In image processing, integer-order differentiation is often used to detect edges, maxima, minima, and points of inflection, especially order 1 used by the gradient and order 2 by the Laplacian. In the present study, non-integer (fractional) differentiation is applied to images which are processed in Fourier space, and the characteristics of differentiation filter are presented.

View full abstract

It is known that random optical intensities with fractal properties (fractal speckles) can be generated by illuminating an ordinary plane diffuser with coherent light having an average intensity obeying a negative power law. So far, such an illumination condition was realized by the laser light scattering from a random fractal aperture, giving rise to extremely weak intensities. In this paper, we introduce a new approach to generating power-law intensities on the basis of a liquid crystal spatial light modulator. This method provides us with much brighter fractal speckles than those due to the preceding method. Some properties of fractal speckles are also discussed, particluarly, in relation with the intensity clustering and multifractality.

View full abstract

The paper deals with handwritten numerical character recognition by use of a 3-way categorical data analysis with CCD cameras for control of an autonomous robot Khepera. We installed two CCD cameras on Khepera as human binocular eye. By the CCD cameras, a character was extracted from the wall, and the distance to the wall could be calculated. When a character was recognized, Khepera moved corresponded with a meaning of the character. The distance was determined for calibrating CCD measurement. To judge an accurate depth to the object on the base of the image from two CCD cameras, physical distances to some objects in different depths were measured in advance. Then we calculated differences in horizontal position from the image of right and left CCD cameras according to each object. In the case when the Khepera could detect few clues of depth, the marking point on the wall could facilitate the compensation of the difference of right and left images, and then it could judge the depth. Further, with handwritten character recognition system, we could control the Khepera with instructions on the wall.

View full abstract

Recently, simple and intuitive operation interface to robot becomes important, because robot use has been extended from industrial manufacturing to personal services. We developed hand gesture detecting device by sensing gesture acceleration and operation instruction system utilizing infrared remote controller. Hand gesture device has two directional acceleration sensor. Magnitude and direction of hand gesture dynamic (over 1G) and static (under 1G) acceleration can be discriminated by micro computer and mobile robot moves forward or backward or rotates following 8 kinds of hand gesture. 8 infrared ray receivers covered by 3.5cm long cylinder that gives directivity are attached around mobile robot. They can recognize remote controller key that is first pushed and the relative direction between robot and operator. Following key operation, mobile robot moves forward or backward or rotates. Further it can follow the operator by utilizing on-board CCD color camera. Experiments show improvement of interaction between human and robot.

View full abstract

This paper presents a method to calculate the acoustic characteristics of three-dimensional vocal-tract models. A cascaded structure of rectangular acoustic tubes, connected asymmetrically with respect to their axes, is introduced as an approximation of the real vocal-tract geometry. Computation of the acoustic field in the models are performed using higher-order modes. A small change of the vocal-tract shape is regarded as a geometrical perturbation of the axis position of each vocal-tract section. The computational results of sound pressure distributions and transfer characteristics for a large number of vocal-tract models are also presented. The results indicate that acoustic characteristics in the higher frequencies are highly sensitive to the small change of the vocal-tract shape.

View full abstract

In this paper, the transfer functions and the active sound intensities of a vocal tract model with and without a nasal cavity were computed by a three-dimensional FEM. The models were based on vowel MRI data of the vocal tract with a nasal cavity during phonation of the Japanese /a/. An oral cavity were also coupled with the nasal cavity in a three-dimensional volume of radiation. The effects of wall impedance were also examined. The coupling of the nasal cavity to the oral cavity indicated the following aspects: The additional peaks appeared below 3 kHz for the lossless condition. However, they disappeared in the simulation for the soft wall condition. The sound energy circulation did not occur in the simulation for the soft wall condition. As for the effects of the wall boundary condition on the spectral envelope, the upward shift of lower formant frequencies were confirmed based on the three-dimensional simulations. However the disagreements of the formant frequencies between the simulation and real speech should be further investigated by adjusting the wall boundary condition to a more realistic one.

View full abstract

Inalienable possession is a specific relation between two objects, the possessor and the object possessed(the possessee), for example a human body and its part, head. Inalienable possessive object is the possessee in the inalienable possession. In normal situations, inalienable possessive objects cannot be separated from their possessors. In this report, mainly about Japanese and the Ainu language, and from a point of view of the contrastive study of world languages, we consider the fundamental categories of inalienable possessive objects and patterns of expressions for inalienable possession in texts. Patterns of expressions for the possessors and the possessees in some languages are compared and the results are listed. We also consider some factors to control the generation of the expressions.

