Information Visualization is a technology for or-ganizing and presenting data so that humans can easily examine it. Recent growth of computer systems and the Internet has brought us various kinds of, vast amount of data. As the cost of getting information decreases, how to make use of the obtained data is becoming the crucial problem for us. There are various ways of making use of data with the help of computer systems, such as data analysis, data mining, information retrieval, and filtering. However, such processing should not be performed by computer systems only, but by the collaboration between humans and the systems. Information visualization has been paid attention to one of the important technologies for supporting the interaction be-tween humans and computer systems, and applied to various fields. Furthermore, information visualization can also support the interaction between humans, by helping the share of data and context. Therefore, information visualization is one of the key technologies for establishing fruitful interaction as a basis for our intelligent activities. In this talk, I give the survey of information visualization and its application to interaction support. The talk briefly explains the concept of information visualization as well as interactive information visualization, and introduces our recent research outcomes including BBS with visual representation, visualization of spatiotemporal trend information, and informa-tion visualization for interactive information retrieval.

View full abstract

Biomimicry, the design of articial systems that mimic natural behavior, is recently attracting considerable interest. Biomimicry requires a reverse engineering process; the behavior of a biological agent is analyzed in order to mimic this behavior in an articial system. In many cases, biologists have already studied the relevant behavior and provided a detailed verbal description of it. Mimicking the natural behavior can then be reduced to the following problem: how can we convert the given verbal description into a well-dened mathematical formula or algorithm that can be implemented by an articial system? Fuzzy modeling (FM), with its ability to handle and manipulate verbal information, constitutes a natural approach for addressing this problem. The application of FM in this context may lead to a systematic approach for biomimcry, namely, given a verbal description of an animal's behavior (e.g., the foraging behavior of ants), apply FM to obtain a mathematical model of this behavior which can be implemented by articial systems (e.g., autonomous robots).

View full abstract

IAFC(Integrated Adaptive Fuzzy Clustering) neural network, which has both stability and plasticity, is a fuzzy neural network. IAFC neural network uses a fuzzy learning rule that is based on fuzzification of Kohonen learning rule and fuzzification of leaky learning rule. Supervised IAFC neural networks are supervised versions of IAFC neural network. Learning rules of Supervised IAFC neural networks are based on fuzzifications of LVQ (Learning Vector Quantization). Iris data set is used to compare the performance of IAFC neural network with those of Kohonen Self-Organizing Feature Map and Fuzzy c-Means and to compare the performances of Supervised IAFC neural networks and those of LVQ algorithm and backpropagation neural network.

View full abstract

Fuzzy c-means-based classifier derived from a generalized fuzzy c-means (FCM) partition and tuned by particle swarm optimization (PSO) has been proposed. Since different types of classifiers work best for different types of data, our approach is to parameterize the classifiers and tailor them to individual data set. The procedure consists of two phases. The first phase is an unsupervised clustering, which is not initialized with random numbers, hence being deterministic. The second phase is a supervised classification. The parameters of membership functions are optimized by the PSO and cross validation (CV) procedures. The FCM classifier has following advantages. 1) Classification performance in terms of 10-fold CV and three-way data splits (3-WDS) is high. 2) Missing values can be estimated based on the least square Mahalanobis distances. 3) High-dimensional feature vectors can be classified taking into account the covariance structure of clusters. 4) The relational version can be used when many of the feature values are missing, or when only relational data are available instead of the object data.

View full abstract

A method to automatic extraction of brain tissue from CT images for content-based image retrieval is proposed based on fuzzy c-means clustering algorithm with spatial constraints. In order to reduce the computation time, the modes of the three major components of the CT head image, the background, the soft tissue and the skull, are estimated, which can be used as the initial clusters centroids in the following clustering stage. The characteristics of this method is that the extraction process is parameter-free and completely automatic, hence, it dose not need any user interactions. The proposed method has been validated on real CT head images and has shown satisfactory results.

View full abstract

This article is dealing with the fuzzy clustering method which combines the deterministic annealing approach with entropy, especially fuzzy entropy and Tsallis entropy. By maximizing fuzzy entropy or Tsallis entropy within the framework of the fuzzy c-means (FCM) method, membership functions similar to the statistical mechanical distribution functions are obtained. We examine the characteristics of these entropy and membership functions from the statistical mechanical point of view. After that, both fuzzy entropy based and Tsallis entropy based FCM are formulated as DA clustering. Numerical experiments are performed and the obtained results are compared and investigated.

