The purpose of this study is to develop general methods which discerns differences among feature distributions. In this paper we propose two methods, NG+SOM and NGxSOM, which are combination of Neural Gas (NG) and Self-Organizing Map (SOM). These two methods, NG+SOM and NGxSOM are based on Bag of Features and Higher rank Self-organizing Map, respectively. Those definitions of ``difference among feature distributions'' differ from each other. So in this paper, we do comparative verification of our two methods through two classification tasks with artificial feature distribution and real image feature distribution (SIFT).

View full abstract

This study aims to develop an observation-invariant classification algorithm of multi-class models.The task of the architecture is to classify a set of datasets, each of which is transformed differently when it is observed.Therefore it is required to remove observation-dependent component from the datasets.The higher-rank of SOM (SOM^n or `SOM of SOMs') was employed for multi-datasets classification task, and an observation-invariant distance measured was introduced into the SOM^n algorithm.In this paper, we applied the algorithm to a classification task of a set of `skyline shapes'.

View full abstract

Recently, we have proposed Self-Evolving Modular network (SEEM), in which number of modules is increased depending on input data episodes in self-organizing manner.While SEEM can find the number of modules for a task, it is necessary to preliminarily set the number of modules in conventional modular network. In this study, various experiments are conducted to verify the effectiveness of SEEM.

View full abstract

A fraction-applicable self-organizing map (FASOM) is an extended version of the standard SOM for dealing with partial input data. In this article, how to identify the category of applied unknown data is investigated mainly on the assumption that a feature map has been already developed successfully. As a result of computer simulations with a real data set, it is observed that the selected area as candidates in the competitive layer becomes smaller and smaller with decreasing the missing elements in the applied data. At the same time, however, it is also confirmed that how to assign each region affects the FASOM's performance.

View full abstract

Self-Organizing Feature Map (SOM) is a layered neural network consisting of an input layer and a competitive layer for the data visualization and vector quantization. The accuracy of SOM vector quantization depends on the number of competitive layer's neurons because the codebook vectors correspond to the competitive layer's neurons. Therefore, when an unknown data set is given, it is difficult to decide the sufficient competitive layer size. In this paper, we propose a Tree-Structured SOM (TS-SOM) based method in order to adaptively change the competitive layer size and structure. TS-SOM is a faster SOM method applying a tree search algorithm. We applied the pruning of neurons and layer creation to the tree structure of TS-SOM by using the error among neighboring neurons.

View full abstract

In recent years, image recognition technology, such as becomes to popular digital camera. Image recognition is performed using classifier that previously learned pattern to be detected. But it is difficult to learn all of the appearance of the target detection. In this paper, we proposed classifier that using the principle for self-organizing maps (SOM). In proposed method, the classifiers are aligned on SOM in the learning process. If the unit is adjacent to other units, mutual units have similar nature, or else have a different one. In that recognition, the system comprehensively decides by response SOM units. The effectiveness of proposed method is demonstrated by experiments on some real and artificial data.

View full abstract

Character recognition is a classical issue which has been devoted by a lot of researchers.Making character recognition system more widely available in natural scene images might open up interesting possibility to use as an input interface of characters and an annotation method for images. Nevertheless,it is still difficult to recognize all sorts of fonts including decorated characters such as characters depicted on signboards. The decorated characters are constructed by using some special techniques for attracting viewers' attentions.Therefore,it is hard to obtain good recognition results by the existing OCRs.In this paper, we propose a new character recognition system using SOM.The SOM is employed to extract an essential structure concerning topology from a character.The extracted topological structure from each character is used to matching and the recognition is performed on the basis of the topological matching.Experimental results show the effectiveness of the proposed method in most forms of characters.

View full abstract

Recently, the spread of the Internet makes familiar to the incident concerning the Internet, such as a DoS attack and a DDoS attack. Some methods which detect the abnormal traffics in the network using the information from headers and payloads of IP-packets transmitted in the networks are proposed. In this research, the method for the analysis of the flow of IP packet based on SOM (Self-Organizing Map) using the occurrence rate of the elements in IP-packets as input vectors is proposed. The flowofthepacketscanbevisualizedonthemapanditwill be available for detecting the abnormal flows of packets.