View full abstract

In this paper, new learning method have been proposed to acquire translation knowledge from corpora. The learning method automatically acquires rules, which are immanent in bilingual sentences, as translation knowledge. That is, the system acquires translation knowledge only from corpora without any prior preparation of a resource (e.g., a bilingual dictionary, a machine translation system). The learning method was applied to a machine translation system and a system for automatic extraction of bilingual word pairs. Evaluation experiments indicated the effectiveness of the learning method.

View full abstract

Objects detection algorithm for real surveillance system is proposed based on color exception instances for color dynamic images under low illumination, where the proposed color exception instances consist of color-difference and moving possibility region calculated by detected results in previous frame. It provides useful detection result for too low illuminated situation. Experimental results for dynamic image taken under low illumination in streets show that detected frames with the proposed algorithm increase by 20% compared with detection result without exception instances. The proposed algorithm is under consideration for use in a relatively poor security downtown area in Japan.

View full abstract

Mentality estimation system on affinity pleasure-arousal space using fuzzy inference is proposed for the communication between human beings and an eye robot agent. Mentality status on the proposed space is calculated by fuzzy inference which uses language category as outputs of speech understanding. The constructed eye robot system expresses mentality on the space by eyelid and ocular moving which is defined in advance. Experiments of mentality expression with two scenarios are done. Since the result of questionnaire shows that evaluation value is 3.6 out of 5.0, the proposed system is suitable for communication architecture between interlocutor and robot. The system provides the information terminal robot that becomes common in home environment.

View full abstract

One of applications of fuzzy measures is a subjective evaluation model. It is introduced by the Choquet integral with respect to a fuzzy measure, which is viewed as the weights of criteria. The bootstrapped Choquet integral model has been proposed and shown that it expresses the subjective nature of evaluation more suitably than the additive model and the ordinary Choquet integral model by an assessment test and numerical simulations. In order to use the bootstrapped model the fuzzy measure should be identified. This paper examines a way to identify the fuzzy measure interactively.

View full abstract

An automatic generating system of torus type impossible figures that is a subclass of multibar type impossible figures are proposed based on the shapes and positions of corners. It is shown that some impossible figures are determined only by a number of bars and a type of a link stracture, and other impossible figures are generated by knots in the knot table with a small number of bars that is achieved by the presented inequality. The proposed system aims a basic tool for experiments on visual psychology and further development of intelligent vision systems.

View full abstract

The orthogonal transform is mainly used in the video compression methods in order to concentrate energy on a part of coefficients without increasing the number of the coefficients after transform. It is the main technique of the image compression to preserve the coefficients of transform with large energy. Since the orthogonal transform has smaller number of bases than other transforms doesn't have the shift invariance, and as the result of it, the contour disappears after compression. On the other hand, the contour can be expressed well by the use of the redundant transform with larger number of bases. However, it has been thought that this transform which increases the number of coefficients is unsuitable for compression. Recently, Iterative Projection-Based Noise Shaping that reduces the coefficients by using projection has been developed and proposed as an effective technique to compress the image. The authors will discuss the effectiveness that uses redundant transform for the video compression.

View full abstract

In robot vision, it is important to recognize moving objects in dynamic image, and omni-directional camera is effective for this purpose. Omni-directional camera has an advantage that more visual information can be obtained at the same time than human. In this work we propose an approach of tracking multiple moving objects (human etc.) in dynamic image of omni-directional camera. The dynamics of position and velocity of target object and measurement process of the object by omni-directional camera are represented by a state space model based on finite random set. In general, a scene may have occlusion and appearance of objects, and the multiple moving objects may be observed with missing and false detections. Finite random set is suitable for representing these situations. State estimation is performed by Sequential Monte Carlo (SMC) implementation of Probability Hypothesis Density (PHD) filter. The multiple moving objects are simultaneously tracked by the filter. Numerical simulation and real image experiment results show tracking performance of proposed method.

View full abstract

The restoration of the video signals corrupted by the additive noise is important for getting high quality image and high compression ratio of video signals. In this paper, we propose a novel restoration method for the video signals corrupted by the only Gaussian noise. The proposed method makes choice into four filters (i.e., spatio-temporal filter, spatial filter, temporal filter, identity filter) depend on the local information. Thus, the proposed filter can reduce the additive noise while preserving the signal edge/detail and motion.