View full abstract

An image retrieval system that retrieves images such as paintings and illustrations with Kansei words and image-content keywords using a visual user interface was proposed and implemented. Although Kansei image retrieval systems, which retrieve images with a query containing only Kansei words, have been investigated extensively, these systems have a number of disadvantages. For example, the user often wants to retrieve images using not only sensibility terms but also words that expresses the subject of the images. In addition, since the system requires the user to input query in his/her own Kansei words, it may be difficult for the user to express a request using the correct words. Moreover, the user interface displaying the retrieval results should allow the user to review the retrieved images in order to facilitate the determination of which images in the retrieval results satisfy the user's request. Therefore, a system with the following functions was proposed: The system takes a query as a pattern of Kansei words and image-content keywords, and the user can select the Kansei word from the map based on the Color Image Scale; The system shows the retrieval results in a two-dimensional plane allowing the user to know the fitness of the retrieved image at a glance; The system implements the user relevance feedback framework to adapt the retrieval to changes in the user requirements. The system was evaluated subjectively and was found to provide superior accuracy and usability compared to conventional systems.

View full abstract

This paper describes the pedestrian navigation method reflecting individual preference for route selection. The presented method selects the route with the highest subjective satisfaction degree which is estimated by a Road Satisfaction degree Evaluation Model(RSEM). The RSEM applies fuzzy measures and integrals to calculate the subjective satisfaction degrees of a road. The input to the RSEM is a set of road attributes expressing subjective impression of a road. The collaborative filtering method is applied to the acquisition of road impressions evaluated by other users is proposed. Experimental results imply that almost the same satisfaction degrees as the ones for routes selected using questionnaire data by subjects own are obtained even if the road impressions are acquired by estimation using other users' evaluation.

View full abstract

In this paper, we propose a action generation model for entertainment robots, whose action makes them appear to be living. One problem common to entertainment robots is that it is difficult to generate a variety of actions, because robot action is generally based on simple rule-based algorithms. Thus, the simplicity of the robot's actions quickly bores users. In this paper, we propose an action generation model based on secondary needs. The secondary need indicates the needs which have a close relation to primary needs in psychology. This model uses need parameters and a neural network for choosing action. Need parameters increase gradually, and then decrease through specific action. Because the neural network learns actions that decrease the need parameters, the model begins to behave spontaneously. Simulation of action generation shows that the model does accommodate secondary needs.

View full abstract

Many computer-based educational programs are designed with more consideration on their functionality, where as very little consideration is given to the aesthetic needs of the users. This on-going research attempts to explore the color preferences of children and discuss how a designer can choose preferable colors for interactive learning environments using Kansei Engineering techniques with the aim to provide visual aesthetics while increasing the usability.

View full abstract

This paper describes a human adaptive conceptual design system which designs the shape of a couple of Tableware. In the system, a framework of human adaptive approach is implemented based on the interaction between human and the system, by means of Interactive Evolutionary Computation. Users / beginner designers only carry out simple and subjective evaluation toward the automatically generated tableware shape by the system. The system re-generates the shape of tableware based on users' evaluation results, and presents users again. Repeating the interaction, users acquire satisfied design. The system uses rules of linguistic feature expression of the tableware's shape. Also, the system preserves rules, which are evaluated well by users, toward Database. Therefore, the users' own conceptual favorite shape features are preserved. These rules can be repeatedly applied in order to generate new kinds of tableware, which have the same concept.

View full abstract

This paper summarizes intriguing results obtained by our recently-developed stagewise backpropagation algorithm that evaluates the Hessian matrix of a given objective function explicitly in a block-arrow matrix form. Its computational organization facilitates the exploitation of layered structure embedded in a multi-stage neural-network model. Notably, in nonlinear least squares learning, our stagewise procedure evaluates the Hessian matrix of the squared-error function at the essentially same cost as the Gauss-Newton Hessian, faster than standard rankupdate methods; this computational convenience is immensely significant.

View full abstract

Two methods of optimizing feed-forward Artificial Neural Networks (ANN) through Genetic Algorithm (GA) are presented. The connection weights of ANN are extracted out. The functions formulated having connection weights as variables are optimized through GA. The optimized connection weights replace the old ANN connection weights. The first method proposed gives closer results. The second method is a novel approach of ANN optimization through GA. This method produced exact values. The method is proved by designing spindle of cylindrical grinding machine.

