This paper considers fuzzy random two-level integer programming problems under noncooperative behavior of the decision makers. Having introduced fuzzy goals of decision makers together with the possibility measure, following expectation optimization, fuzzy random two-level integer programming problems are transformed into deterministic ones. Extended Stackelberg solutions are introduced and computational methods are also presented.

View full abstract

In this paper, we focus multiobjective 0-1 programming problems under the situation existing stochastic uncertainty and vagueness at the same time, formulate it as fuzzy random multiobjective 0-1 programming problems that its objective function is fuzzy random variable, suggest a decision making model based on an idea of possibility.

View full abstract

Two major approaches to deal with randomness or impression involved in mathematical programming problems have been developed. The one is called stochastic programming, and the other is called fuzzy programming. In this paper, we focus on multiobjective integer programming problems involving random variable coefficients in constraints. Using the concept of simple recourse, such multiobjective stochastic integer programming problems are transformed into deterministic ones. As a fusion of stochastic programming and fuzzy one, after introducing fuzzy goals to reflect the ambiguity of the decision maker's judgments for objective functions, we propose an interactive fuzzy satisficing method to derive a satisficing solution for the decision maker by updating the reference membership levels.

View full abstract

This paper describes a prototype system which generates operetta songs fitting to story scenes represented by texts and/or pictures. Inputs to the presented system are original theme music, numerical information on given story scenes and story texts. The presented system outputs sounds, playing songs generated based on impressions of story scenes. A song generation process consists of two phases, the initial song generation phase and the interaction phase. In the initial song generation phase, the presented system generates variations on theme music and lyrics according to image of music and lyrics obtained from numerical information on given story scenes. Evolutionary computation is applied to this phase. Using a vocal synthesizer and a general midi synthesizer, the present system plays songs as the variations on theme music with the lyrics. In the interaction phase, the present system reflects uses's Kansei to variations on theme music and lyrics by using interactive evolutionary computation. This paper also describes the pre evaluation experiments to confirm whether the generated songs reflect impressions of story scenes appropriately or not.

View full abstract

Stock Layout problem exists in many industry fields, which is one of the mathematical problems of NP-Complete. In practice, the problem is solved under the various technological requirements for production process such as guillotine cutting, cutting line width, texture direction, and stock selection from enough etc. The purposes of this paper are to propose (i) flexible coding method of the problem, which enable us to solve the problem under the requirements, and (ii) a fish swarm optimization method. Appling the proposed method to practical projects and comparing with other algorithms shows its validity.

View full abstract

In this paper, a new optimization method is proposed to minimize the evacuation time of passenger using the urgent evacuation in an aircraft accident. In order to minimize the total evacuation time of the passenger, a quasi-optimization technique is examined as a multi-agent problem using an effective evacuation guidance by cabin attendant or the urgent safety system in an aircraft accident. The urgent evacuation behavior is formulated as an autonomous agent and multi-agent system (AAMAS) model evolving over a two-dimensional grid cell that represents the aircraft cabin and passenger. However, it is impossible by AAMAS simulation to forecast the minimum evacuation time in such chaotic situation of aircraft accident, because it is too difficult to solve the minimum evacuation time of multi-agent. Therefore, a new quasi-optimization technique is proposed by the recursive procedure for obtaining the adaptive and satisfactory minimum evacuation time from the cabin to the ground on the airport during the emergency accident of aircraft.

View full abstract

In this presentation, Kansei engineering is applied to the landscape analysis of the historical town. Chofu area in Shimonoseki city was taken up as an example of the analysis, and the landscape is evaluated based on an actual photograph. The design that reflects the climate and the history in the region is seen well in the landscape. This shows a characteristic of the region. In the city planning, it is the necessary to unify such images in the regional landscape. However, some problems are frequently seen in such area where the image is not unified around some specific streets. Then, "Image held in landscape and image of the points it the region" was analyzed. SD scaled questionnaire is taken in the kansei experiment and multivariate analysis, and self-organizing neural network (SOM) were used as methods of the data analysis.