View full abstract

In order to visualize the whole structure of concept lattice obtained by formal concept analysis with respect to the image/video databases, a hierarchical representation method based on fuzzy clustering (especially, FCM) is proposed. Experiments using real image/video databases ('Corel Gallery 1000a' (1,000color images) and a video selected from standard motion database (150 frames)) are performed to confirm the eectiveness of the proposed method. Through the investigation of the concept lattice obtained by the proposed method, it is confirmed that the proposed method is helpful to do video clipping and grasp the whole structure of image database.

View full abstract

In web advertisement, commercial messages are automatically posted on the monitor of a user. If they meet with the user's interest, not only the user can get to the necessary information immediately but also the effectiveness of the advertisement will increase. The authors have developed a dynamic delivery system in web advertisement. Any prior registration of the user's preference is not necessary. The system we propose here not only grasps the change of user's preference in the long term, but also predicts the user's current preference in the short time. A Markov model and kMER are used to analyze the user's site transfer history. Application to the actual web advertisement system is reported.

View full abstract

Our laboratory deals with an obstacle recognition method using Self Organizing Map(SOM) for wheelchair type autonomous mobile robot DREAM-3 which is served to the care for elderly people or a physically handicapped person. In this obstacle detection method using the SOM, the operation for the image processing is not suitable in its computational time for a practical use. In this paper, a coding method of image data is proposed and advantageous features have been proved by the numerical computation.

View full abstract

We consider to the 3-dimensional topological Self-Organizing Maps and more higher order Maps origin from the 1-dimensional topological network (line) and 2-dimensional topological network (plane) which are well used from the former. Since a compression ratio becomes low by 3-dimensional SOM as compared with 2-dimensional SOM, it may be able to leave more information on the original data. Research of 3-dimensional SOM is not popular. The main reasons are 1) The topology of a unit is various and 2) The general-purpose topology which can be used in common with many data sets is not established. The data visualization technique of having been suitable for 3-dimensional SOM is proposed in this paper.

View full abstract

Recently, a lot of social systems have been computerized using Information Technology (IT). It is no exception in hospitals. Hospital support systems are also rapidly being computerized using such as Diagnosis Reservation System, Ordering System for Drug Prescriptions, Electronic Medical Record (EMR) and so on. In the immediate future, all of medical data will be stored to the database server in the hospital, and it is considered that complete paper-less and film-less system in hospitals become reality. These stored data, however, is not used effectively for analysis because of the great deal of the data and its variety --- image data, numerical data, text data and so on. Previously, knowledge discovery methods from numerical medical database were studied and discussed actively. Text mining method for medical data, however, has not been enough discussed. This study discusses knowledge discovery method from Electronic Medical Records using Self Organizing Map (SOM) and its possibility. As the first step of this research, this paper describes the outline of report mapping method using SOM from Incident Reports. This paper also discusses the keyword extraction method from EMR and the coding method of SOM.

View full abstract

In this paper, a learning algorithm of a self-organizing map in which an input space is represented as graph is proposed. In the conventional SOMs, the input space is defined by Euclidean space. Recently a new method for enhancing the input space to the function space was proposed, however, these methods have been limited to a countinous input space. The expansion of a further application can be expected by applying the self-organizing maps to graph such as route maps and the electric circuit. The distance of the input node and the reference element is defined by the distance in the graph, and the reference node is updated aling the shortest route on the graph. The proposed method is applied to the travelling salesman problem with obstacles.

View full abstract

In this paper, we proposed Evaluation based Topology Representing Network (E-TRN) is proposed to improve learning accuracy of self-organizing relationship (SOR) network. In case of that a dimension of inputoutput space is high, an approximation of a target system sometimes includes large error, because the topology of the network is limited by the topology of competitive layer, usually 1- or 2-D. TRN can extract a desired topology precisely by using an idea of Competitive Hebbian Learning rule (CHL). By hybridding SOR network and TRN, both of an evaluation based learning and a topology extraction can be realized.

View full abstract

A modular network SOM (mnSOM) proposed by authors is an extension and a generalization of a conventional SOM in which each nodal unit is replaced by a module such as a neural network. The mnSOM has the learning algorithm in which both the supervised and the unsupervised learning are combined. Thus, each module of the mnSOM implements the supervised learning, while, roles in the modules are allocated self-organizingly. In this paper, the architecture and the algorithm of the mnSOM are shown, moreover, the applications are presented.

