In this paper we have dealt with prediction of fish motion under the vision system provided by CCD camera and embedded chaos motion into the system for more effective catching action. Taken into the consideration of emotional aspect, the fish can suddenly change its escaping trajectory or speed up when being threatened such as the net attached at hand approaching to it. Furthermore, as the time of tracking process flows, the fish can somewhat get accustomed to the environment and begin to learn new strategies about how to escape from the bothering net. For example, the fish tends to stay within a corner where it is forbidden for the net to reach for safety or stays away from the net by keeping a constant distance. For the sake of such innate ability being widely existed in animal behavior, the effective intelligent method will need to be conceived to go beyond the fish intelligence. The purpose of this paper is to construct an intelligent system that is able to exceed the fish intelligence in order to track and catch the fish successfully like fish-eating animals in nature to survive.

View full abstract

This paper deals with closure properties of models of fuzzy attribute implications. Fuzzy attribute implications are particular IF-THEN rules which can be interpreted in data tables with fuzzy attributes and/or ranked data tables over domains with similarity relations. We show that models of any set of fuzzy attribute implications form a fuzzy closure system with hedge. Conversely, we show that each fuzzy closure system with hedge can be seen as a system of models of some set of fuzzy attribute implications. Furthermore, we show applications of the closure properties: we describe semantic entailment from fuzzy attribute implications using least models and show a method for generating of non-redundant sets of dependencies which are true in a given data table with fuzzy attributes.

View full abstract

In this paper, we pursue application of Gibbs measure theory to LBP in two ways. First, we show this theory can be applied directly to LBP for factor graphs, where one can use higher-order potentials. Consequently, we show beliefs are just marginal probabilities for a certain Gibbs measure on a computation tree. We also give a convergence criterion using this tree. Second, to see the usefulness of this approach, we apply a well-known general condition and a special condition, which are developed in Gibbs measure theory, to LBP. We compare these two criteria and another criterion derived by the best present result. Consequently, we show that the special condition is better than the others and also show the general condition is better than the best present result when the influence of one-body potentials is sufficiently large. These results surely encourage the use of Gibbs measure theory in this area.

View full abstract

We propose a new similarity measure between two fuzzy sets based on their relative sigma count and extend it to define two other measures, one a similarity measure between elements in fuzzy sets and the second a similarity measure between fuzzy sets in which all elements in the universe of discourse are weighted. We compare our proposal to several previous measures proposed in [1-6].

View full abstract

The paper presents a general method of imposing constraints in formal concept analysis of tabular data with fuzzy attributes. The constraints represent a user-defined requirements which are supplied along with the input data table. The main effect is to filter-out conceptual clusters (outputs of the analysis) which are not compatible with the constraint, in a computationally efficient way. Our approach covers several examples studied before, e.g. crisply generated concepts and constraints by hedges.

View full abstract

In this paper, we present a new similarity measure between generalized fuzzy numbers. It combines the concepts of geometric distance, the perimeter and the height of generalized fuzzy numbers for calculating the degree of similarity between generalized fuzzy numbers. We also prove three properties of the proposed similarity measure. We use fifteen sets of generalized fuzzy numbers to compare the measure results of the proposed method with the existing methods. The proposed method can overcome the drawbacks of the existing methods. It provides a useful way to measure the degree of similarity between generalized fuzzy numbers.

View full abstract

This paper describe a new topological map generation method for mobile robot. The proposed system only requires a laser rangefinder and can automatically generate topological map based on Delaunay triangulation method. To confirm the proposed new topological map generation method, an electric wheelchair based mobile robot is used to implemented and tested. Experiments in indoor environment demonstrate validity and feasibility of the proposed methods of topological map generation.

View full abstract

Path planning for an autonomous mobile robot is very important for route navigation. This paper describes a smooth path generation method for an autonomous mobile robot. To achieve smooth path even in a high-speed situation, we employ omni-directional image and laser range profile data to generate an appropriate path based on a clothoidal curve. The generated clothoidal path can minimize the effect of lateral force. The validity of the proposed method is verified by both simulations and actual implementation.