View full abstract

The liver has a lot of important functions and is needed metabolisms in the living body. The liver diseases are many signs but their diagnosing has some difficulties. For example, the diagnosis of liver diseases is needed knowledge of a great of number of blood test items (e.g. T-BiL , Alp, gamma-GTP, LDH, ChE, GPT, GOT, Plt, Alb, AFP, and so on) and of the liver diseases(Acute hepatitis, Chronic hepatitis, Liver cirrhosis, Hepatoma)under the mutual relation. Then the researchers have challenged to this problem with several computational technique. The standard distance in general is Euclid distance but The Taguchi Method using Mahalanobis distance appreciated at diagnostic support of the medical. Mahalanobis distance is a distance measure including the variation in data. SOM (Self-Organizing Map) is drowning the second dimension map which there is calculated node near parallel nodes, and a spherical SOM is modified and progressive version of original SOM. We propose a new diagnostic method for liver disease using a spherical SOM constructed by the Mahalanobis Distance. This thesis will give an examination the usefulness of the data analysis of the medical from now on.

View full abstract

Recently, SOM (Self-Organizing Map) is noticed as one of the analysis techniques of multi-dimension data, and it is actively studied and applied in various fields, such as medical treatment and environment. In clustering by the conventional SOM, data is classified according to compressing the inputted multi-dimension data into two dimensions. However, information on the cause of existing inside data cannot be analyzed. Therefore, ICA (Independent Component Analysis) is taken in to the conventional SOM, data is decomposed into some components, and the extraction and analysis of an independent component having contained the amount of the characteristics of input data are conducted. As a result, we thought that it enables a high level clustering which solved the above-mentioned problem. In this research, we develop the analysis technique (ICASOM) which combines SOM and ICA, and prove the effectiveness by applying this method to the acceleration plethysmogram.

View full abstract

The new classification method was proposed where the multi-dimensional data set can be easily classified visually. There, a phase distance on the spherical surface between the label data was computed and the dendrogram was constructed using this spherical surface distances among the labels. Here, based on this constructed dendrogram, the classification of the cluster group on the spherical surface based on the label data was classified and colored. As for this way, it is possible to apply to a variety of data set. However, first, the AES spectrum (the chemical analysis data) was successfully classified and colored. Other examples will be also included for the presentation.

View full abstract

In the environment with uncertainty, to continue the appropriate behavior, the agent has to be equipped with means to judge whether change of the environment caused by his own behavior is good or not. In this paper we propose a novel reinforcement learning system which incorporates the above means, that is, a value system. Through computer simulations using maze problems, it is verified that the proposed system is valid by comparing with the conventional reinforcement learning system.

View full abstract

Human beings are able to acquire the motion patterns efficiently by observing and imitating the demonstrator's motions. This paper presents a method, which enables the real robot assembled by LEGO Mindstorms NXT to acquire the walk movement by the imitation learning. At first, we prepare two real robots, each of which is shaped like a measuring worm. Then, we use the motion capture system in order to observe the motion patterns of the demonstrator robot. Based on the observed time series data of the joint angles, we realize the imitation learning by using neural networks. Through the experiments, we show the learner robot is able to acquire the motion patterns, which are similar to the motions of the demonstrator robot.

View full abstract

The introduction of the reinforcement learning to the multi-robot systems is researched. The performance of reinforcement learning is quite sensitive to the segmentation of state and action spaces. To overcome this problem, we have been developing a new technique, Bayesian-discrimination-function-based RL (BRL). BRL is capable of the adaptive division of the state space. In BRL, the over-fitting occasionally occurred. This paper introduces an extended BRL method to restrain over-fitting. The extended BRL method is the technique that introduced meta learning into BRL. This method coordinates learning rate to the learning degree of progress, and uses rule ignition entropy for an index of the learning degree of progress. The standard BRL and the extended BRL are comparing verified through the task of movement by the connected robots. Through a cooperative task by the connected robots, performance of the extended BRL is compared with standard BRL.

View full abstract

Swarm robotic systems is the research field of multi-robot systems which consist of many homogeneous autonomous robots without any type of global controllers. Generally, since a task given to this system can not be achieved by a single autonomous robot, emergent cooperative behavior is expected in a robot swarm by a certain mechanism through the interactions among the robots or with an environment. In this paper, an evolutionary robotics approach, i.e., the method that the robot controllers are designed by evolutionary algorithms with artificial neural networks, is adopted. It has been found that the topology of an ANN must be important for the improving evolvability, but the systematic design method has not been known so far. In this paper, we build ANN with the hidden layer of small world network, and examine whether ANN with the hidden layer of small world network performs better than ANN with the hidden layer of regular or random graph.