View full abstract

In this presentation, Kansei engineering is applied to the landscape analysis of the historical town. Chofu area in Shimonoseki city was taken up as an example of the analysis, and the landscape is evaluated based on an actual photograph. The design that reflects the climate and the history in the region is seen well in the landscape. This shows a characteristic of the region. In the city planning, it is the necessary to unify such images in the regional landscape. However, some problems are frequently seen in such area where the image is not unified around some specific streets. Then, "Image held in landscape and image of the points it the region" was analyzed. Kansei engineering experiment is taken, and the results are transformed to landscape image maps.

View full abstract

This paper proposes an emotion-generation model for complex change using a chaotic neural network (CNN). Using a CNN, the proposed model will solve the problem of past studies that have indicated that robotic emotion changes are simplistic. The model uses the principle of an adaptation level, which is used in Russell's emotion model to generate emotion. This paper considers the effectiveness of this approach using simulation, and shows that the model can express a change of "adaptation". In addition, through the chaos of CNN, the proposed model can express different changes, even if the values of CNN's input values remain the same.

View full abstract

In this paper, we propose an information processing model for a kansei robot. This model handles memory based on human psychology. We expect that on incorporating the model, a robot can exhibit human characteristics because of using psychological memory. To verify the model, we first perform a comparison between the results of the experiment performed using this model and that of an actual psychological experiment. The results of the comparison suggest that the memory functions of the model are similar to the human memory functions. Second, we conduct the process of learning movement actions to verify that the robot on which the model was implemented learned movement for moving to many places and decreasing its curiosity.

View full abstract

In this thesis, it proposes the sensibility retrieval model to imitate user's sensibility by using the neural net work. The neural net work is a system that models the function of the brain, and the output part is assumed to be stimulation to which the person received the input part of the neural net work corresponding to feelings when the person receives the stimulation. It optimizes it to the relation to feelings to stimulation and the stimulation that the user received the I/O relation of the neural net work by adjusting the uniting gaining the weight of the neural net work.

View full abstract

AHP (Analytic Hierarchy Process) has been widely used in decision making. Inner dependence method AHP is one technique even in case of criteria have dependency. However using original AHP or Inner dependence method, the results often lose reliability because the comparison matrix does not always have sufficient consistency. In these cases, fuzzy representation for weighting criteria is useful. In this paper, we refine fuzzy weights using results from a sensitivity analysis. With experiment data, we show its usefulness when the comparison matrix is not sufficiently consistent.

View full abstract

The interactive evolutionary computation (IEC) is a technique to the development support of a product fitting human's Kansei by a computer. The great problem of the IEC is reduction of the burdens when a user evaluates it. One of the methods to reduce the evaluation burden of the IEC user is improvement of the evaluation interface. A general evaluation interface of IEC is an n-stage evaluation method of all of the individuals that the IEC system shows. However, it becomes the big evaluation burden for IEC user if many individuals are shown immediately. Therefore we suggest the interactive tabu search (ITS) to aim for the reduction of the evaluation burden of the IEC user.

View full abstract

This paper aims at the construction of a partner agent, which plays a seven-card stud poker game interacting with a human player against an opponent player. Seven-card stud poker is a kind of a game with imperfect information and it is difficult to consider optimal strategy in such a game. Therefore, a human player has a discussion with the partner agent on game strategy cooperating with the agent. The partner agent presents agent's own decision and facial expressions according to the game situation. And if a human player has some different decision, these decision may be stored as case examples, and these case examples are reused as one of solution candidates of a new situation. This paper also performs subject experiments to evaluate the effectiveness of the proposed partner agent and case examples obtained through discussion.

