We report a verification results about comprehensive evaluation of a fuzzy integral output on fuzzy analytic hierarchy process (AHP) for alternatives evaluations. Fuzzy AHP is a method which uses fuzzy integration instead of weighted summation in normal AHP method for calculation of overall evaluation outputs. In this study, we use fuzzy integral calculation WEB (produced by Takahagi). We assume users would like to buy personal computer that suits themselves and investigate the outputs comparison when changing ξ; parameter of deciding whether alternative evaluation or complementary evaluation, with fuzzy AHP.

We report an investigation of a collaborative filtering using Womnavi application for Official characters of prefectural PR. We investigate degree of preference for Official characters by user and recommend his/her favorite characters to users through collaborative filtering. We collect a questionnaire data of user about cute or not cute of Official characters and find the most preferred Official character. Also, we recommend his/her preferred Official character using collaborative filtering by Womnavi application.

In this experiment we analyzed the questionnaire data on features of adult children using hierarchical clustering tool software. There are 5 types of adult children, and we confirm whether data can be successfully divided into five types by clustering from questionnaire data. From the experimental results, we will also examine the relationships of each type and expand the new findings.

The number of elderly people with dementia increases year by year, and about 20% of people in Japan over 65 years old are estimated to be dementia patients in 2030. Because dementia is difficult to be treated after the symptoms advance, early detection of dementia is important. The peak latency of P300 is one of the EEG (Electroencephalogram) features. It has been reported that the peak latency of P300 is dependent on MMSE (Mini-Mental State Examination) score. If MMSE score can be estimated by measuring the latency of P300, it is expected to be used for the early detection of dementia. However, no regression equation between MMSE score and the latency of P300 had been identified yet. The authors measured the peak latency of P300 of dementia patients who were attending the Geriatrics of Nagoya University hospital and identified a multiple regression equation with MMSE score as the objective variable, and the latency of P300, age, educational background as the explanatory variables. The 95% confidence interval of estimated MMSE score was ±3.15. This paper incorporates alpha wave as a new explanatory variable for the regression equation. The result shows that the alpha wave is statistically significant for the estimation of the MMSE score, and the 95% confidence interval of estimated MMSE score is reduced to ±3.09.

P300 is one of ERPs (event-related potentials), which arises when an infrequent stimulus appears. Oddball tasks are often used for measuring P300. I try to make a game-based oddball task for users to enjoy the tasks. There are a lot of elements in a game, BGM (background music) is one of them. This paper discusses the effects of BGM in oddball tasks.

In this presentation we discuss the characteristic of sound localization for the Emergency guide system, which provide the escape sound so as to escape quickly and accurately to the exit. We performed the evacuation experiments with several conditions, and recorded the results of accuracy rate of subjects for sound localization.

In recent cooperative behavior study with goal type sports, some devices such as action cams and acceleration sensors are used for measuring personal behaviors. In these measurements, we mainly analyze the output of player’s behavior. But in cooperative behavior, self and other persons confirm mutual actions and share their intention estimation. And they decide their purpose and judge the next own action. In this research we analyze the intention estimation by player’s head movement and model it to simulate the estimation process. We experimented soccer pass behavior and measure player’s head movement by a flying drone. We describe the measurement accuracy of the player's head movement by the drone and the analysis result of the timing when a ball holder shares the intention of other players and passing.

Data obtained from multiple observations for the same object generally take different values. Normally, the differences within a dataset are regarded to be caused by some errors. The errors are usually assumed to be the normal distribution. Then, the average value AV and standard deviation SD are calculated performed from the dataset. On the other hand, Oda (1992) proposed the new indicator C .. of the sum of contradictions of multiple observed values. The idea for the new indicator is that the irrelevancy-contradiction (C) index may be used for estimating psychological feelings to the distribution of multiple observations better than traditional indexes like SD values. C.. is the sum of C values. C .. is an extended form of the so-called "Range" index, and takes different function forms for the observation number. In the function forms of C.., the linearly ascending weight values are multiplied to the internal Range values. Also, if n is an odd number, the median data is not used. Furthermore, we obtain the ambiguity index AMB=2*(C../n^2), so that it can take almost near values of the SD values obtained from the same dataset. In this research, in order to reveal the characteristics of the AMB index for the small n values, we conducted the Monte Carlo Simulation for the purpose of clarifying the zone in the multiple dimensional space, where data with large differences between the SD and the AMB value.