View full abstract

Reinforcement Learning(RL) is one of the popular method in machine learning,which can adapt environment automatically. Most of RL such as Q-Learning guarantees to acquire optimum policy. The typical applications of this method are deciding the behavior of complex systems such as multi joint robots and multi-agent systems. However,it is difficult for Q-Learning to treat Partially Observable Markov Decision Process which exist in a lot of real environments. Profit Sharing(PS),which is one of the RL, can treats this process though the obtaining the optimum policy is not guaranteed. There is some proofs of acquiring rational policy in not only single agent systems but also multi agent systems. However,the PS needs a lot of trials to obtain the rational policy in the following environments: the agent requires many steps to get rewards, there are many actions which agent can execute. These problems occur in designing multi joint robot and a lot of agent in multi agent systems. In this paper , we proposed a new PS method which can learn fast in these environments and provide the availability of this method through numerical experimentations. The results show the proposed method can acquire the rational policy fast in comparison with the conventional methods.

View full abstract

Recently, biologically inspired approaches have received much attention for controlling behaviors of robots. A typical example of biologically inspired approaches is control of rhythmic behaviors using Central Pattern Generator (CPG). This method is useful to control adaptive walking behaviors of robots, however, this method has a problem that there are few theories to design CPG. Hence, in order to design CPG, we proposed a learning method called the Three Steps Method which is a combination method of Genetic Algorithms and Reinforcement Learning. In this study, in order to show the effectiveness of the Three Steps Method, we apply it to a quadruped robot with the CPG controller. The simulation results show that the Three Steps Method leads robots to acquire walking behaviors without the knowledge of designers. Furthermore, we also show that the robot can adapt to various environments using the proposed method.

View full abstract

In this paper, we describe a constructive computational model for basic capability of human imitation learning skill. When a person observes another's motion, he/she can project the visual image onto his/her body schema. By calculating the observed visual image back to posture data, an observer can behave as the observed person. In this study, we develop a computational model which integrating visual and posture information based on kernel canonical correlation analysis. We also evaluate the computational model through an experiment.

View full abstract

In this presentation, two-staged dimensionality reductions are examined for Evolutionary Learning: The first stage of dimensionality reduction aims to yield compact representation of each of sensory input. The second stage tries to synthesize reduced sensory inputs. We apply these synthesized inputs to the Instance-Based Policy Optimization, a sort of Evolutionary Learning.

View full abstract

It is essential technique to accurately localize a signal source in human brains for successful surgery of intractable epilepsy patients. In order to improve the accuracy of the localization, in this paper, we will introduce a localizing method by using Self-Organnizing Relationship (SOR) network. The SOR network was extended based on Self-Organizing Maps by our others and has a feature of realizing an approximation of a desirable input/output relationship of a target system. We think that the feature can be also applied the signal source localizing.

View full abstract

We observed polar excitations on bullfrog sciatic nerves by an optical stimulation using a Holmium YAG (Ho:YAG) laser with the laser power over 2.0 [W]. The laser power was higher the amplitude of the action potential was bigger. However, when the laser power over 10.0 [W], the laser gave damage to the sciatic nerve and the action potential decreased. The laser light did not give heat to the region of laser irradiation in the case of laser power 2.0 [W]. These results show the possibility that the single wavelength laser can use to position estimation of epileptogenic focus and irreversible blocking of propagation of epileptic seizures. We guess that the optical stimulation is evoked to optical energy of a laser light.

View full abstract

Establishing a noninvasive method to identify the epileptic focus is important for the treatment of patients with epilepsy. Near infrared spectroscopic topography (NIRS) is a noninvasive method to measure the regional cerebral blood volume (rCBV) dynamics coupled with neuronal activities. Therefore, optical topography (multichannel near-infrared spectroscopy) was applied for detecting rCBV changes in the epileptic focus. Optical topography was applied to 5 female patients with epilepsy; patients ranged in age from five months to 52 years of age (mean, 13.8). The rCBV dynamics was measured with an ETG 7100. Probes were placed over the scalp, including the putative epileptic focus which was estimated according to seizure symptoms, and electroencephalograms (EEGs) were performed. The rCBV dynamics during the seizures were analyzed for any changes in oxidized hemoglobin. In all cases, rCBV changes were examined in the site where the epileptic focus was estimated to be. In a patient with West syndrome, who was believed to have generalized epilepsy, the rCBV focally increased, and the patient was diagnosed with partial epilepsy. After changes in prescription, the number of patient seizures was markedly reduced. A patient with hemimegalencephaly who underwent a functional hemispherectomy became seizure-free, according to the EEGs and rCBV changes.