View full abstract

Kohonen's self-organizin map (SOM) is an architecture that generates a map of a given dataset. In this paper, a novel extension of SOM called SOM2 is proposed. The mapping objects of SOM2 are SOMs themselves, each of which represents a set of data vectors. Thus, the entire SOM2 represents a set of data distributions. In terms of topology, SOM2 organizes a homotopy rather than a map in self-organizing manner. SOM2 is expected to be a powerful tool for the classification, estimation and recongnition tasks relevant to nonlinear manifolds.

View full abstract

Recently, the rapid progress of computers introduce evolutionary computations to next step, which is the demand for the variety of Pareto solutions in multi-objective optimization problems. We can calculate a large amount of Pareto solutions in a short time. However, it is difficult to use the acquired Pareto solutions effectivelly, because the Pareto solutions have multi-dimension of fitness values. This study tries to develop "Solution Mining" technique with visualization. This paper proposes a visualizing method for Pareto solutions which have multi-objective fitness values. The proposed method enables us to grasp the distributed structure of Pareto solutions and clarify the relationship among multi-objective fitness values. This paper shows that the visualized data enables us to interpret the characteristics of Pareto solutions through experimental result.

View full abstract

Recently, particle methods have been receiving attractive attentions in the filtering of nonlinear and non Gaussian systems. This paper is addressed a novel particle filtering method by combining interactive multiple model particle filter~(IMM PF) with evolution strategies~(ES). The analogies between particle filters and evolutionary computation~(EC) are considered then, ES is introduced to replace the resampling step in the IMM PF. The proposed filter is applied to simultaneous state and mode estimation problem of a kind of switching systems such as hybrid systems.

View full abstract

In this paper, we apply genetic rule selection to large-scale datasets. Since genetic algorithms require many iterations of the computation, it is difficult to apply genetic rule selection to large-scale datasets directly. In order to overcome this problem, we extend our genetic rule selection. First, we partition a dataset into several subsets. Second, at each generation, we externally store the nondominated solutions in terms of average fitness and survival times. Through computational experiments, we show that genetic rule selection not only improves their classification accuracy, but also significantly decreases the number of rules.

View full abstract

In this paper, we propose a method generating initial solutions in a three-objective vehicle routing problem with two demand periods. We also propose a local search algorithm to increase the similarity of a solution for a high demand period and one for a normal demand. We consider three objectives in vehicle routing problems: minimizing the number of vehicles, minimizing the maximum routing duration, and maximizing the similarity between the solutions in the normal and high demand periods. Through computer simulation, we show the effectiveness of the generating method to improve the objective values. We also show that the proposed local search improves the similarity of solutions.

View full abstract

This paper describes a distributed multi agent simulation (MAS) for financing model using Message Passing Interface. Recently, the social simulation with MAS is widely noticed social scientists, a more realistic simulation is necessary for them. So, we have been developing a social simulation tool for the Grid computing environment that enables a large-scale computing. In this paper, we propose a parallel algorithm for MAS with Message Passing Interface (MPI) on the Grid. The simulation results show that there is no significant gap between the experimental data obtained by the proposed algorithm and the traditional algorithm.

View full abstract

Interactive Evolutionary Computation (IEC) is one of the effective methods for optimization problems which are difficult to formulate the evaluation function such as human sensitivity. This method, however, gives a big burden to the user because he/she himself/herself has to evaluate the candidates of solutions a lot of times. This paper employs fitness inference method to reduce the burden for evaluation. When fitness inference method is applied to IEC, it can be a problem that the evaluation criterion for user has changed with the passage of time and/or impression given by the effects of other candidates. Because the fitness inference method infers the fitness values of candidates using the information of the actually evaluated solutions and their fitness values in previous generations. This paper proposes the fitness inference method with varying the number of actual evaluation for candidates based on the accuracy of inference for fitting the change of evaluation criterion of user. The proposed method can reduce the burden of evaluation following the user's evaluation criterion. This paper applies the proposed method to hearing aid adjustment support system using IEC and investigates the effectiveness of this method.