View full abstract

This paper deals with control of auto-driving cooperative system considering the safety of vehicle. Fuzzy instruction is fuzzy set of control instruction which could be as control command candidates. According to the current situation of the vehicle the fuzzy instruction is calculated and auto-driving system is constructed. The cooperative control system was purposed to support the human's operation based on the fuzzy instruction, and the support is shown through the variable reaction force to human. An impedance controller was designed for the human's operation, and the controller which can most effectively represent the safety of the running vehicle was identified. The cooperative control system is evaluated by an experiment using driving simulator, and the validity of the proposed design procedure is confirmed.

View full abstract

In this research, we focused on using pointing behavior for a natural interface. We investigated to find a system that recognizes users? intentions by using their gestural information in particular situations. The system recognizes objects or the pointing direction, and decides its own action by automatically considering context. In the first, the user pointed out a specific parking space in which the system was to park a car. In the second, the user pointed out a space in front of a car to call the car. And we make this system by RT middleware.

View full abstract

Recently, the number of accidents for an electric four-wheeled vehicle increases. User's driving skill and knowledge shortage are enumerated as the cause. Especially, it is difficult to judge other traffic participants' movement and to decide the content of the evasive action. Up to the present, we had developed an automatic driving system that can correspond also to a dynamic obstacle by building the driving algorithm based on the predictive fuzzy control into the controller, and adding the driving knowledge to human safety. In this research, we add the function to acquire the action of the future by the conversation with a dynamic obstacle based on this system, and construct an automatic driving system of an electric four-wheeled vehicle that does driving that considers the acquired action of the future.

View full abstract

The K-means algorithm is well known as the most popular clustering method because of its good performance. However, we cannot extract meaningful clusters by applying Kmeans to data that are not linearly separable. For overcoming this difficulty, some researchers have adopted high-dimensional data mappings such as kernel mappings to clustering methods including K-means. We may extract meaningful clusters by using such mapping but requires high-computational costs and a large amount of memory. We present a simple algorithm including Kmeans for extracting arbitrary-shaped clusters by mapping data into a one-dimensional space. This 1D data mapping is achieved by minimizing the square error of the average 1D coordinates of the neighbors of data in input space. After this mapping, Kmeans is applied to the histogram of the distribution of the data in 1D space. The performance of our method is verified with the experiments on synthetic 2D data, image and video segmentation.

View full abstract

In this paper, one of soft computing techniques, fuzzy, is applied for automatic detection of a humanitarian land mind. A ""feature in-decision out"" fuzzy sensor fusion algorithm for a ground penetrating radar, (GPR), and a metal detector, (MD), for mine detection is introduced. The inputs to the fuzzy fusion system are features extracted from both GPR and MD measurements. The output from the fuzzy fusion system is a decision if there is a landmine and at what depth it would be. Fuzzy fusion rules are extracted from training data through a fuzzy learning algorithm. Experimetal test results are presented to demonstrate the validity of the proposed fuzzy fusion algorithm and hence its influence in minimizing the false alarm rate for mine detection.

View full abstract

The ink drop spread (IDS) method is a modeling technique that has several advantages in robustness, real-time capabilities, tractability, and interpretability of models; thus, it has a good potential to be a useful soft computing tool. In this method, the structure of models is determined by the partitioning of the input domain; finding the optimal number of partitions is the most effective means for achieving high model accuracy. This paper proposes a basic constructive algorithm for the structural optimization of IDS models and presents the performance of the IDS modeling in regression and classification tasks using three-input nonlinear systems.