View full abstract

Focal brain cooling has distinct power on controlling neuronal excitability through reversibly modulating synaptic transmissions. However, the therapeutic feasibility of focal brain cooling on central nervous system (CNS) diseases still remains unknown. To address the issue, an implantable cooling system was developed and cooling-induced neurological and neurovascular changes were invested in animals and, intraoperatively, in epileptic patients. The results clarified the existence of therapeutic temperatures for CNS diseases and supported the feasibility of a hypothermal device-based therapy.

View full abstract

We propose a subdural electrode array guided by a 0.3mm-diameter SMA guidewire for a minimally-invasive method of electrocorticogram recording. The measured electric characteristics show that the proposed electrodes are compatible with the application of electrocorticogram recording. Somatosensory evoked potential is measured by the proposed method in the animal test in vivo. The results confirm that the proposed electrode array is available for the ECoG recording under a minimally invasive surgery.

View full abstract

Authors aim to describe extensions of human's body, and the event (such as things happened in some space) in Kukanchi. The authors also aim to understand and apply collective process by configurational approach in the informational space. For informational clarification of the extensions of human's body in Kukanchi, the authors used knowledge of Kansei Engineering, Psychology, Sociology, Human Engineering, and so on. By the RT ontology, human and robots can understand each other,and the robot will be able to interact with human, and able to provide the suitable services for human intention. In this paper, the authors introduce the examples of such service support systems in life space.

View full abstract

A structured platform is important to gather, store, transform, and provide informations for realizing the ubiquitous society. Such an environment is called informationally structured space. It is necessary to extract some features from the gathered data in order to obtain the required information in informationally structured space. This paper proposes intelligent sensor networks and a method to laern temporal relationships of sensor for constructing informationally structured space.

View full abstract

We have developed an exit alarm set automatically by cameras mounted on a robot. Our aim is more accurate detection of getting up motion. In a case of using cameras, a problem is that they take only vague infrared light images in a dark room. To resolve the problem, we use information obtained by fine visible light images while the room is light. Our method sets a detection area automatically, after it detects a bed from maps of the room by using visible light images. In the dark room, we use a getting up motion model based on the detection area. Experimental results showed our robot could locate the bed by an 8.8[cm] error. A getting up detection ratio is 95.7%.

View full abstract

The authors propose daily support system for elderly people using an interaction monitoring robot. Daily support system uses the robot to implement exercise support function, fall down sensing function and information presentation function. Exercise support function includes an interaction module which is for encouraging person exercise by action and utterance. Fall down sensing function uses a monitoring module to detect human motion and tells user who fells down. Information presentation function shows the usage condition of tableware and the consumer electronic by using AR.

View full abstract

This research shows the possibility and utility of Kukanchi and mining in time-series logs which supports the system to understand human and help people directly. The authors show an experiment which is constructed based on "Kukanchi". This experiment proves the possibility and utility of using the concept of Kukanchi to construct a system, recognizing human action by mining in time-series motion logs, and the mining results can be improved by using interactive data mining which includes the visualization techniques.

View full abstract

Association rules that express relations between itemsets is effective in the field of data mining such as basket analysis. However, the number of rules tends to be huge for actual dataset, which makes it difficult for an analyst to understand the mining result. To solve this problem, the information visualization is expected to be useful for the analysis. This paper proposes visualizing method that can be applied to large dataset. The proposed method focuses on analyzing the combination of item categories in extracted itemset, and visualizes frequent closet itemsets that are extracted in the previous step of extracting association rules. A prototype system based on the proposed method is implemented, which shows that insights into the combinations of item categories can be obtained. finding concerning the combination of the item categories.

View full abstract

This paper proposes a file management method based on the coordination of folder and tag management models. Most of existing file management systems employ exclusively either folder structure or tag space. The folder management model hierarchically organizes files, whereas tag management model is suitable for flat and flexible information management. Therefore, more flexible and efficient file management would be possible by combining the merit of both models. The proposed method mainly employs folder management model, of which structure can be dynamically modified based on user's operations on tag space. This paper discusses the relationship of the operations for file management between folder and tag management models.

View full abstract

Recently, the number of elderly people increases in Japan, and the development of welfare apparatus having the effect of minimizing care is required. In this research, authors develop Intelligent Power Wheelchair along with minimizing care needs for frail elderly people. This Intelligent Power Wheelchair features the four functions; approach function by human gesture recognition, standing up assistance, walking assistance, and operation assistance by posture recognition. And it shows experiments which three assistances function in useful.