In this study we discuss how to represent trust values. Two metrics of “trust” and “distrust” are introduced to represent a trust value of an agent, and by applying the idea of the reverse-item averaging method we obtain an actual degree of trust on the agent. The behavior of an agent can be modeled with a state machine, and we can define an extended state augmented with the trust value of the agent. We discuss the transitions of trust values with the extended states.

The purpose of this research is to investigate the serval characteristics concerning the pain with the described subject evaluation method using the bipolarity multi-axis concentric circle scale in a university students group. The participants were asked to enter an appropriate onomatopoeia in the parentheses at the top of 12 axes and to put an x mark on the suitable position on each bipolarity scale and perform some evaluations of the degree of presumed characteristics concerning the pain. The participants were instructed to assume the intensity of pain that they felt when a snapped rubber band hit their fingertip as a unit of pain intensity. The participants were also ask to evaluate each word showed the pain of how many times of strength of the painful unit with a numerical value more than 0. The correlation coefficient of the median, mode, and mean of evaluated scores with the bipolarity multi-axis concentric circle scale and the mean score of the numerical value evaluation of more than 50 onomatopoeias were calculated. The high positive correlation coefficient (all r>0.65, p<0.01) suggests the criterion related validity of the described subject evaluation method with the bipolarity multi-axis concentric circle scale in this investigation.

The complex refractive index of a subwavelength-in-diameter silver pipe has been analyzed by a model of two-layer concentric cylinders, i.e. an inner-layer silver pipe (inner pipe) and an outer-layer silver pipe (outer pipe). A field emission scanning electron microscope image of the silver pipe made by a sputtering process shows that the outermost surface of a cluster of silver nanoparticles can be assumed to be composed of two layered concentric pipes, the inner pipe and the outer pipe, whose complex refractive index is different from each other. The pipes are illuminated by a plane-polarized laser of wavelength 660 nm at normal incidence. The angular distribution of the light intensity scattered by the pipe is measured and fitted using the model calculation. The fitting results show that the optimum value of the index can be determined, but that uncertainty ranges cannot be determined. It is presumed that it is caused by fuzziness of the bound between the inner and outer pipes.

Recently, in order to view TV lives and appreciate sports watching or analyze video images for those sports videos taken by multiple video cameras, a technique interlocking multiple cameras using wired or wireless and taking videos with a time synchronization has been developed. However, the existing method has a problem that the cost of special device introduction is large. On the other hand, there is a method using an inexpensive device such as a smartphone has also been proposed. However, according to the different processing speed depending on performances of varies terminals, there will be a time lag occurs in the videos captured by different terminals Also there is a problem that the video cameras usually used in sports have not the information communication function to synchronize time. Therefore, in this research, we try to clarify the problem in time synchronization of multiple videos taken by different cameras usually used and consider the synchronization technique using image information projected in videos taken by those cameras.

According to the Ministry of Education, Culture, Sports, Science and Technology guidelines in 2013 on instruction in athletic club activities, it is important that coaches use not only their own experience, but also understand and use the results of studies by sports science specialists. In recent years, a device that provides measured results immediately after the batting swing has been developed by downsizing of the device following advances in sensor technology and IT techniques. This device measures the bat swing characteristics from the measurement data. However, coaching methodology based on scientific data is not yet established in amateur baseball clubs, possibly because of a mixture of rational and irrational instruction and difficulty in interpretation of the calculated numerical values. Therefore, the purpose of this study was to investigate the effect of using a data-based evaluation sheet on the understanding and motivation for batting practice in high school baseball players, and to examine the coaching method for baseball hitting using a swing measurement device.