View full abstract

Recently, Evolutionary Computations (ECs) have been employed to minimize modeling errors between robotic movements on the computer simulation system and the trajectories of an actual mobile robot. Generally, this task is important but it is so difficult. In this paper, we propose the method to minimize the modeling error of between robotic movements and simulation results using co-evolutionary computations with image processing technique. In the proposed method, we employ the video camera system on the ceiling for capturing the robot movements, and the trajectories of the actual mobile robot are detected from captured images by image processing technique and the modeling error is estimated. We experimented using an actual mobile robot to validate the effectiveness of the proposed method, and the results shows that modeling errors are deduced effectively by the proposed method. Finally, this paper describes the problem and future works of this study.

View full abstract

In this paper, we discuss the application of Genetic Algorithm (GA) to the collaborative systems operation optimization. Collaborative systems are server client systems in which a group of collaborators come together to provide service to end-users. The cost function to be optimized is the sum of the service cost and waiting cost. Service cost is due to the hiring of professionals that provide service to customers in the collaborative system. Waiting cost is incurred when customers who are made to wait in long queues do not come to receive service from the system a second time. The number of servers operating at each of the collaborative places, as well as the service time, acts as system constraints. GA finds the minimum value of the cost function under these operational constraints.

View full abstract

This paper presents a context dependent textile image auto annotation system. Typical image auto annotation systems have training data, images that have been annotated manually. Images similar to an input image are found in the training data, and the input image is annotated using the keywords of the similar images. However, the similarity of images may be changed by the context. To solve this problem, we propose a context dependent image auto annotation system that recognizes context by using specialist knowledge. We adopted textile images for experiments that verified the effectiveness of the system.

View full abstract

Surplus bands inevitably occur in VBR encoding that is generally used for data streaming. These bands can be used efficiently by adding the data of static contents. However, there are some delays for adding the data, so that the transmission rates of the total data frequently exceed the maximum communication band. This is because each surplus bandwidth is calculated from the formerly observed data. In this paper, we propose a method to estimate the surplus bandwidth by using the multi-order Markov model together with the quantization of the bandwidth. The proposed method requires so little computational time that it is quite suitable for real-time data processing. The efficiency of the proposed method is evaluated through some computer simulations. Discussions for some problems in future studies are also included.

View full abstract

To robustly and inaudibly embed copyright information as watermark into music, a two channel digital audio watermarking system based on the Least Significant Bit (LSB) method and Echo Hiding (EH) with new Exponential Time-spread Echo Kernel (ETEK) is proposed. Owing to complementary combination of the LSB method and EH, the proposed system exhibits robustness against various attacks. Furthermore, the proposed ETEK for EH, which creates multiple echoes as reverberation in a real room, makes the sound of the watermarked music more natural than the one obtained by the conventional single echo kernel. Experiments of extracting watermarks from watermarked clips which are processed through various attacks show that the proposed system obtains higher watermark detection rate, or robustness, 26.09% higher than the LSB method and 7.42% higher than EH, moreover, robustness of the proposed ETEK also increases 14% higher than that of the conventional EH. These results of robustness, together with subjective evaluations of sound quality confirm the internet availability in business of the proposed system.

View full abstract

This paper presents changes in the EEG pattern observed after Bhramari Pranayama (BP) or bee like breathing exercise. BP has been found effective in healing many mental problems such as tension, stress, hypertension, etc and it brings relaxation in the practitioners of it. It has also been said that BP is very quick in producing results. The analysis of EEG data recorded before and after BP shows similar and positive changes in different brainwave patterns, related to relaxed state, both for an experienced and a new subjects.

View full abstract

Accounts of real situations which cause problems for domain users are relatively rare in academic literature because the problems which confound users often occur in a single way. This article argues that while such accounts cannot be based on traditional research methods and the data lack the ability to be easily replicated and verified, such accounts do have value in serving as a guide to the possibility of the problem having wider occurrence. An anecdotal account of a problem encountered in the real domain of train travelling with a bicycle in the UK is given and this is used to illustrate the points made about anecdotal evidence in general.