View full abstract

In our laboratory, we have already proposed the method based on the Multi-Layered Fuzzy Behavior Control, which describes behavior decision fuzzy rules for each subtask after dividing complex tasks and acquires the behavior strategy, and the Fuzzy State Division-Type Reinforcement Learning that is able to input the continuous value for the input state. And we have also proposed the Differential Reinforcement-type Shaping Q-Learning applied a general idea of "differential reinforcement" to reinforce the special behavior step by step such as in real animal training. Therefore, in this research, by adopting the concept of shaping to the fuzzy state division type reinforcement learning we propose a method incorporated an automatic Shaping system to make decrease the burden of the trainer which occurs when he gives a Shaping reward. Furthermore, we also propose the method to let it learn the behavior strategy according to the situation for the behavior acquisition of the soccer robot by obtaining the weighted value of attack and defense for the robot by using the fuzzy rule.

View full abstract

In this paper we introduce a novel method of multi dimensional fuzzy set (KFs : Konohen Fuzzy set) with GPGPU. When we can obtain samples those belong same group in some experiments, The set of these samples indicates a fuzzy region in the N-dimentional property space Rn. Making KFs from these data, we introduce 3 step algorithm ; 1) neighbour net which denote topological super surface, 2) density of samples on a metric restricted neighbour net, 3) fuzzy grade interpolation on arbitrary point. We show exprimental results ; compare runtime by using GPGPU with Not using GPGPU

View full abstract

Studies of autonomous and decentralized robot systems have mainly focused on efficient coordination controllers that lead an individual robot contributes to a society or a group which the robot belongs to based on a status evaluation of the society. It is sometimes difficult to design the controller beforehand because of unpredictable disturbances of the environment, condition changing, and so on. Learning approaches for the cooperative behavior in the society are attractable because of their adaptation capability, however, it takes, in general, an unpractical long learning time to reach an acceptable level of social behavior as the result of individual robot behavior learning from scratch. Therefore, it is reasonable to introduce human instructions in order to realize a practical and efficient adaptive controller for the autonomous distributed robots with minimum autonomy. However, conventional human interactions sometimes are not intuitive or effective for the multi-robot systems. In this paper, a human-friendly intuitive instruction system for autonomous distributed mobile robots is proposed. The instruction system includes miniature robots that can be picked up and moved by hands so that the human operators can intuitively instruct cooperative behavior to the autonomous distributed robots.

View full abstract

In this paper, we introduce a novel method to model human behaviors that are described on a multi variable parametric space, using fuzzy set on multi dimensional region. Multi dimensional fuzzy sets which have arbitrary boundaries are powerful describe tool for complex systems. This multi dimensional methodology is called the Konohen fuzzy. Using konohen fuzzy, we show a experimental results for modeling human behaviors.

View full abstract

After total knee arthroplasty (TKA), analyzing the implanted knee kinematics is very considerable problem in the research area of orthopaedics. There are some studies based on 2-D/3-D image registration of 2-D X-ray fluoroscopy images and 3-D geometrical model of knee implant. However, these conventional studies are classified into statics image analysis, and there are a few studie for dynamic images analysis with continuous knee movement. In addition, the conventional studies have a problem of falling into local maxima in high-dimensional search space. And the estimated result did not consider continuous knee movement. This paper proposes a method for analyzing continuously implanted knee kinematics based on particle filter. The proposed method searches 3-D pose position parameter using multiple candidate parameters by particle filter, and estimates 3-D pose position of the implanted knee joint using 2-D/3-D image registration. The experimental results showed that the proposed method analyzed the continuous movement of the knee joint and estimated the knee joint angles.

View full abstract

Aimed at face posture estimation and tracking in real time from image sequence captured by camera, we have proposed a method using state space model for estimating face direction and position with simplified flat face and two eyes model. Real-time implementation of state estimation for the model has been done with particle filter algorithm. Experiments with two obvious marks, picture face, face photo, and real face have demonstrated performance of the proposed method.

View full abstract

Particle filter is latest method for state estimation in state space model. Fundamental of state space model, state estimation, and particle filter are described with definitions of mathematical notations in the followings. Latest CPU technology allow us to use many-core CPU and parallel computation on it. In this paper, we have extensively paid our attention to many-core CPU to implement particle filter algorithm.