Recent years have seen an increasing interest in the use of preventative care to prevent disease and disabilities. Such a care relies on people adopting a healthy lifestyle and in particular a healthy eating habit. To eat healthy, people would need to be aware of the number of calories which they consume. In this paper, we propose a method which could be used to automatically predict the calories of cooked food based on their recipes. We extract the various food recipes, including information about their ingredients and cooking methods from various cooking websites. Doc2Vec is then used to vectorize the recipes, the features of which are used to train a calorie prediction model. This model is then used to automatically calculate the calories of unknown food recipes submitted by the user.

The smart wellness city is proposed as a typical type of healthy city that provides safe and sound life for citizens with physical health as well as value of living. In the study, smart wellness cities are classified according to the structural factors in healthy city. Firstly, five aspects are defined for urban health. Secondly, structural factors of smart wellness city are described quantitatively. Thirdly, fuzzy cluster analysis such as fuzzy c-means method is applied to classify the smart wellness city. Furthermore, essential structural patterns for health policy factors are derived from the previous results. Finally, the effectiveness of health policies in the smart wellness city is evaluated.

Onomatopoeias are useful expression to communicate intuitive image of designs. There are some approaches adopt onomatopoeias to design computer graphics, robot motion or calligraphic fonts. On the other hand, it is difficult to achieve optimal design using only onomatopoeias. Nakamura et al. introduced onomatopoeias to set proper initial values of design parameter in order to obtain optimal design of the calligraphic font. Nakamura’s experimental result showed that the proposed method is useful to reduce user-load to design the calligraphic font. But the experiment was conducted for the fonts whose appearances were easily described by using onomatopoeias. Our study aims to re-evaluate Nakamura’s method. This paper utilized the fonts whose appearances were difficult to describe by onomatopoeias. The paper showed the experimental result and discussed the usefulness of Nakamura’s method.

We propose a new loss function derived from variance of generated images of GAN(Generative Adversarial Networks), and applied the function to a generative model of GAN. Using the function, it is expected to avoid mode collapse and generate various images. In this study, GAN with the function was examined how various images can be generated and which parameter can contribute to the variance value through illustration dataset. Outcome of the study has potential ability relevant to auto-generators of illustration. Auto-generators of illustration can assist human creative activities and entertainment .

The previous study has proposed a system for estimating a user’s field of view by an RGB-D camera and making a map of the his/her awareness on the assumption that he/she recognizes objects in his/her view. Furthermore, the system has been improved to be able to update the awareness map correctly even if objects on the maps move. However, the previous system cannot update the maps when the moving objects are occluded from the camera. In this study, we propose a system that can update the awareness map robustly against occlusions by tracking a moving object using color and depth information.

Conventional research suggests that emotional expression of robot affects human’s motivation. There are studies that studied the learning effect of children by an emotional expression of a robot, but there are few studies using the emotional expression by a humanoid robot to improve human motivation in sports or games. In this research, in order to investigate the influence on human motivation improvement by emotional expressions of a humanoid robot during a game, we conduct experiments of repeatedly performing scissors between a humanoid robot that expresses emotion and a person. Experimental results verify the influence of the humanoid robot emotional expression on human motivation.

A problem which arises in a human organization is often a multi-objective optimization problem or a single-objective one in which different fitness functions brought by humans coexist. In such cases, we need a two-stage process which consists of producing diverse solution candidates by computers and selecting a final solution from among the solution candidates by humans. Then, a method which unifies the two stages for efficient search is needed. In this study, we propose an evolutionary algorithm in the style of multi-agent system as one of the unified methods. In addition, we apply the proposed algorithm to the task in which a human group creates its representative avatar and show the basic usefulness of the algorithm.