View full abstract

In this paper, we present a method of unsupervised spiking neural network. Recently, Spiking Neural Network has been much considered in attempt to achieve a more biological plausibility artificial neural network. In this paper, we suggest spiking neurons with delays to encode information. Thus each output node will produce a different timing which enables competitive learning. The suggested mechanism is design and analyses to perform self-organising learning and preserve the inputs topology. The mechanism is further assessed in real world data clustering. The simulation results show that the model behave similar to the conventional self-organising neural network.

View full abstract

This paper presents a new constructive algorithm (NCA) in designing artificial neural networks (ANNs). Unlike most previous studies on designing ANNs, NCA puts emphasis on architectural adaptation as well as functional adaptation in its design process. This algorithm uses a constructive approach to determine both the number of hidden layers in an ANN and of neurons in each hidden layer automatically. To achieve functional adaptation, NCA trains hidden neurons using different training sets created by employing a similar concept used in the boosting algorithm. Eight classification problems were used to evaluate the performance of the proposed approach. The experimental results show that NCA can produce compact ANNs with good generalization ability in comparison with other algorithms.

View full abstract

In most of conventional approaches used for pattern classification using unsupervised ANN either method of clusterification is used or the entire feature set is used. Redundancy in such cases makes the dimensionality of the feature space too complex to handle. Also early convergence is another desired factor for training phase of such networks for which different architectures or learning algorithms are to be tried. A rough set is one such tool of approximation, which works well when there is lots of inconsistency in data or even missing data is there. Hence approaches using rough sets can be used at preprocessing level, learning level or neuron architectural level. Thus this paper discusses preprocessing and architectural level approaches using Rough sets for overall performance improvement of such pattern classifier for case of character recognition.

View full abstract

Recently, system verification plays an important role in Design of large scale and complex systems, embedded systems, and other critical systems. However, it is very difficult for designers other than the specialist who is well versed in Temporal Logic to specify behaviors of the system. In this paper, we consider where designers of systems can extract check-points, necessary signal events, in model checking of formal verification. Moreover, we demonstrate some specification examples, and some verification results by NuSMV model checking tools.

View full abstract

Due to multichannel broadcasting, much contents consecutive in means of time are being offered. Much of these contents require real-time characteristics. In existing methods for viewing two simultaneous TV programs, there are several issues. For channel switching and displaying two programs on one display simultaneously, since the user misses parts of the program, the comprehension level decreases. In this paper we focus on raising the comprehension of two simultaneous TV programs. We therefore propose a crossover viewing support method using time compressed playout. In this method two TV programs are first buffered in real-time and the buffered contents are alternately played out in a time compressed fashion. Since the TV program may be switched in the middle of a statement, comprehension may become challenging. Overlapped restart playout is therefore introduced to alternate playout. This makes semi-real-time viewing of two simultaneous TV programs possible. Through evaluation experiments we showed that the proposed method raises the comprehension level compared to existing methods.

View full abstract

A fuzzy trend model is a time series model based on fuzzy if-then rules. In this paper we propose a vector fuzzy trend model for nonstationary time series. The model provides a decomposition of each trends into the common trend and individual trends. The applicability of the model is show by a practical analysis.

View full abstract

Aging society is approaching rapidly. Cancer, heart diseases and apoplexy became top three murders of people in Japan. These diseases are named lifestyle-related diseases. Research has shown that lifestyle-related diseases are refractory under medical treatment but preventable by removing bad lifestyles. Therefore, a home doctor system for lifestyle-related diseases based on Distance-Type Fuzzy Reasoning Method is in developing. In previous research, current health condition could be checked by the system, health Prediction algorithm according to current health condition, lifestyles and heredities has been discussed. In this paper, besides the impact of lifestyles and heredities, adverse health effect of lifestyles duration is also considered. Firstly, effect of lifestyle duration is analyzed. Secondly, an improved prediction algorithm is proposed. Lastly, the availability of proposed algorithm and developed system is discussed according to experimental data.

View full abstract

Motion control of mobile rehabilitation robot during centre-of-gravity shift is an intrinsic problem with mechanism. This paper deals with the motion control of an omnidirectional lower limbs rehabilitation robot during centre-of-gravity shift. The main research contents of this paper consist of two parts: 1) The dynamic model that considered centre-of-gravity position is integrated within the control framework. 2) A robust control strategy is developed and used to achieve the desired trajectories. Simulation tests are performed and demonstrate the feasibility and efficacy of the proposed method.