View full abstract

This paper presents a mutation-based evolutionary algorithm that evolves genes for regulating developmental timings of phenotypic values, which is meant to bring different evolution speed to each phenotypic variable. A genotype in the evolutionary algorithm time-sequentially generates a given number of entire phenotypes and then finishes its life at each generation. Each gene represents a cycle time of changing a probability for determining its corresponding phenotypic value in a life span of the genotype. This cycle time can be considered a sort of information on developmental timing. This paper also discusses a new approach to depicting an evolutionary optimization process. The approach depicts an evolutionary optimization process as change in a network topology that emerges in the process. An evolutionary optimization process involves identification of linkage between variables, and ability of an evolutionary algorithm in identifying the linkage influences the search efficiency. Therefore, network structures formed by using the identified linkage information in the evolutionary process draw how the evolutionary algorithm solves a given optimization problem. The experimental results show that evolving developmental timings helps to sequentially solve problems with linkage between variables and also that power-law-like network topologies emerge in the optimization process of the mutation-based evolutionary algorithm for any problems used.

View full abstract

In this paper, for accurate classification in arbitrary distributed patterns in sparse distribution, we propose a classification method using nearest neighbor method based on the mean of norm in prototypes in each class which are minimum distance from an input pattern and Its k neighbor prototypes.

View full abstract

Suppose there are normal populations each with unknown mean and unknown variance. We consider two problems in this paper. The first, for some given values of s and t, we are interested in selecting the one whose mean is between s and s+t'(t'>0) and simultaneously its mean is closest to t, otherwise, none is selected. The second, for given values of s, a square of c, and t'(t'> 0), we are interested in selecting the one whose mean deviates from s within t' and those variance is no large than a square of c and simultaneously its mean is closest to s. The problem is formulated in a Bayes set up. Empirical Bayes rules are derived and they have been shown to be asymptotically optimal.

View full abstract

A learning method using a weighted integration of individually trained multiple component predictors as an ultimate predictor is generically referred to as ensemble learning. The ultimate predictor is called an ensemble predictor, and the parameter for the integration is called a weight parameter. The present paper proposes a weight parameter estimation method for ensemble learning in the following situation. Although we do not have any additional training data for the weight parameter estimation, we can assume that the accuracy of the individually trained multiple predictors are approximately the same. The proposed method is naturally derived from an ensemble learning model, which is based on an exponential mixture of probability density functions and Kullback divergence. We show that the proposed method gives the theoretically best strategy for the weight parameter estimation under the above-mentioned situation. We verify the effectiveness of the proposed method through numerical experiments.

View full abstract

Mentality expression system by eye robot on affinity pleasure-arousal space is proposed for the communication between human beings and robots, where mentality status is calculated by fuzzy inference for speech understanding module. The constructed eye robot system expresses mentality by comprehensible moving, and realizes an interface for new type of cozy information terminal. 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 a basic interface for mascot robot which is easily understandable information terminal at home environment in ubiquitous society.

View full abstract

This paper proposes a TV program recommendation system based on human-robot casual communication under living environment. In these days, partner robots which live together with human are promising for various applications. For the purposes, partner robots do not only have to follow user's every command for prearranged procedures and services, but also communicate with humans in similar ways as human to human communication. In this paper, we propose a human - robot communication system for recommending TV program and related information based on user's interests. The system gathers user's TV watching log and his/her utterances with speech recognition module, based on which his/her interests are estimated and represented as a user profile with the form of bookmarks. A user profile is used for recommending information in which the user is interested. This paper introduces the system architecture and the algorithm for generating user profiles from TV watching logs.

View full abstract

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

View full abstract

Architectonics of systems and communities are becoming more complicated everyday. Communities help one another by riding on their own strength. Similarly, each matured technology needs help mutually to develop. In particular, the wide development that covers not only Robot Technology (RT), but also Automotive Technology (AT) and Information Technology (IT) is expected in nowadays. They are growing in a spiral by interacting one another. We introduce systems in which these technologies collaborate and our approach for intelligent systems. Finally, we show an experiment of Networked Robots and ontological network.