View full abstract

Car Racing Game is a competition game of computer programs in IEEE CEC 2007 Car Racing Competition, where two car agents compete with each other for taking waypoints in a two dimensional plane. If the current waypoint is taken, the next one becomes the current and a new one is placed as the next waypoint on a random position. The agent can get some information on itself, the other agent, the current waypoint and the next waypoint.In this paper, we evaluate the state of an agent for the current waypoint with a fuzzy rule from its speed and the distance and angle to the waypoint. We decide our agent's strategy, the waypoint to take and the speed, with a fuzzy rule from the evaluated values of the two agents for the current waypoint. When our agent decides to take the current waypoint, it goes to the current waypoint not only straight but also around with considering the next waypoint.We illustrate this method is higher in performance than the second-ranked program in the preliminary league in the Competition.

View full abstract

Car racing game is a game of computer programs in IEEE 2007 Car Racing Competition, where two car agents compete with each other for taking waypoints in a two-dimensional real-number plane. If the current waypoint is taken, the next waypoint becomes the current. In each action, the car agent can get some information on itself, the other car agent, the current waypoint and the next waypoint. In this paper, we apply to it a fuzzy Q-learning that is an extended version of Q-learning, one of methods of reinforcement learning, to acquire the actions that car agent takes the current waypoint without the other car agent. We define the state space of fuzzy sets about the distance and angle to the current waypoint and the speed of car agent. We illustrate this method with several patterns of rewards.

View full abstract

Behavior learning or skill acquisition is one of the important issues of reinforcement learning schemes, in order to realize intelligent agent. Generally, simple tasks such as goal exploration can be easily acquired by reinforcement learning techniques, as many simulation studies are demonstrated. However, complicated tasks such as behaviors in sports like soccer are difficult to acquire substantially. It is caused by difficulties of objective modeling and multi-agent environment. In this study, we develop a behavior acquisition system of soccer agent based on reinforcement learning. We discuss the results of acquisition through numerical examples.

View full abstract

In the Multi-Agent Reinforcement Learning, the learning time and the number of trials increase in proportion to the view width and square of the number of the agents according to the state of the perceptive environment. In this regards, as the means to improve performance by reducing the number of the states, a method of coarse-graining of perceptive information has been proposed. This method has achieved faster learning, however the accuracy of behavior selection decreases, which is the indicator to evaluate learning outcomes. In this report, based on Parallel Learning and Average Residual Entropy, we propose a method of coarse-graining adaptively according to the progress of the learning. This achieved both faster learning and maintaining the accuracy of behavior selection.

View full abstract

Reinforcement learning, which is inspired by behavioral psychology, is a sub-area of machine learning. It is concerned with how an agent takes actions in an environment to maximize the reward given by the environment. On the other hand, there is a condition of "slump" in the process of human skill learning in the real world, and it is reported that the occurrence and the resolution of the slump cause promotion of skills. In this paper, we propose to introduce the notion of slump in the reinforcement learning. By simulating the condition of the slump in the reinforcement learning, the efficiency of learning increases in some conditions. The formulation and the implementation of the slump in reinforcement learning are described, and the result of the evaluation of the performance is presented.

View full abstract

Conventional systems usually use retrieval images using titles and keywords. We developed ``Feelimage,'' an image retrieval system, with fuzzy sets for color distributions and exif information such as date and time. We process color distribution on each pixel of the image to calculate degrees of various colors for the entire image. We can retrieve images with a query such as ``Red flower.'' However, a query of two color expressions like ``Red flower and green leaf'' retrieves images with one of the colors. In this paper, we formulate a methed to retrieve images with a query of two color expressions.

View full abstract

In this paper, we propose a method for classification and retrieval of news contents based on automatically extracted tags. There are some researches on automatic tagging that use TF-IDF to extract important tags in a news article. However, some useless tags are extracted, on the other hand the important ones are not extracted by the method. To solve this problem, the keywords extracted from the title and category of the article are added as a tag to the article e in addition to the tags extracted by TF-IDF. Besides, we performed an experiment that subjects extract keywords from each of the articles with different length. From the result of the experiment, we discuss the appropriate number of tags for the length of an article.