View full abstract

Home robot is highly expected to help our everyday life for the coming aged society. In order to realize the success applications of home robots, it is of most important to develop effective human-robot interface, which can not only transform what human asked via speech as well as body motions but also catch what human wanted by measuring the brain's biological activities directly. In this research, we propose to apply NIRS (Near-Infrared Spectroscopy), a new type of brain measurement setup that can measure the brain activities during the subject's free body motions, as an interface to control the motion of a human-interactive robot RI-MAN. In detail, we first studied the time response from the brain activities in primary motor areas to the reaction as seen from NIRS. We then developed a BRI interface for the robot RI-MAN, in which we considered the 7 basic locomotion patterns of the robot including STOPPING and classified the brain measurements so as to control robot's motion directly.

View full abstract

In this study, multichannel near-infrared spectroscopy (NIRS) was used to measure the difference between cerebral hemoglobin concentration during a complete-letter recognition task and that during an incomplete-letter recognition task. In the incomplete-letter recognition task, 90% of the black pixels in the letter images were erased randomly. 13 subjects participated in the experiment. The results demonstrated that the relative oxygenated hemoglobin concentration (oxyHb) during the incomplete-letter recognition task was higher than that during the complete-letter recognition task. Furthermore, the oxy-Hb difference between the two sections in the bilateral frontal areas was more significant than that in the occipital areas, indicating that the frontal cortex plays a more important role in recognizing incomplete objects than the visual cortex.

View full abstract

Gene expression profiles could be a powerful tool for therapeutic decisions and prognostic estimation of disease. It is useful but transforming such huge data into a mechanistic understanding of disease is very difficult. Recently, researches aimed at uncorvering the modular organization and function of transcriptional networks and responses in disease. In this study, such modules[1] are extracted from the gene expression profiles as statistically independent components(ICs) obtained by independent component analysis(ICA). Therefore, a probit model for classificaion of disease is constructed by the ICs, and the model parameters are estimated by Markov Chain Monte Carlo(MCMC) method. The present method is applied to gene expression profiles of glioblastoma and anaplastic oligodendroglioma. As a result, we obtain a probit model which can classify the two kinds of gliomas. In addition, biological functions are extracted as the ICs.

View full abstract

Binocular disparity is one of the most important cues for depth perception of human. Previously, some of the present authors have measured electroencephalogram (EEG) from subjects who were watching random dot stereograms (RDSs) with three types of binocular disparities: no or small or large disparity, and estimated brain areas where the visual information is processed, by use of equivalent current dipole source localization (ECDL) method. The results showed that: 1) the postcentral gyrus (PstCG) is involved in visual processing of stereopsis; and 2) all-channels of averaged EEGs for all subjects had converged and the convergence time for the large RDS disparity is longer than that for the small one[1]. Application of the ECDL method to averaged data for small and large disparities showed that visual processing route before the PstCG consists of two pathways: one is from V1 to V4 and then to the TE field; and the other is from V1 to the MT field and then to the PstCG. This result did not depend on the types of disparity. After the PstCG, ECDs were localized to the superior colliculus (SC) and the frontal visual field (FEF), both of which are involved in ocular movements. At interval between the FEF localization and the EEG convergence, ECDs were localized to the inferior frontal gyrus (IFG) and the middle frontal gyrus (MFG). For RDSs with large disparity, the IFG and MFG ECDs were estimated earlier than those for small disparity, while for large disparity the convergence time and the time when ECDs were localized to the IFG just before the convergence time were later than those for small disparity.

View full abstract

The aperture problem is a motion perception through a small window involved ambiguity both in a speed and a direction of the motion. It has been analyzed in various experimental approaches with knowledge about the early motion process in visual pathways of the human brain. Some of the present authors had investigated correct answer rate of the moving orientation and analyzed brain activities with different experimental parameters, such as line speed, radius of the apertures, and length of a line. The present authors recorded electroencephalograms (EEGs) from a subject and estimated their sources and latencies in the brain using the equivalent current dipole source localization (ECDL) method. We compared localized ECDs for two different line speeds (Type 1: 10msec/pixel and Type 2: 20msec/pixel). At the latency of the appearance of aperture, ECDs were localized along the ventral pathway concern with the recognition of form. After appearance of the line, ECDs were localized along the dorsal pathway concern with the recognition of movement. In addition, after appearance of another aperture, ECDs were localized to the middle frontal gyrus and the inferior frontal gyrus.