View full abstract

We focuses on emotion recognition using prosodic features in speech. There are some differences in prosodic features among indivisuals. Therefore, the accuracy of emotion recognition goes down. This paper proposes speaker adaptation technique for emotion recognition. It's an important technique that how to estimate the correct emotion of adaptation data for unsupervised speaker adaptation. The flow of conversation is usuful in estimating the emotion, because human emotion has a habit of expression continuity. So, we apply this habit to the method of emotion recognition in speech. Experimental results show the proposed technique can estimate the correct emotion of utterance data that unspecified emotional model can't estimate. Furthermore, the proposed technique shows the ability of emotion recognition technique for natural conversation.

View full abstract

This paper proposes modular neural networks for trajectory learning and a steady-state algorithm for trajectory generation used in the imitation of a partner robot interaction with a human. Various type of genetic algorithm have been applied for trajectory generation of robot manipulators. In this paper, we propose a trajectory motions pattern, and compare the proposed method with its related methods. Finally, we show experimental results of trajectory generation through interaction with human.

View full abstract

Collaborative filtering is a technique for reducing information overload, and personalized recommendation is performed by predicting missing values in a data matrix. Linear fuzzy clustering is a technique for local principal component analysis and can be used for estimating local prediction models considering data substructures. This paper proposes a new algorithm for estimating local linear models that performs a simultaneous application of fuzzy clustering and principal component analysis based on sequential subspace learning. In numerical experiments, the diagnostic power of the filtering system is shown to be improved by predicting missing values using the proposed local linear models.

View full abstract

Fuzzy c-Varieties (FCV) is a tool for linear fuzzy clustering and is also applicable to local principal component analysis (local PCA). While the clustering criterion in FCV is composed of distances between data points and prototypical linear varieties, the criterion can also be defined based on least square approximation. Optimal scaling is a useful approach to multivariate analysis for mixed databases. This paper proposes two formulations of local PCA for mixed databases based on optimal scaling which include an additional step of calculating numerical scores of categorical variables.

View full abstract

Independent Component Analysis (ICA) is a technique for blind source separation and is also useful in regression (prediction) task when only a subset of random variables is observed. Local independent component analysis (Local ICA) is a non-linear extension of linear ICA models that extracts local feature values by applying linear ICA in conjunction with suitable clustering algorithms. This paper proposes a switching regression model, in which local linear structure is first captured by fuzzy c-regression, and then a non-linear regression model is estimated by a modified ICA model considering fuzzy memberships in each cluster.

View full abstract

A postsupervised FCM classifier using a modified Cauchy weight as a membership function has been proposed. The performances of the classifier in terms of ROC and rejection curves are compared by using the benchmark datasets of the UCI ML repository. Cauchy weight function gives lower confidence to the classification decision and this distinguishes the FCM classifier from the Gaussian classifier. The FCM classifier outperforms the Gaussian classifier and the k-nearest neighbor classifier on many datasets in general.

View full abstract

Miyamoto et al. derived a hard clustering algorithms by defuzzifying a generalized entropybased fuzzy c-means in which covariance matrices are introduced as decision variables. We apply the hard c-means (HCM) clustering algorithms to a postsupervised classifier to improve resubstitution error rate by choosing the best clustering results from local minima of an objective function. Due to the nature of the prototype based classifier, the error rates can easily be improved by increasing the number of clusters with the cost of computer memory and CPU time. But, with the HCM classifier, the resubstitution error rate along with the data set compression ratio is improved on several benchmark data sets by using small number of clusters for each class.

View full abstract

In this paper, we propose a new fuzzy clustering algorithm for cooccurrence matrix. In our proposal, the graded possibilistic approach is applied to estimation of memberships of items for deriving the absolute responsibilities of them. The memberships can be regarded as the probability that an experimental outcome coincides with one of mutually independent events. Then, soft transition of memberships from probabilistic to possibilistic constraint is performed by using the graded possibilistic constraint in the approach.

View full abstract

In this paper we propose kernelized validity measures where a kernel means the kernel function usedin support vector machines. Two measures are considered: one is the sum of the traces of the fuzzy covariances within clusters. Why we consider the trace instead of the determinant is that the calculation of the determinant will be ill-posed when kernelized, while the trace is sound and easily computed. The second is a kernelized Xie-Beni's measure. These two measures are applied to the determination of the number of clusters having nonlinear boundaries generated by kernelized clustering algorithms. Another application of the measures is the evaluation of robustness of different algorithms with respect to fluctuation of initial values.