View full abstract

We have many kinds of data of time series such as stock prices. We have proposed a method to express their global trends in a natural language via the representative values on the fuzzy intervals in the temporal axis and their local features via the position of large differences between the original data and the data generated from the global trends. We have also proposed a method to tune the fuzzy intervals in the temporal axis. To retrieve similar time series, we have defined a degree of similarity between two time series with global trends in natural language. In this paper, we propose an integrated degree of similarity of global trend, local features and oscillation. We show that the result becomes more appropriate by the tuning of fuzzy intervals in the temporal axis.

View full abstract

Humans do not always learn using the same knowledge representation. They use concrete knowledge representation at first, but they gradually come to use abstract one as they learn more. We have studied a system that switches knowledge representation, reconstructs knowledge set and switches reasoning method. The system does them as soon as conditions are satisfied.In this paper, we extend the system to switch knowledge representation smoothly. The extended system reasons by using not only a new knowledge representation but also an old one depending on the ratio of swtching. We take the cases into account where the system starts to switch knowledge representation under switching reasoning method. As a result of learnig for some kinds of data, the system shows a different result from the old one. In particular, the rate of correct answer increases for complex data by decreasing incorrect answers with a lack of knowledge.

View full abstract

In human learning, knowledge representation for a little the quantity of learning is different from that for a lot of quantity of learnning. We proposed a learning that switches knowledge representation. Our emotions influence switching of knowledge representation of human. In the previous paper, therefore, we applied the incorrect ratio to Russell's circumplex model by using a fuzzy rule to calculate the degree of emotions, by which the threshold of the incorrect ratio for switching knowledge representation was increased or decreased. We simulated using the iris data set. The result showed emotions settled to joy, and the threshold kept on increasing. In this paper, we propose two methods of controlling the threshold, one is a method for greatly decreasing the threshold if there is even a few incorrect answers after the middle stage of inference and the other is a method for considering saturated state if the same emotion continues. These two methods are compared and examined by simulating the data other than iris data set.

View full abstract

In the representaion, Pavelka's fuzzy logic is reported. He fuzzified not only propositions but also axioms and the rule of inference.

View full abstract

Data Envelopment Method (DEA) has been applied to an efficiency estimation of energy utilization of some cases using CCR and BCC efficiencies. The efficiencies are schematically expressed by the role of CCR and BCC frontiers on the 2-inputs, 1-output diagrams. The frontiers are compared and discussed.

View full abstract

It is well known that fuzzy mapping plays an important and fundamental role in fuzzy analysis. In this paper, we attempt to determine extensive classes of these fuzzy mappings that are represented by their Taylor expansion. We first obtain that a continuous function f (x) can be transformed into a fuzzy mapping f (u) by using the extension principle. Furthermore we discuss some properties of the series of fuzzy numbers. From these properties we get that f (u) has the analytic properties similar to those of the analytic function f (x). Finally as applications, the author gives some examples.

View full abstract

In this paper, we simulate virtual societies under several communication conditions between agents using multi-agent systems whose agents are based on cognitive economic efficiency, to get the clue to understand the modern society which consists of complex human relationships. The model simulates social actors making positive and negative selection of each other in efforts to reach balanced cognitive states. We observe the influence on dissimilarity or the formation of groups on our social simulations as a result of several communication networks and the ways of choosing communication partners, especially for sharing common perceptions, which influences their society.

View full abstract

This paper proposes a new approach to collaborative filtering based on sequential use-item co-cluster extraction, in which personalized information filtering is performed by extracting users and items that are mutually familiar. In order to extract exclusive clusters, a fuzzy clustering method for weighted data is applied to sequential fuzzy cluster extraction. The exclusive constraint is forced only to users so that the clustering process plays a role for connecting promising items to users and achieves appropriate information recommendation.

View full abstract

We have proposed an agent simulator including fuzzy model-based action decision mechanism for behavior analysis of guests having various preferences in theme parks. In this research, we consider the influences of both great preference varieties and alternative selection schemes. It is demonstrated that the alternative selection in the action decision mechanism considering the congestion degrees plays a role for avoiding an over congestion in a certain attraction.