View full abstract

Aperture problem is a psychological experiment for analyzing binding mechanism of the spatial recognition in an early stage of visual pathway. In this paper, we measure perceptual rate in the aperture experiments, and discuss the dependency between the perception and various parameters in the experiments. We also record Electroencephalograms(EEG) of subjects who are recognizing the perception. By the electroencephalograms(EEG) analysis, we measure reaction latency of visual evoked potential (VEP) and event related potential(ERP) related to visual pathway, and estimate the localized equivalent current dipole(ECD) in the visual pathway.

View full abstract

This study performed psychophysical experiments addressing how we perceive global closed contours whose polarity of border-ownership(i.e., figural side) are pre-assigned by using small rectangles which has the nature of figure. When ownership side was assigned toward inner side of the global closed contour, saliency of the contour was high; on the other hand when ownership was toward outer side, saliency was low. This difference may reflect the existence of interaction between global and local ownership signals.

View full abstract

We propose an application of socially intelligent agents (SIAs) for enhancing e-Learning. SIAs are software agents which manifest social intelligence. Although there are many types of social intelligence, we focus on social intelligence for providing emotional support. By applying this type of social intelligence, our SIAs are able to simulate human social behavior. We developed a software platform for applications of SIAs. We demonstrate how we created the SIAs and describe the conceptual architecture of our software platform. We also show the mechanism for generating the agents' social behavior, which makes our software platform unique. According to the media equation theory, we expect SIAs to benefit users. To examine the effects of SIAs, we designed and implemented an applications for e-Learning. We provide explanation about how SIAs function in this application and discuss further implications of social intelligence.

View full abstract

We have developed a method of syllabi systematization to help course developers design curricula. A good curriculum is crucial for a successful university education. When developing a curriculum, topics, such as economics, natural science, and informatics, are set first, and then course syllabi are written accordingly. However, there is no guarantee that the topics actually covered by the course syllabi are identical to the initially set topics. To find out if the actual topics covered by the developed course syllabi, we developed a method of systematically analyzing course syllabi that uses latent Dirichlet allocation (LDA), which was proposed by Steyvers and Griffiths. Latent Dirichlet allocation is a method used to extract latent topics based on a generative probabilistic model of documents, which in this context are syllabi. Using the distance between topics, we can map the topics and syllabi on a two-dimensional space using an isomap. We also developed a method to relate other syllabi to given topics. We applied these methods to the syllabi of the University of Tokyo and the Open University of Japan and found them to be promising methods of systematizing syllabi.

View full abstract

Information and communication technology is progressing in the activities of higher education institutions. We have so far collected and accumulated document information mainly including syllabi concerning curricula of higher education. Analyzing the structure of accumulated information, general data structure of curricula was examined, and database has been developed on relational database with XML format. Moreover, comparative analysis of curricula by utilizing the database has been considered. As a development of our former research, this article presents an investigation on the document structure of syllabi of advanced engineering programs of colleges of technology in Japan.

View full abstract

Correspondence analysis is frequently utilized in text mining. It will deepen the global understanding on the characteristics of accumulated text information and may lead to new knowledge discovery. However, in the case of data fluctuation, result of correspondence analysis might be greatly changed. This article presents a mathematical consideration on sensitivity of correspondence analysis and its application to comparative analysis of curricula of higher education institutions.

View full abstract

Control of sluice gates of dams is a complex, nonlinear, non-stationary control process and is significantly affected by hydrological conditions which are not predictable beforehand. In this paper, control methods based on dynamic fuzzy logic are proposed for the operation of sluice gates of dams during floods. The proposed methods are not only suitable for controlling gates but also able to maintain target water level in order to prepare a draught. In the proposed methods, we use dynamic fuzzy logic that the membership functions can be varied by changing environment conditions for keeping up the target water level, instead of conventional static fuzzy logic. Simulation results demonstrate that the proposed methods based on dynamic fuzzy logic produce an accurate and efficient solution for both of controlling sluice gates and maintaining target water level defined beforehand.

View full abstract

In a metal grating manufacturing industry, the layout design of gratings is the most important activity of the whole design process since it determines a large portion of manufacturing cost. However, due to the complexity of the problem in layout design phase, it is impossible to generate design alternatives and select the best design solution within a reasonable time period. In this paper, we apply genetic algorithms to search a near-optimal solution of the layout design problem, which is focused on the minimization of machining cost. We also employ heuristics to generate sub-design candidates for subproblems. The effectiveness of the proposed method is shown by experimental results.