View full abstract

We focused on using natural interface for interaction between a human and robots. Recently, robots that fit into the human community are needed. A gestural interface is important for these robots. Actually, gestural recognition has already been studied for use in a human-machine interface. We investigated to find a system that recognizes users' intentions by using their gestural information in particular situations. We aim constructing interactive communication. Therefore, firstly, we constructed a system, which a use point out an object using virtual room interface. Our aim is not to calculate accurately the point indicated by a user. The system recognizes objects from interaction with user. Moreover, we show that a robot direct a user's attention to an object using pointing behavior. To acquire a pointing behavior, we constructed virtual robot. We implement the motion, which was acquired by virtual robot to actual robot.

View full abstract

The feature selection is to select a set of attributes that is relevant for a given task. Due to the presence of irrelevant and redundant attributes, higher predictive accuracy can be expected by selecting only the relevant attributes from the dataset. Conventional feature selection was done after discretization. In this case, however, network size and computation would be increased drastically. In this paper, we introduced a rank-based feature selection (RFS) method using positive correlation learning (PCL). A higher-ranking attribute is the one that has a lower distance from a reference point. We consider an attribute to be relevant if its rank is higher. We empirically proved that if irrelevant features are removed, higher classification accuracy can be achieved. This is because PCL tries to create smaller coherent weights. We applied this approach to diabetes problem. We show empirically that RFS can easily remove the irrelevant features and produce better accuracy and generalization.

View full abstract

We propose a neural network model for word sense disambiguation using up and down state neurons and morphoelectrotonic transform. Relations between stimulus words and associated words are implemented on this neural network by using an associative ontology. This new neural coding model enables word sense disambiguation in an input sentence by using firing dynamics on the neural network. It is decided whether to put a new link between two neurons by using a co-occurrence frequency between two words corresponding the neurons and an attenuation rate of morphoelectrotonic potential between the two neuron. The distance of the new link is obtained by learning from calculating the morphoelectrotonic transform from the two neuron's morphoelectrotonic potential. It is shown that this model has a small-world structure by analyzing the learning behavior using average shortest path lengths and clustering coefficients.

View full abstract

Neural networks can approximate arbitrary nonlinear functions. Neural networks are often designed by trial-and-error. Though the design methods such as brute-force approaches, network construction and pruning have been proposed, a generic approach in which other conditions such as learning parameters are taken into account is rarely met. This paper presents an efficient approach to decide multiple design parameters of multilayer neural networks by using the Design of Experiment, whose features are efficient experiments by an orthogonal array and quantitative analysis by analysis of variance. We pick up not only the number of hidden nodes but also initial conditions and learning parameters as the design parameters. In this paper, our approach is applied to 3-layer feedforward neural networks (FNNs) and 5-layer FNNs. We also show that the approximation accuracy of multilayer neural network can increase by picking up much more parameters.

View full abstract

This paper presents a new constructive algorithm known as multilayered constructive architecture (MCA) for designing and training multiple hidden layered artificial neural networks (ANNs). Unlike most previous constructive algorithms, MCA puts emphasis on both simplicity and generalization ability of designed ANNs. In order to maintain simplicity, both the number of hidden layers and nodes in a hidden layer are determined by a constructive approach. The use of modified single layer backpropagation algorithm for training output layer and hidden neurons increases the generalization ability of the ANN. MCA has been tested extensively on a number of benchmark problems in machine learning and neural networks, including Australian credit card assessment, breast cancer, diabetes, glass and heart disease. The experimental results show that MCA can produce compact ANNs with good generalization ability.

View full abstract

This paper presents construction of an adaptive neural network ensemble i.e., try to solve the given problem with single neural network and build ensemble with minimum architecture when required. Exiting methods always think multiple networks for ensemble case though some problem may solvable by single network. This new method utilizes constructive approach to determine ensemble size automatically where it adds networks one by one and uses cumulative number of hidden nodes for coming networks. Also, at the time of network addition, a new network is motivated on previously unsolved portion of training space. Finally all the networks are trained simultaneously. This new method has been tested extensively on several benchmark problems of machine learning and neural networks. Experi-mental results show that this method able to construct adaptive ensemble in which some problem is solved by single neural network and for multiple networks case used minimum ensemble structure.