View full abstract

DEA (Data Envelopment Analysis) is a non-parametric method for evaluating the relative productivity of DMUs (Decision Making Units). The evaluation value is estimated by comparing objective DMU with DEA-Reference Set which is composed by efficient-DMUs. In order to clarify the relations among these DMUs by intuition, this paper proposed a visualization method used PCA with variable selection. Numerical study shows that proposed method visualizes these DMUs in 2-D feature map maintained the intrinsic structure of DEA-Reference Set.

View full abstract

When the decision maker gives a pairwise comparison matrix on alternatives, their priority weights are obtained by Analytic Hierarchy Process (AHP). In a group decision making, we assume that each group member gives his/her own pairwise comparison matrix. In this paper, a method for obtaining the interval priority weights of alternatives as the collective evaluation from the given many matrices is investigated. We basically presume that the individual matrices are the possible realizations of the collective interval priority weights in order to reflect all group members' opinions and their widths are minimized. However, as is often in the real world, the individual opinions are not always similar so that the collective intervals might be extremely wide. To avoid this, we relax the presumption by introducing the deviations for the individual intervals from the collective ones. Then the problem is formulated as the biobjective programming problem which minimizes the widths of the collective intervals and the deviations from individual intervals.

View full abstract

Recently research in robot is intended for use outdoor and domestic environments. Structural changes such as joint failure or frame distortion might occur because the operating environment could not be expected. Structural changes make designed motions doing wrong operations. Though with limitation to the motions position control, we propose the sequential motion modification technique in robot structure changes without identification of structure changes and structure exploring. Adjustment factors of joint angles are estimated by observation of control point in structure changed robot and calculated joint angles of inner robot model. When the robot structure has changed, the robot motions are estimated by modified volume of joint angles. Experimental results show that the possibility to modify the robot motions to changed structure with less exploring cost.

View full abstract

We have proposed parallel distributed genetic fuzzy rule selection to design fuzzy rule-based classifiers for large data sets. It can drastically reduce the computational time without deterioration of test data accuracy by dividing not only a population but also a training data set into subgroups. In this paper, we propose ensemble classifier design using classifiers obtained by each subgroup in our parallel distributed implementation. Through the computational experiments with large data sets, we examine the effect of the proposed ensemble method on the generalization ability for the test data.

View full abstract

Japan Aerospace eXploration Agency (JAXA) carries out the sound analysis in rocket launching using Computational Fluid Dynamics (CFD) as a part of the research and development of the rockets. The computation of this analysis employs a parallel distributed processing using some CPUs. To estimate and reduce the total computation time, it is necessary to consider not only the calculation time for the allocated jobs but also the communication time between jobs crossing CPUs. This problem is defined as a single objective optimization problem whose objective function is minimization of the total computation time, which is defined as the weighted linear sum of the calculation and the communication amount at the CPU having maximum value of them. However, in the real-world problem, it is difficult to know the actual value of the weights for them beforehand. Thus, the purpose of this study is to acquire Pareto solutions acquired by the search with various weights, and this paper investigates the effective search method for this problem.

View full abstract

Recently, many evolutionary computation methods such as GA for multiobjective optimization problems have been reported, and then many objective optimization problems, which have a lot of objective functions, have been especially focused on with the background of the improvement of computer performance. One of the goals in multiobjective optimization problems is to obtain various solutions satisfied with at least one objective function. However, the solutions which have better values in an objective function but not in others are not what users need in practical problems, which makes the multipoint search of MOGA ineffective. This paper investigates the method which enables a user to find practical solutions for him/her with changing the direction of the search interactively, and studies the effectiveness of the proposed method using a practical problem.

View full abstract

In this study, we propose an interactive fuzzy data mining technique from time-series data. In our method, fuzzy rule-based systems are extracted as knowledge by evolutionary multiobjective optimization with user preference. A user often does not know own preference before obtaining some solutions. The preference is sometime changeable during search process. To deal with these situations, we define an objective function changeable according to user preference anytime. We simultaneously optimize two objectives: accuracy maximization and user preference maximization.