View full abstract

In this paper, the performance improvement of a magnetic position sensor for autonomous vehicle using 1-axis magnetic field sensor array is described. In the data of magnetic marker's position versus its magnetic fields, there is a region which is reasonably linear. The proposed system uses the region to detect marker's position from magnetic field. The proposed magnetic position sensing is verified by the experimental results.

View full abstract

This paper discusses a fuzzy support agent using ontology-based linguistic instructions for ship steering control and desires to testify the validity of the proposal by applying the fuzzy control model to the steering control system based on linguistic instructions. The fuzzy support agent is presented to build the maneuvering models of steersman and the miniature model for steering control system. The proposed system contains three main mechanisms, including the interpretation mechanism of linguistic instructions, the self-regulation mechanism, and the task performance mechanism. Among three, the task performance mechanism includes the kinematics module and the performance ontology. The simulation results show that the proposed approach can work effectively for ship steering control.

View full abstract

Automated collaborative filtering is a computational realization of ``word-of-mouth'' in network community and the applicability of each item for users is predicted based on missing values estimation in a matrix of users versus items. The original memory-based system of GroupLens uses the weighted averages of ratings given by the ``neighbors'' considering similarities to the active user. A similar idea was applied to the model-based system based on linear fuzzy clustering, in which missing values are predicted considering local substructures. This paper considers combining the numerical evaluation matrix with other categorical information and proposes a collaborative filtering system based on linear fuzzy clustering with nominal variable quantification. Numerical experiments demonstrate that categorical information is useful for improving the performance of the model-based prediction model.

View full abstract

PCA-guided k-Means derives a continuous (relaxed) solution of the cluster membership indicators in k-Means. A robust k-Means model can be formulated based on a fuzzy PCA-guided clustering procedure where a responsibility weight of each sample in k-Means process is estimated based on the noise fuzzy clustering mechanism, and cluster membership indicators in k-Means process are derived as fuzzy principal components considering the responsibility weights in fuzzy PCA. In this paper, kernel method is applied to the fuzzy PCA-guided k-Means in order to extract a larger number of clusters than the dimensionality of a data set. Considering mapping to a high dimensional space, we estimate meaningful principal components that are used for capturing nonlinear classification boundaries. Cluster structures can be visually assessed by the spectral ordering, in which samples are arranged based on mutual connectivity weights, and are emphasized in the diagonal block structure in the connectivity matrix.

View full abstract

Linear fuzzy clustering is a local PCA technique, in which the Fuzzy c-Means (FCM)-like iterative procedure is performed by using linear varieties as the prototypes of clusters. In Fuzzy c-Medoids (FCMdd), cluster prototypes are selected from data samples and clustering criteria are calculated by using only mutual distances among samples. Then, it can be easily applied to clustering of relational data. This paper proposes an extended FCMdd approach for linear fuzzy clustering of relational data, which is useful for extracting plane-like sub-structures spanned by three representative objects (medoids). In the algorithm, new prototype is given by solving a combinatorial optimization problem for searching medoids and the computational complexity is reduced by searching only from a subset of objects having large membership values. The information summarization approach can be regarded as a multi-cluster-type multi-dimensional scaling for summarizing data into several 2-D planes.

View full abstract

Fuzzy c-Regression Models (FCRM) is a fuzzy clustering-based switching regression method and can be used for pattern classification. In this paper, a hybrid approach to FCRM and category partitioning is applied to physiological measurement data for brain activity analysis, which is a physiological measurement approach to qualitative modeling of Kansei information based on measurement of brain functions. A task prediction problem is formulated by using time series of oxidized hemoglobin measured by Near-infrared spectroscopy (NIRS), and FCRM with partitioning of categories is used for revealing mutual relation between a particular task and external factors represented by categorical observations. In the modified FCRM algorithm, switching prediction model is constructed considering quantification and partition of categories with the goal being to classify external factors based on task dependencies.

View full abstract

Four algorithms for sequential clustering using fuzzy cluster memberships are developed and their effectiveness and efficiency are compared. First is a known technique of the mountain clustering, and the second is a new method of mountain medoid clustering which uses individuals instead of grid points in the mountain clustering. The third is an algorithm using possibilistic clustering which the authors have developed, and the fourth is an sequential algorithm using the noise clustering with one cluster. In these algorithms, the number of clusters need not be specified beforehand. These four algorithms are described; their performances are discussed and tested using different numerical examples.