We develop two types of simple population-based search methods that model two types of scale-free behaviors of individuals. The scale-free behavior is a particular behavior of individuals of species that search for food not cooperatively but independently. One type of the scale-free behaviors is that a moving distance of an individual from the present food source follows a power low distribution, which is called the spatial scale-free behavior. The other is that a staying duration of an individual at the preset food source follows a power low distribution, which is called the temporal scale-free behavior. In addition, we conduct simulations assuming static and dynamic environments in which a position of a food source is not changed and changed, respectively, and compares the characteristics of the two search methods.

In the application of Genetic Algorithm (GA) to real problems, acquiring multiple satisfied solutions which satisfy a certain conditions is often required. In addition, to help design choices, satisfied solutions are needed to be different design variable patterns one another. In this paper, a method for acquiring multiple satisfied solutions is proposed. To investigate the applicability of the proposed method to real problems, the proposed method is applied to a many constrained two-objective optimization problem, and compered with Island model which is the representative conventional method maintaining the diversity of design variables.

In recent years, evolutionary computations have been actively applied to engineering applications. In these real engineering problems, the characteristics of objective functions are different one another, which causes the differences of search difficulty among objective functions. As the result, while the optimization of the objective functions with low search difficulty progresses, that with high search difficulty does not proceed. Therefore, there is a possibility that the search is biased. In the real problem, the objective functions should be optimized equally. In this paper, we propose a method to distribute the search resources in preference to high difficulty objective functions based on their search difficulties, and we study the performance of the proposed method using a benchmark problem.

According to the White Paper on Tourism for 2017, approximately 24 million foreigners come to Japan for sightseeing in 2016. It can be said that it is just the time of sightseeing right now. Therefore, the necessity for using English, which can be said to be a world common language, has been increased. As a measurement of English communication skills needed on the site, the “Certification for Bilingual Medical Staff” started in 2011. In this study, English sentences in a workbook that follows the “Certification for Bilingual Medical Staff” were investigated to educe the metrical linguistic characteristics. In short, frequency characteristics of character- and word-appearance were examined.

On March 14, 2015, the Hokuriku Shinkansen line started its operation between Nagano and Kanazawa Station. The “Nagano Shinkansen” operated between Tokyo and Nagano was extended to Joshin-etsu and the Hokuriku region, and it came to connect Tokyo to Kanazawa in Ishikawa Prefecture in about 2 hours and 30 minutes (about 3 hours and 45 minutes before the Shinkansen started its operation). Moreover, although it had been called the “Nagano Shinkansen” for convenience’ sake, it was renamed the “Hokuriku Shinkansen” at this time. The approach from the metropolitan area to “Toyama prefecture,” whose charm is nature including Tateyama and Kurobe, and “Ishikawa prefecture,” whose capital Kanazawa is popular for its history and tradition of Kaga-hyakumanngoku was greatly improved. In this study, what evaluations the Hokuriku Shinkansen has received by its users for these three years were investigated using online reviews.

Deep cases are semantic relations between nouns and verbs. Deep case identification is important to grasp semantic information of sentences. Previous studies used syntactic and dependency features to identify deep cases. Recently, there have been studies using word embedding and artificial neural network. In this study, we propose the method to identify deep cases for Japanese using word embedding.

I aim to establish a dialogue system non-dependent with language. in this paper, the implications of what input / output sounds contained are verified.

”WoM-NAVI” JAVA application is a tool software for constructing recommender system utilizing questionnaire data on users’ hobbies and preferences, where collaborative filtering is achieved based on simultaneous fuzzy clustering of users and items. In this research, a case study of constructing travel recommendation system is reported, where questionnaire data on preferences on various tourist spots in Osaka.

LogAnalyzer3 is a tool that analyzes game logs of RoboCup soccer simulation 2D league. The recent development of a soccer team requires quantitative analysis of soccer games. Therefore, we developed LogAnalyzer3, a tool that extracts useful features for analyzing the game logs. This tool can extract features used in real soccer. This presentation introduces LogAnalyzer3 and its applications.

