur in the system. In this experiment, then, we tried to extract rules without any manually managed and supervised information, by using simple translation technique: regular expressions. As a result, 50 agents in the best individual were divided into 16 groups from 322 log files. This means that 16 rules were acquired. We confirmed these rules could detect some errors such as DNS configuration error. We could also find the importance of the rules because each rule with more agents tended to have a higher adopted frequency by evolutionary computation. Therefore, we consider that our method using ADG is useful for the diagnosis of computer systems, and helps administrators manage their systems without expert knowledge.

View full abstract

In this paper, we get results for an inference of fuzzy quantified and truth qualified - especially qualified by false - natural language propositions. We get the the inference results Q's for the inferences QA are F is false -> Q'A are mF is false, and QA are F is false -> Q'(mA) are F is false (Q, Q': fuzzy quantifiers, A: fuzzy subject, F: fuzzy predicate, m: modifier).

View full abstract

In this paper, the fuzzy reasoning demand forecasting model is examined by using the business control data and meteorological factor during year. The analysis of sales of every day and the relation between the number of visitors and the meteorological factor is first tried here by applying a decentralized analysis, the principal component analysis is applied next, and the analysis of the relation of the meteorological factor of every day is tried. The fuzzy reasoning model which forecasts demand from these analysis results using the acquired combination is constructed.

View full abstract

Yubazaki proposed single input rule modules connected fuzzy reasoning method (SIRMs method) whose final output is obtained by summarizing the product of the importance degree and the inference result of single input fuzzy rule module. We hope that inference results of SIRMs method will have monotonicity if the antecedent parts and consequent parts of SIRMs rule modules have monotonicity. However, this paper points out that even if fuzzy rules in SIRMs rule modules have monotonicity, the inference results do not necessarily have monotonicity. Moreover, it clarifies the conditions for the monotonicity of inference results by SIRMs method.

View full abstract

In this paper, we propose a fuzzy reasoning method with plural regularization parameters. It is known that the annealing of the regularization term brings improvement of the generalization ability for the model by adding the regularization term to the error term in the cost function. In this study, we focus on the regularization parameter that is used to control the influence of the regularization term in relation to the error term. We consider the model that each parameter works on each rule by using not the regularization term as a unitary coefficient but as a matrix. Using the regularization parameters for each rule, a regularization parameter corresponds to one weight of the consequent part in a one-to-one relationship. Therefore we can know the influence of the weight to reasoning output. In the numerical simulations, the validity of the proposed method is discussed.

View full abstract

Linear estimation based sequential importance sampling methods for particle filters are proposed that can be used to detect the rapid change of object motion in a video sequence. First a linear least-squares estimation is used to build an importance function from observations, and then it is extended to a robust linear estimation. These sampling methods gives a framework for tracking objects whose motion cannot be well modeled by a prior model. Finally a switching algorithm between the proposed method and the prior model based sampling mehtod is proposed to achieve a filtering of both smooth and rapid evolution of the state. The ability of the proposed method is illustrated on a real video sequence involving a rapidly moving book held by hand. In addtion it is shown that the proposed method can achieve real-time processing.

View full abstract

In this paper, we propose a new method to remove shadows from a real image. In this method, the image is divided into shadow areas and sunshine areas by using the probabilistic relaxation method. The level of color and brightness in the shadow area are adjusted to those of the sunshine area, respectively. The adjustment rate is modified according to the distance from the boundary. The effectiveness of the proposed method is verified by applying it to images containing a shadow area.

View full abstract

This paper proposes a new image enlargement method accompanied by the estimation of high frequency components lost in a data aqcuisition process. The high frequency components are estimated by using the codebook composed of the low and high frequency components in a target image. The effectiveness of the proposed method is verified by applying it to some images.

View full abstract

Atrophy of the human brain is observed as the progress of dementia such as the Alzheimer disease (AD) and the frontotemporal dementia (FTD). According to the patient's symptoms, the atrophy occurs at the different rate on each gyrus. Therefore, estimating the atrophy for each gyrus helps us for distinguishing diagnosis. For the purpose, identification of gyri on the cerebral surface is required. In this paper, we propose a fully automated method for identifying the gyri in three-dimensional (3-D) human brain magnetic resonance (MR) images. To clarify the effectiveness of the proposed method, it was applied to identify the superior frontal gyrus, middle frontal gyrus, precentral gyrus and central gyrus on synthetic data and eight subjects (four normal subjects and four AD patients). Experimental results showed that the proposed method correctly segmented the appropriate gyri.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

Although early detection and correct diagnosis of early-stage melanoma is the most important, discrimination between melanomas and benign nevi is often difficult even by expert dermatologists, especially when these lesions are still small. We developed an Internet-based melanoma diagnostic system and opened it for the public. In present, our system was built based on 319 dermoscopy images and achieved 90.7% of sensitivity and 80.7% of specificity by cross-validation test. Our present system provides the diagnostic results by the expression "malignancy score" between 0 and 100. Considering a translating the system into a practical application, it is required that the system also provides the reason of the diagnosis with readable format for dermatologists. In this report, we investigated to the relationship between dermatologists' recognition of 15 clinical items for dermoscopy and objectively computable image parameters. We developed linear regression models for each clinical item based on 356 calculated parameters and they provided almost equivalent results to dermatologists in recognizing clinical items. Even though it is a preliminary experiment, we confirmed that the capability of computer-based method for recognizing dermoscopic patterns with a view of dermatologists.

View full abstract

A goal of this study is to construct 3D rough map form streetscape images automatically. For the sake of buildings reconstruction, buildings areas in streetscape are extracted. Typically, streetscape images contain the sky, road, street trees, buildings and so on. These regions other than buildings cause some errors in the following stereo matching process. Thus, a preprocessing to eliminate the areas other than buildings is required. The preprocessing is achieved by the newly proposed segmentation algorithm which integrates clustering with color information and wavelet analysis. The effectiveness of the proposed method was verified by experiments using actual streetscape images.

View full abstract

In this paper, we propose an ultrasonic nondestructive evaluation method for estimating cellular quantity in artificial culture bone by analyzing the ultrasonic wave. In order to measure the cellular quantity, we have to crush the culture bone and observed by the electro microscope. Therefore, they can't use the crashed culture bone in the research and the clinical. So we transmit ultrasound over the culture bone injected stem cell for getting a wave pattern. Our method analyzes these waves by fuzzy inference. As a result, we can estimate cellular quantity in culture bone with 60% accuracy.

View full abstract

Recently, Japanese society is aging so quickly that nearly one-third of the population will be people by 2050. We are assuming several problems will occur in the aging Japanese society. We are constructing a system to presume of lifestyle patterns of people living alone using pet robots, with which people feel at ease. We make the system about the established idea to take any lifestyle. We checked possibilities of those data to presume health condition and made a presume system. Also, the recent pet type robot has a high quality CCD camera. It is possible for the information to be acquired with the image processing by the proposal system. In this research, the pet type robot can give a healing to the person. We tested the collection of the data using the type robot. In this paper, we discuss the specific presumption system of lifestyle patterns using Hidden Semi-Markov Mode.

View full abstract