View full abstract

The idea of "Deep KANSEI", which represents admiration and healing subconsciously induced by contents of a high artistic quality, is quite important for better understanding of the whole picture of human feeling or KANSEI. As a possible device producing Deep KANSEI, we have been developed a remarkable sound system, named Extra HI System M (EHISM). Our previous psychophysical studies showed that EHISM could affect a human psychological state. However, it has not yet known if such a change in psychological, thus verbalized, state brought by EHISM were accompanied with a change in physiological states. In this study, we investigated the relationship between the psychological and physiological stress responses induced by EHISM. With regard to the psychological stress scale, the term "tension" and "relaxing" were interviewed before and after the sound stimulus by EHISM or other conventional sound system (control). With regard to the physiological stress scale, 3 kinds of possible stress biomarkers, i.e., cortisol, immunoglobulin A (IgA), and chromogranin A (CgA), were assayed by saliva taken 3 min before and after the sound stimuli. The results suggested the close relationship between these psychological and physiological scales. Finally, we discussed our results in relation to the possibility of objective assessment scale of Deep KANSEI.

View full abstract

This study aims at an application of rough set theory to illustrate the relationship between human psychological and physiological states. Recent behavioral medicine studies have revealed that various human secretory substances change according to mental states. These substances, the hormones and immune substances, show temporal increase against mental stress. Thus, it is frequently introduced as biomarkers of mental stress. The relationship between these biomarkers and human chronic stress or daily mental states was also suggested in the previous studies. However the results of these studies were inconsistent. Some technical reasons were indicated for this discrepancy. Among that, we focused on the analysis technique investigating the relationship between human psychological state, i.e., scores of a psychological scale, and physiological state, i.e., level of the secretory biomarkers. In this paper, we introduced Rough Set analysis method instead of using a conventional linear correlation analysis method. In the experiment, the salivary secretory immunoglobulin A (IgA), which is a major stress biomarker, of 20 male students was assessed before and after a short-term stressful mental workload. Also, 65 items of psychological mood scale was assessed as a psychological index. The result showed that some items strongly related with the change in the IgA, while no significant linear correlation among that was obtained.

View full abstract

Complicating the daily life style, mental tiredness and the physical fatigue will be trigger factors of human errors and relating to the accident. It has been getting important that a computer knows user's fatigue. Fatigue is the subjective quantity which only the said person can recognize. Therefore, to know the user's fatigue it is necessary to calculate the probability of the causal relation from physiological information such as facial and voice changes observing continuously. In our prior research, we constructed the models of fatigue based on the Bayesian network using the psychological and physiological data and proposed convergent inference system framework using multiple Bayesian networks based on the ontology and agent technology. In this work, the learning mechanism of a new causal relation was tried to construct from the observation of the disagreement of the expectation by non-continuous process by supervised learning through the question on the reason.

View full abstract

People can separate and comprehend voices from a mixed signal of simultaneously spoken voices. This kind of ability was defined as voice separation ability, and it is considered to relate to brain cognitive function. To develop an easy-used check and recovery system of age-related cognitive deficit, a quantitative measurement method of voice separation ability was proposed in this paper. Correspondingly, the measurement experiments on subjects in different ages were done. The measurement result showed that the voice separation ability obviously declines with aging, and the proposed method was proved feasible to evaluate brain cognitive function to a certain extent. Furthermore, through brain function mapping analysis by near-infrared spectroscopy (NIRS) system, the brain cortex areas associated with voice separation ability was primarily determined. The activated cortex areas showed that voice separation ability correlates to brain cognitive function, and the proposed method was further proved valid. Consequently, this quantitative measurement method is considered possible to be applied to early-detection and early-recovery of cognitive deficits in future work.

View full abstract

Interactive evolutionary computation (EC) is an effective method for reflecting user's favor to creation of media contents. In this research area, user's fatigue by selecting or evaluating contents remains as a problem. Takagi et al. proposed a framework of novel EC technique which uses user's physiological indices as evaluation value instead of subjective evaluation. In our previous study, we constructed EC system that creates musical melody based on evaluation of listener's heartbeat information, however, the system provided only simple melody. In this study, as an advanced system, we constructed an EC system that creates musical piece with code development which was used in Canon by Pachelbel. In the constructed system, user's heartbeat intervals during listening musical pieces were used for evaluation value in Genetic Algorithm. To investigate the effectiveness of the constructed system, subjects participated to two listening experiments; roulette selection and elitism strategy vs. random selection. Furthermore, some musical melodies were shown and compared.