We developed a fuzzy-set manipulation system in Common Lisp, registered to SOFT-CR, which provides facilities for manipulating various kinds of fuzzy sets in Lisp environment. We can define ordinary, type n, level-m, L-fuzzy sets and relations and their combinations, and operate them with a lot of built-in functions. The built-in functions can apply to any kinds of fuzzy sets. In this paper, we illustrate basic facilities of the system with several examples. The system runs on Windows and Unix operating systems.

For collaborative idea generation, we proposed an AR system that satisfies the following: 1) to participate by multiple people, 2) no limitation of environment, 3) various input methods, and implemented a prototype system. However, each function of prototype implementation was evaluated only. For that reason it has not been verified whether our system was able to provide intellectual support or not. In this research, we analyze the problems in the prototype system for improvement and implement a system for evaluation. And also we aim to verify and evaluate whether new system supports collaborative idea generation intellectually.

In this paper, we report a real-time feed-back system for a making report on mutual viewing-based learning. The system has a function to feed-back shortage viewing time. We adopted summary viewing time of self and other learners' learning logs as a feature vector. To develop the system, we solve a condition to increase academic scores based on entropy. As the results, theaters can understand learners that use ineffective viewing, real-timely.

Electronic bulletin boards are utilized in various ways in modern society. These days, researches related to data science with vast data accumulated in bulletin boards or learning analytics in educational institutions are increasingly demanded. This study analyzed targeting the data written into the electronic bulletin board for information sharing used by the student staffs who we are collaborating to make continuous improvement of an active-learning classroom created at a technological college. Discussion process visualization system was used for this analysis. Various analyses of the data have been in the progress already, and we conducted network analysis this time from the viewpoint of multi-vocality, which has been noticed in recent years. Network analysis enables to visualize change in how people interact with each other based on data and also analyze interaction or chronological changes between participants. We particularly focused on the collective cognitive responsibility of the student staffs from the interaction at the bulletin board and carried out the analysis.

The rubber hand illusion (RHI) is well-known phenomenon, that the sense of ownership is extended over to the external objects in external area, is one of useful experimental models to understand the characteristic of somatosensory ownership. In this study, we made a real extended rubber hand with variable length of arm, and performed RHI by synchronous visual and tactile stimulation. The intensity of RHI was objectively evaluated by the questionnaire report, by the correlation of the electroencephalogram (EEG) activity between the electrodes labeled as C 3 -C 4 in international 10-20 system, and by electromyography accompanied with hand movement. As a result, the correlation of the EEG of C3 and C4, and the score of the questionnaire report increased during repetitive induction of RHI, suggesting that the effects of RHI-induction accumulated. In addition, even without subjective RHI, tension of muscle and c3-c4 correlation of EEG signal increased in some experimental participants, suggesting that RHI is often performed without awareness.

Engineering for estimating human emotions based on cognitive science are recently applied to neuromarketing and so on. In this study, we focused on the extraction of specific brain activities accompanying the emotional context of video contents. Brain activity was measured from the forehead by using wearable fNIRS. In order to reduce the influence of the stress with measurement environment and the influence of contents other than the alternative scene part, expected to induce emotional activity, every contents consist of the same characters and scene place. In addition, the video contents was composed in monochrome without sounds. As a result, the cerebral blood flow was remarkably reduced accompanied with viewing the alternative scene part expected to evoke sorrow emotion, in which the characters crying. Significant difference was observed in the brain activity between in watching the video with sorrow scene and the scrambled video without any emotional contexts. These results suggest that the activity of the frontal lobe is suppressed by seeing negative emotional scene.