View full abstract

The aim of the research is to make decision for the best MA alternative under multi-objectives. In this paper, criterions for evaluating of MA are suggested. Criterions are based on convenience, engine room size, interrelation of equipment, and pipe producibility of an alternative. Fuzzy theory was adopted to quantify the linguistic considerations. After evaluating alternatives, TOPSIS algorithm was used for determining the best alternative reflecting multiple attributes.

View full abstract

Implementation of a context-awareness framework for ubiquitous computing environment

View full abstract

This paper presents the generalized Petri Nets (PN) structure to describe an application instruction of PLC. The generalized PN structure for the application instruction is newly constructed using the register arc. The state equation that corresponds to a data move function, a data add function and so on, is created. After converting a ladder program into a PN graph, we construct the generalized state equation from the PN graph.

View full abstract

In this paper, a new method is proposed to recognize human side view faces in a video stream. This work is aimed at ultimately being used as an aid for the gait recognition because the gait biometric alone is not enough for the reliable identification. We employ a probabilistic approach to fuse the information in the successive video frames. In the proposed method, we use the ICA (FastICA), histogram filter, to extract features from side view face images and then employ PSVM (Probability SVM) as a classifier in the probabilistic framework. Experimental result shows the suggested method is valid and effective for the identification in a video stream.

View full abstract

One of the obstacles to building practical fuzzy database systems is to get semantic data such a proximity relation or similarity relation. The proximity relation is represented by the degree of ""closeness"" or ""similarity"" between data objects of a scalar domain. A fuzzy database system evaluates imprecise queries with the proximity relations. In this paper, a systematic proximity elicitation and efficient representations of the proximity relation are proposed. The proposed generation method is simple and systematic. It is based on the well-known fuzzy set theory and applicable to the real world applications because it has tuning parameter and weighting factor. The proposed representations of proximity relation are more efficient than the ordinary matrix representation since they reflect some properties of a proximity relation to save space. We show an example of quantitative calculate for the proximity relation. And we analyze the time complexity and the space occupancy of the proposed representation methods.

View full abstract

The conceptual framework is presented for dynamic visualization of evolving networks which represent a historical transition of social relationships and interactions. We consider a sequence of large number of networks embedded in a stack of 2-dimensional layers, and provide an alternative viewpoint of traversing a lifetime of a system along the third axis as well as staring an each snapshot of a graph individually. In general, social systems are described as a complex emerging system based on the aggregation of individual human behaviors caused by their purpose and whim. As a result, system dynamics has intrinsic contingency, which makes identifying the causal efficacy of social events very difficult. The objective of the proposed research is to tackle such issues by developing a methodological tool for visualizing the historical changes of network structures over time, space and different analytical conditions. We discuss a cooperative working between a graph layout algorithm of each snapshot and a time series representation of superposition of networks under interactive topological operations.

View full abstract

The concept of an intuitionistic fuzzy set is a generalization of the concept of a fuzzy set. An intuitionistic fuzzy set has three values for any element of universe of discourse U, that is, a degree of membership(A), a degree of non-membership(B) and a degree of unknown (or hesitancy )( C). We can treat a degree of contradiction by using an extended interval truth value. In this presentation the relationship and the algebraic properties of intuitionistic fuzzy sets and extended interval truth values are clarified.

View full abstract

This paper presents an overview of Kansei System.The primary purpose of this paper is to construct a framework of Kansei system. It is shown that the Kansei system consists of four quadrants that correspond to Kansei Expression Theory, Kansei Recognition Theory, Cognitive Science and Modeling. Also a survey of the paper in each area is given to realize the framework. Therefore the objective of this paper is to discuss the basic concepts and methods of the system and provide a framework for future Kansei oriented research.