View full abstract

Music recommendation system using the content based music retrieval plays an important role of multimedia information retrieval. Clustering songs using music features is a basis of music informaion retrieval. However the parameter of a clustering algorithm and music features have not been studied yet. We investigate the various numbers of music features and compares with humans' sensitivity by subjective experiment for 200 MIDI music. The result will be the foundation for the future songs retrieval systems.

View full abstract

This paper report the basic inspection which used an image processing system and making system hardware about the possibility of the application through the implementing experiment. This system is one kind of the example using FPGA. In recent years, a lot of practical example is already reported and the high-speed and real-time processing is progressing for a manufacturing technology by using FPGA. Also, this type system is used at the production site in a lot of companies. The engineers have to design the image processing hardware by the computer program and the technical knowledge about the image processing design. This system research would support those engineers for the future. Therefore, using evolutional computation is new approach about computing of the best control parameter as offline. Finally, this paper discusses the possibility of the application of the manufacturing technique field using our technique and the linear sensor camera for huge digital image process.

View full abstract

A classification algorithm for blood gas test is proposed based on acid-base balance in integrated medical information system. It provides degree of each balance disorder using not only blood gas test value but also other test item in other medical information system based on fuzzy inference. Classification results for 11936 data taken in clinical laboratory of Kochi Medical School Hospital show that healthcare professional in clinical practice evaluates the results as the same as knowledge of clinical practice. The proposed system is a basic system of diagnostic support in evaluation for clinical examination.

View full abstract

Visual key image retrieval system has been proposed, which depends on a part of images. In the conventional research, the similarity used for constructing clusters was defined by linear combination of Euclid distance of each feature and the retrieval precision is 10%. Linear sum has been used without being discussed enough about the distance function used to calculate the degree of similarity and the weight has been decided empirically. Similarity suitable for person's sensibility can be constructed by changing the combination of the feature vector and the operator more appropriately, and the distance function which is better than conventional techniques can be obtained. However, it takes a great amount of time to search the combination of operators suitable for person's sense, and global optimum is difficult to be derived from among the combination of huge number of operators. In this research, genetic programming is used to search the suitable combination of the aggregation operators and feature vector. Compared with conventional technique, the recall ratio is 8% decrease and the relevance ratio is 19% increase and the F-value increase by 7%.

View full abstract

In the present study, is it accepted with familiarity we in men if acting very when the robot will try to merge in life of us of the acceptance of the company etc. In the near future, and the influence that the movement of the robot gives the person by using the Semantic Differential method is clarified.

View full abstract

The research on fuzzy control is various and performed. Fuzzy control is used for the control methods, such as automatic conveyance vehicles, such as a line trace robot and a factory, in many cases. It is verified whether fuzzy control is applicable to an object with a fast running speed like line trace car. In this research, the simplified reasoning method is used for the steering angle and the motor control of the line trace car. We tried to make smoothly controller without carrying out a zigzag run. About line tracer system, the car calculates a steering angle from the value of the sensor read during the run. Then, the line trace car calculates a motor output from the outputted steering angle.

View full abstract

Proposal of landing operation for biped robot.

View full abstract

This experiment is acquisition the motion data of human-beings by using a motion capture system. This purpose is experiment and verification if motion of robots is making with motion data of human-beings.

View full abstract

Brain-Computer Interface (BCI) is an interface by which the human can directly operate the machinery using brain activity signals measured noninvasively. In this study, a BCI is developed based on the brain bloodstream change measured by Near-Infrared Spectroscopy (NIRS). So far we have processed the brain bloodstream changes during silent reading by the distance-type fuzzy reasoning method to get the BCI control signals. However, the fuzzy rules could not completely reflect the features of NIRS signals, which reduced the recognition ratio. Therefore, we analyze the signal characteristics of cerebral blood flow during silent reading in order to obtain more proper fuzzy rules. We measure the cerebral blood flow changes on both sides of the prefrontal cortex during silent reading task 4 times a week for 4 weeks and extract the peculiar cerebral blood flow change patterns as knowledge for reasoning.

View full abstract

The study proposes an autonomous control system based on the distance-type fuzzy reasoning method to realize the autonomy of robots by imitating the human behavioral strategies, which is extracted from the relationship between the human manipulation to the robots and the circumstances where the robots are in. We develop an experiment platform on an omni-directional mobile robot with a flat board. There is a target on the board and the task is to move a ball to the target by operating the robot. The human behavior strategy is extracted from the relationship between the human operation and the position of the ball. We attempt autonomous control of the robot based on the extracted behavior strategy using the distance-type fuzzy reasoning method.

View full abstract