There are individual differences in acquisition of motor skills, and recognizing of individuality to motor skill learning is useful for efficient training. Some studies suggest that the activating area of the brain differs before and after training of motor skills, suggesting that the brain activity is switched before and after the establishment of the learning. Therefore, it is considered that switching of brain activity mode is related to the speed of motor learning and accuracy of the motor skill. In this study, many experimental participants increased not only the time between task blocks, but also the time for pressing 3 buttons in the block by swapping the order of the blocks, as upper hierarchy. Although the button pressing order in the block is the same, the time for button pressing time in the block increased, because the mode of the brain activity for automatically performing the motor action is switched to the mode that consciously controls the movement of the pressing the button in a certain order. The result suggests that in acquiring hierarchical motor skills, many people do not strongly separate the motor learning among the hierarchies, and it can be interpreted that brain attempts to learn again from the lower hierarchy when new patterns of order in the upper hierarchy is required to learn.

In education for children and guidance of sports, it is important to give appropriate instruction to learners. It is necessary to grasp the ability and characteristic of the learner by observing the learning process and to change the teaching method as needed. In this paper, we consider the learning parameter and appropriate giving rewards method, using simulation data which makes agent learn maze. For learning of the maze, we used Q-learning well known in the field of reinforcement learning. And we conducted experiments using multiple agents with different learning parameters. Agent behavior data at the middle stage of learning is classified by SOM and learning parameters are estimated. After that, we change the giving rewards method, and consider it according to the learning parameters from learning result.

Rough set theory was proposed by Z.Pawlak in 1982. This theory can mine knowledge through a decision rule from a database, a web base, a set and so on. The decision rule is used for data analysis as well. And we can reason for an unknown object using the decision rule. However, there are cases where too many decision rules are found. It is difficult to acquire knowledge from too many rules. However, the C.I. value of the merged decision rule may be low. In this paper, we propose a method to reduce the number of rules by merging decision rules. It is difficult to acquire knowledge from many rules. It is also difficult to find knowledge from rules with a large number of condition attributes. We propose a method to reduce condition attributes. We propose a method to reduce the rule of low C.I. value.

In multi-layered neural networks, the back propagation algorithm provides a means of training the network to perform supervised learning tasks. This is viewed as a means of implementing the gradient descent optimization through calculation of the derivative the error function. However, it is generally known that this function has a large number of local minima, and this algorithm does not guarantee convergence to a global minimum. We derive the Hessian matrix of the error function in the 3-layered neural network under limited conditions and show the sufficient condition for convexity.

The purpose of this study is to investigate the awareness of WJBL (Women's Japan Basketball League) retired players toward dual career and to examine the way Dual Career Education for elite athletes. As method of this study, questionnaires were conducted 45 WJBL retired players for grasping the awareness of dual career. Based on the WJBL retirement survey, this paper considers the dual career education of the athletes.

The purpose of this study is to investigate the awareness of WJBL (Women's Japan Basketball League) retired players toward dual career and to examine the way Dual Career Education for elite athletes. As method of this study, questionnaires were conducted 45 WJBL retired players for grasping the awareness of dual career. Based on the WJBL retirement survey, this paper considers the dual career education of the athletes.

In this paper, we try to apply graphical lasso to a problem for maintenance of sewerage systems. Using graphical lasso in inspection data of sewerage system, we will grasp a change of structure between normal systems and abnormal systems, and as a result, show which inspection items contribute to detecting abnormal sewerage systems.

This research and development aims to contribute towards solving social problems through systems that collect and analyze human's highly accurate position information and data such as biological information with real time. In Japan, policies regarding supports are actively advanced towards the Tokyo 2020 Olympic Games. As the initial target of this achievement, we are focusing on the sports field, and we will contribute to the development of quantitative performance of sports athletes, health promotion, and new business creation. Specifically, we developed real-time sensing devices that can be worn on the human body, reference station devices for simultaneous parallel processing of RTK processes to improve positional accuracy, and station devices for data collection. In order to offer these developed devices at low cost to the market, we studied optimal sensor combinations. In this time, we carried out software development applied to soccer and at the same time carried out accuracy verification / operation experiment for commercialization. This result can be expected to be utilized not only on sports filed but also management of construction equipment and agricultural equipment, application to UAV, management of aged wanderers, management of animals at pastures, etc.