View full abstract

In this paper, we describe advantages of the simultaneous perturbation optimization method for some biomimetic systems. Especially, learning mechanisms for neural networks, neural systems and blind source separation are good examples. We explain about the applications of the simultaneous perturbation for these systems.

View full abstract

A new real-time two-stage blind source separation (BSS) method for convolutive mixtures of speech is proposed, in which a single-input multiple-output (SIMO)-model-based independent component analysis (ICA) and a new SIMO-model-based binary masking are combined. SIMO-model-based ICA can separate the mixed signals, not into monaural source signals but into SIMO-model-based signals from independent sources in their original form at the microphones. Thus, the separated signals of SIMO-model-based ICA can maintain the spatial qualities of each sound source. Owing to this attractive property, novel SIMO-model-based binary masking can be applied to efficiently remove the residual interference components after SIMO-model-based ICA. In addition, the performance deterioration due to the latency problem in ICA can be mitigated by introducing real-time binary masking. We develop a pocket-size real-time DSP module implementing the new BSS method, and report the experimental evaluation of the proposed method's superiority to the conventional BSS methods, regarding moving-sound separation.

View full abstract

Gabor filtering is known as a function model of the primary visual cortex in the brain. It is also known as a powerful feature extraction method for image recognition. This paper presents an image processing system using the dedicated Gabor-filtering LSI that we have already developed based on a merged analog-digital architecture. The Gabor-filtering LSI is controlled by an FPGA connected with a personal computer and a camera. The developed system can successfully output a Gabor impulse response with even and odd functions, and also detect a stripe-pattern image with a given direction in real time.

View full abstract

This paper proposes hardware architecture for the self-organizing map (SOM). Due to the parallel structure of the SOM, a considerable speed-up can be expected by the custom hardware. After verifying the operation of the hardware SOM, it is extended to implement a feedback SOM. The feedback SOM is used to identify the location of a mobile robot. The proposed SOM design is described by VHDL and its performance is tested by simulations using actual sensory data from the experimental robot. Results show that the feedback SOM successfully segments the environment where the robot operates. Also the circuit size of the SOM is considerably small and its performance is estimated as 3,400 million weight update per second.

View full abstract

The growth of mobile services on mobile phones or PDA has been significant in the recent years. This has opened up a new arena for many organizations to implement services on these mobile devices. In this paper, we propose an extension to our work in Artificial Life; of having our embodied conversational agents (ECAs) framework to be extended for mobile handheld devices. We present an overview and investigate new challenges on the implementation of the ECAs on the cross-platform architecture suitable for small end user devices. We demonstrated the proposed downsizing framework suitable for mobile devices with the application in the area of crisis communication.

View full abstract

Our earlier work proposed and discussed the issues of a method for obtaining weights, which are associated with the weighted relevance aggregation method, for hierarchical fuzzy signatures from real world data. This method also handled the non-differentiability of conventional max-min aggregation functions, using a mathematically proved method in the literature. This paper applies the proposed method to extract weights for two real world hierarchical fuzzy signatures structures namely Salary Selection and SARS Patient Classification. Based on the results of the experiments for weights extraction with SARS and non-SARS patients data we show that our weights learning method for hierarchical fuzzy signatures not only performs better in separating SARS and non-SARS patients, but also separating non-SARS data into further significantly distinguished and ordered available categories.

View full abstract

Traditional researches on the classification problem concern that a complete dataset is given as a training set without missing. However, incomplete data usually exist in real-world applications. In this paper, to handle incomplete numerical data in the classification problem, we propose a new approach based on fuzzy entropy. The proposed approach of handling incomplete data uses the technique of granular processing of fuzzy similarity measure to fill missing values of attributes. The experiments were made and the results were compared with the method of AMSC (attribute mean with same concept) through a few famous classification models to evaluate the performance of the proposed handling method.

View full abstract