View full abstract

This paper describes a family of methods of fuzzy clustering handling objects with weights. Weighted objects easily appear when an individual is a representative of several data units. Fuzzy c-means and possibilistic clustering algorithms for weighted objects are proposed. Relationships as well as differences between solutions of possibilistic and fuzzy c-means methods are described. Moreover a kernelized fuzzy c-means algorithm with weighted objects is studied. Numerical examples show effectiveness of the proposed methods.

View full abstract

Recently, the document data of news, websites and papers have accumulated in the database every year. And the amount of information increases substantially. Therefore, we can not understand and search the matter from all data. In this paper, we proposed a new clustering model for information retrieval using kernel aggolomerative hierarchical clustering(KAHC). Moreover, we show a new visualization technique for the result of data using KAHC and kernel Fisher discriminant analysis(Kernel-FDA). This visually will be able to verify information on related to the result of the information data. As a result, the data that was classified on the feature space can be visualized on the low dimentional space. Through some examples, we show to display the retrieval result more effectively becomes possible.

View full abstract

This paper proposes a fuzzy clustering method considering weights of variable-based dissimilarities over objects. In order to estimate the weights, we propose two methods. One is a method in which we directly use conventional fuzzy clustering for the variable-based dissimilarity data. The other is to use a new objective function. Exploiting the weights into the conventional dissimilarity, we define a dissimilarity assigned with the significance of each variable for the classification and implement a fuzzy clustering method under the intrinsically weighted classification structure of data.

View full abstract

This paper describes algorithms for classification and clustering of information objects on the basis of similarity measures using the concept of a fuzzy neighborhood. Algorithms for application to information retrieval are proposed. Illustrative examples are given and effectiveness of different algorithms are compared.

View full abstract

The Fisher kernel, which refers to the inner product in the feature space of the Fisher score, has been known to be a successful tool for feature extraction using a probabilistic model. If an appropriate probabilistic model for given data is known, the Fisher kernel provides a discriminative classifier with good generalization. However, if the distribution is unknown, it is difficult to obtain an appropriate Fisher kernel. In this paper, we propose a new nonparametric Fisher-like kernel derived from fuzzy clustering instead of a probabilistic model, noting that fuzzy clustering methods such as a family of fuzzy c-means are highly related to probabilistic models. Numerical examples show the effectiveness of the proposed method.

View full abstract

In this research, we propose a color extraction method for deciding the best color threshold to remove noises except an ellipse target object selected by human from color static image obtained by an omnidirectional camera. In this method, the automatic tuning of color extraction threshold is performed by genetic algorithm (GA) that searches the best color threshold according to the color information in the area selected by human. We report experimental results with the threshold information before or aftar GA search. By these experimental results, the performance of proposed color extraction method are also evaluated by comparing the extraction result by human to GA.

View full abstract

At present day, public infrastructures such as bridges and roads can be seen all around. The continuous construction of such infrastructures is posing an important problem of how to maintain and efficiently manage an existing structure. In this paper, system of extracting characteristics of cracks showing up on damage levels is attempted by using these results. So, in this research, in order to aim at maintenance of a structure and the increase in efficiency of management, and to extract a crack from a crack picture, it experimented using the Gabor function.

View full abstract

A slant correction is one of the important preprocessings before an image processing such as a character recognition. In general, the results of that preprocessing influence the performance of the image processing. We present an effective slant estimation method employing a half-cosine function wavelet network, which is superior in feature extraction. The proposed method executes an character area extraction and slant estimation simultaneously. In this paper, the slant estimations of a character sequence in various slant angles are examined and the validity of the proposed method is discussed.

View full abstract

We show a method for extracting physical movie features that may influence movie viewers' physiological responses without physiological measurements and extracted physical features by the method. For our new challenge, emotion control by controlling physical movie features, we need to choose physical movie features that influence human physiological responses, find the relationship between the physical features and the physiological responses, and control the physical features to control the physiological responses indirectly. In this paper, we discuss the first stage.

View full abstract

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

View full abstract

We propose a model for expressing speed impression caused by switching speed of movie shot. Speed impression is one of the producers' intentions, and realized by changing colors, moving objects, and others. As there are many time-sequential data that make us feel different speed impression, it is difficult to express our speed impress numerically even if we restrict the factor to switching speed of movie shot. We propose a speed impression model that inputs the time-sequential data and outputs one value of our speed impression. First, we evaluate human speed impressions by several movie shot scenes and subjective test. Second, we fit our proposed model to the measured subjective data by tuning the model parameters with GA. Finally, we discuss a relationship between model outputs and human speed impression, and effectiveness.

View full abstract