View full abstract

The liver has a lot of important functions and is needed metabolisms in the living body. The liver diseases are many signs but their diagnosing has some difficulties. For example, the diagnosis of liver diseases is needed knowledge of a great of number of blood test items (e.g. T-BiL , Alp, gamma-GTP, LDH, ChE, GPT, GOT, Plt, Alb, AFP, and so on) and of the liver diseases(Acute hepatitis, Chronic hepatitis, Liver cirrhosis, Hepatoma)under the mutual relation. Then the researchers have challenged to this problem with several computational technique. We propose a new diagnosis procedure for liver diseases using Mahalanobis distance in MTS with Rough Sets. Mahalanobis distance of MTS is a distance measure including the variation in data. By using Rough sets, we can extract knowledge of mathematical principle. They will be applicable to knowledge extraction, prognosis prediction, etc. besides a diagnosis method like this. This thesis will give an examination the usefulness of the data analysis of the medical by MTS from now on.

View full abstract

We are using color in many aspects of daily life. For instance, when seeing a color-coded image such as a subway map and a weather map, we distinguish objects or identify an object by color. Additionally, when seeing a traffic signal and when selecting ripe tomatoes from ones at various stages of ripeness, we determine their condition according to their color. In the former case, it is necessary to distinguish regions painted in different colors, but it is not necessary to know color of each region. By contrast, in the latter case, it is necessary not only to distinguish colors but also to know color of each object because each color represents a state. However, in both cases, it is often difficult for people with color blindness. With the aim of solving the problem in the latter case, we developed a color transform method for assisting people with color blindness to guess a normal color, and also developed a software in which that method was implemented.

View full abstract

In this paper, in order to reduce the business burden that the caregivers in welfare center, we propose a development of system that can efficiently conduct meal management. General meal management support systems have been used at the welfare centers might be tough to use for caregivers, although those systems have many kinds of functions for dietitians. Our purpose of development this proposal system is to record and inspect data easily. In the concrete, this system makes caregivers possible to presume the intake calorie from the content of the meal, and input data using mobile phone.

View full abstract

The nurse scheduling problem(NSP) has been solved by several methods. Among them it is pointed out that the Cooperative Genetic Algorithms(CGA) are suitable for NSP. In CGA there usually exist two objects, one is to maximize the average evaluation value, and the other is the minimization of the variance. Kawaguchi et al. have proposed the weighting method to give an effective intersection of CGA. They also showed an algorithm for NSP using the method. However, the weighting method has a difficulty in finding suitable the weights. In this paper we propose CGA using the extended information ratio, an index for a portfolio optimization problem, and analyze the characteristics of the obtained solution.

View full abstract

Recently, semi-supervised clustering has been remarked and discussed in many research fields. In semi-supervised clustering, pairwise constraints, that is, must-link and cannot-link are frequently used in order to improve clustering properties by using prior knowledges or informations. In this paper, we will propose a semi-supervised clustering by using clusterwise tolerance and pairwise constraints. First, the concept of clusterwise tolerance and pairwise constraints are introduced. Second, the optimization problem of fuzzy c-means clustering using clusterwise tolerance based pairwise constraints is formulated. Third, a new clustering algorithm is constructed based on the above discussions. Finally, the features of proposed algorithm is verified through numerical examples.

View full abstract

In recent years, there are many studies to classify data automatically. There are many methods to realize data classification, and there are two types: supervised clutering and unsupervised clustering. Each method has both advantage and disadvantage.Thus, semi-supervised clustering has payed much attention these days, aim to cluster many data and get any result as one's like. One of an important method is pairwised constraints, using must-link and cannot-link. In this paper, we will propose a new method of pairwised constraints using penalty vectors. We will show some examples of the new method and demonstrate our new algorithm.