View full abstract

RatCar system, a vehicle-formed brain-machine interface for a rat, has been applied to analyze bidirectional adaptation in brain and machine under direct neural connections. A rat with neural electrodes implanted in its motor cortex and basal ganglia regions was mounted on the vehicle which were designed to move around by estimating intention of the rat. Recorded neural activities, however, had been restricted to those generated nearby our electrodes, which had resulted in low accuracy of the estimation and instable control of the vehicle. In this paper, another control strategy were introduced to the system; a predefined model determined a basic operation of the vehicle (e.g., circular locomotion) while neural activities modified its global behavior. A state space representation composed the model solved by Kalman filter algorithm. This framework enabled adaptation of vehicle control mathematically dissected from adaptation in the brain. As a result, more stable and practical control of the vehicle besides observing time-varying parameters during the adaptation.

View full abstract

Microelectrode arrays are commonly used to measure neural activities in the brain, and arrays with some 100 electrodes are commercially available to date. However, insertion of dense grid array deforms the brain, resulting in deterioration of the measurements. In order to overcome this problem, we propose a piezo-driven vibrating insertion device to reduce the inseritoninduced deformation of the brain. We attempted under various conditions of the insertion of the array into an agarose substrate, whose hardness was adjusted to that of the cerebral cortex of rats. Our experiments demonstrated that inverse-sawtooth vibration reduced the insertion-induced deformation of the substrate in proportion to the logarithm of an upstroke velocity when the velocity was higher than 10 mm/s, and vibrating insertion of the maximum velocity at 36.7 mm/s reduced the deformation by up to 40% as compared to vibration-free insertion. In addition, we tested the vibrating insertion device in an electrophysiological experiment in the rat auditory cortex in vivo, and successfully measured toneevoked neuronal activities, suggesting that the vibration during the insertion did not cause fatal damage to the brain.

View full abstract

In order to understand the function of neuronal circuits, the spatio-temporal activity of multiple neurons have to be measured. In this regards, imaging of neuronal activity using fluorescence dyes is one of the most promising techniques. We conducted imaging studies on nervous tissues of in vitro and in vivo preparations using several different fluorescence dyes. Ca²⁺ is an important messenger in signal transduction of neurons and the intracellular Ca²⁺ concentration ([Ca²⁺]_i) is known to increase during the cell excitation. We made the Ca²⁺ imaging study in the brain slice preparations. [Ca²⁺]_i was measured using a high-speed cooled-CCD imaging system equipped with a excitation wavelength changer. In the basal ganglia striatal slices from mice, we observed the spontaneous [Ca²⁺]_i changes from individual neurons and glial cells. Long lasting spontaneous [Ca²⁺]_i changes, which lasted up to about 100 s, were found in both neurons and glial cells. In the visual cortical slice preparation, we measured the [Ca2+]_i changes of the neuronal population evoked by electrical stimulations. We could measure the signal propagation in the neuronal network of the visual cortex, and study the functional neuronal connections in the visual cortex. The membrane potential changes of neuronal population can be imaged with a voltage sensitive dye. In vivo membrane potential imaging reveal how the visual signals are encoded in the visual cortex and how signals propagate in the intact visual cortex. A flash of light applied to contralateral eye induced focal activation in the primary visual cortex in accord with the retinotopic map. In summary, imaging can visualize the population of the neuronal activity and have great potentials to reveal the spatiotemporal properties of the functional network from in vitro to in vivo.

View full abstract

Rat hippocampal neurons reorganized a complex networks on microelectrodes array dish. The living neuronal network can distinguish patterns of action potentials evoked by different inputs, suggesting that a cultured neuronal network can represent particular states as symbols. A neuro-robot-hybrid system with living neuronal network and miniature moving robot was developed. We use a Khepera II robot for interfacing with a living neuronal network and the outer world and succeeded in performing collision avoidance behavior with premised control rule sets. Using self-tuning fuzzy reasoning, we associated a distinct spatial pattern of electrical activity with a particular phenomenon in the outside of the culture dish. The particular relationship between network activity and outer phenomenon was performed by control rules of electrical stimulation to the neuronal network, responding to outer phenomenon, while a spatio -temporal patterns of neuronal activity were linked to output devices by premised control rules. We succeeded in performing collision avoidance, and found that fluctuation of neuronal responses evoked by sensor output of robot body was controlled by the interaction between neurons via synapses.

View full abstract