In Japan, supporting activities in terms of scientific research and so on for the purpose of improvement in performances of Japanese national athletes and competition levels are required towards the 2020 Olympic Games. Under these circumstances, ICT (Information and Communication Technology) technologies have been developed in the area of sports. The authors have also developed the visualization system for field sports. The system can provide coaches with the ground for instruction by quantifying and visualizing the performance of individual athletes. However, it has a problem that team performance like formations cannot be grasped precisely. In this study, we develop a new visualization system that is capable of confirming team formations for soccer. Then the performances of individual athletes as well as the whole team are visualized in actual games, and practical usability of the system is verified based on the point of view whether it can provide coaches with qualitative information of team formations.

In our country, policies regarding sports are actively advanced towards the Tokyo 2020 Olympic Games. One of those policies, "Sports x ICT" considers effective methods of utilizing ICT (Information and Communication Technology), such as development of measurement instruments, measurement and visualization of data, and proposals for new services in the field of sports. Against this backdrop, we have been developing the visualization system for American football games using terminal devices included GNSS and acceleration sensor. Using that system, it is confirmed that the effective information which are the individual condition and the motion analysis of american football players against not only players but also the college football leaders can be grasped. But, in the our existing research, we could not realize the strategy for analyzing games selected for play calls depending on circumstances, to make a prediction of successful ratio, and so on. Then, in this paper, we aim to analyze the match-up for offence and defense players related to pass plays using deep learning. Reasoning the success rate of pass plays, it is capable to provide the new awareness for American football strategies to the manager and coaches.

In recent years, the research on automatic dialogue systems (Chat dialogue systems) attracts attention. Dialogue systems are classified into task-oriented and non-task-oriented one. In non-task-oriented dialogue systems, it is important that a user want to continue dialogue over a long period of time. In order to improve dialogue continuity, it is reported in conventional researches that utterances including humor and back-channel feedback are effective. In this paper, we investigate a generation of back-channel humor for chat dialogue systems.

Humans tend to prefer a little steady profits than a large risky profits. However, they also tend to take a risk to avoid a loss. Therefore, it is considered that properties of agents that support users must be different according to users’ situations that they are obtaining profits and suffering losses. In this study, we design two types of agents, an unstable agent that brings both of the best and the worst results, and a stable agent that always brings moderate results, and investigate their trustworthiness in situations that the users are obtaining profits and suffering losses. Experimental results show that the stable agent is more trustworthy in the situation of obtaining profits.

Recently, educational-support robots have been attracting increasing attention as learning-support devices. Previous research has proposed the sympathy expression method that robot expresses emotions which sympathize with learner. In previous research, the sympathy expression method has been used for the robot which expresses its emotion by face. However, no research for sympathy expression method using the robot which expresses its emotion by using body or unifying body and face. Therefore, this paper examines the learning effect and the impression about three types robots which have different emotion expression methods in long term experiment.

Recently, educational-support robots that support human learning attract many people's attention. However, the problem of these robots is that learners feel that the behavior of the robot is uniform and tires of collaborative learning with the robot. To solve this problem, the previous study proposed using sympathy-expression method that is defined as the robot expressing its emotions autonomously. This method expresses emotion according to the answer time. We think that the writing time should be considered. The writing time is defined as the time of learners write to solve the questions. For example, when learners wrote a lot, the answer time became long time and the robot, which used the previous method, expresses the emotion of the sleep side. To sympathize with learners, we think that the robot should express the emotion of the arousing side. However, since we do not know the impression of the robot that expresses emotion based only on the writing time, we investigate the impression of the robot that expresses emotion based on the writing time.

This study aims to construct a deterioration diagnosis support system for coating of steel structures. This paper discusses an image background removal method, that is proposed on a trial basis, as a pre-processing unit of the system. This method is constructed by a sequential combination of the following three algorithms: Mosaicing, FloodFill, and GraphCut algorithms.