View full abstract

Switching regression models are known to be useful in real applications. Semi-supervised clustering is also well-known to be valuable and many researchers study it recently. Although these algorithms are very useful, there is one drawback. The results have a strong dependency on the predefined number of clusters. To avoid this drawback, we apply a method of sequentially extracting one cluster at a time using noise-detecting method to semi-supervised switching regression models which enables an automatic determination of clusters. We show the effectiveness of the proposed method by using numerical examples.

View full abstract

Recently semi-supervised clustering has been remarked by many researchers. Using pairwise constraints for clustering is one way of semi-supervised clustering. In this paper we discuss agglomerative hierarchical clustering using pairwise constraints. To introduce Cannot-link into centroid method and Ward method, we propose two ways, one with kernel functions and another without them.

View full abstract

While AO have been discussed as an operator for scalars but have not been discussed in a vector space, even though, a lot of data analyses treat data as vectors. We propose generalized aggregation operators (GAO) extended from aggregation operators (AO) to operators based on vectors. We apply GAO to a kernel fuzzy clustering model. Conventionally, the model uses an inner product to represent the common property; however, the real world data has variability for the common property. This means that we can not always capture the common property of objects by using the inner product. Several numerical examples show better performance for the application of GEO to the kernel fuzzy clustering model.

View full abstract

Exercises can be classified into the physical fitness, which is for the physical health care and the emotional fitness, which is for the mental health care. Especially in the emotional fitness, it is required that the exercise intensity is controlled naturally in contrast to mandatorily in the physical fitness. In this report, the authors propose a natural control method of the exercise intensity by movements according to rhythm of music. First, the relation between music and exercise intensities calculated by Karvonen Formula is clarified. Next, we propose a music change timing for the hart rate control based on the fuzzy PI control. Finally, usefulness of the proposed control method is verified through jogging experiments using treadmill.

View full abstract

In order for the robot and human to live together, the flexible understanding ability of human intention is required for the robot. In this research, we aim to realize "Interactive Emotion Communication" (IEC) which is a bidirectional communication based on emotional behaviors between human and robot. The purpose of IEC is to raise the personal affinity which the robot gives to the human by interactive emotional behaviors between both of them. We consider that patterns of emotional behaviors desired to the robot vary from person to person in IEC. In this research, we perform the individual preference analysis for emotional behaviors, and then construct "effective reaction model" which the robot performs desired emotional behaviors. As a result, we aim that the robot gives high interpersonal affinity to the human. To confirm the efficiency of the constructed model, we performed the comparison experiment used a pet-type robot through IEC to the human, and also report the result of this experiment.

View full abstract

In this study, we focus on the gestures appeared in general negotiation dialogs. By analyzing the relation between situation and gestures, we suggested a technique to let the agent perform an adequate human-like gesture and evaluated whether the gesture of an agent could give a similar impression to a human. We collected a number of negotiation dialogs in order to investigate what kind of common gestures people usually use. We investigated the appearance frequency of the gesture in different situations and extracted the gestures with a high appearance frequency. Based on the quantity of the characteristic, the agent gesture model was made. We performed an impression evaluation about the gesture that human and an agent performed by a questionnaire. As a result, we verified that the agent can express the same state of mind as people can by generating an adequate human-like gesture.

View full abstract

In this report, the authors propose a comfort index CP using the acceleration plethysmograph for Emotional Fitness, which is an exercise for mental health cares. The CP is defined by a multiplication of 'coherence' between stable and after exercise acceleration plethysmograms and 'power spectrum' of the after exercise acceleration plethysmograms. First of all, we run for ten minutes by using the treadmill, and measure the acceleration plethysmogram, and the two dimension mood scale (TDMS) before and after exercise, which express human comfort. We propose the value in which multiply the power spectrum by coherence from acceleration pulse wave as refreshment index (CP). Next, we experiment on Emotional Fitness, and shows correlation between CP and index of comfort using TDMS(index of comfort after exercise). Through result of experiment, we transform CP into LF / HF. The change after exercise is different, but we think that all examinee get refreshment. Therefore we think that CP proved effectiveness as the refreshing index. Finally, we confirm that CP expresses human comfort after exercise.

View full abstract