[in Japanese]

View full abstract

As well known as "Cocktail party effect," the binaural hearing assists human being to listen the specific sound selectively and to segeragatate of it. Frequency domain binarual model (FDBM hereafter) is a computer model of binaural hearing realized in the frequency domain based on FFT. In order to realize the selective listening for hearing impared, a binaural hearing assistance system utilizing FDBM is developed with howlling control by a novel howlling canceller. In this paper, the basic algorithm and its implementation of the new binaural hearing assistance system is presented.

View full abstract

In the actual measurement, the acoustic signal is often measured under the mixture of several noise sources. We propose a probabilistic estimation method for extracting the objective noise by using Independent Component Analysis. The effectiveness of this method has been experimentally confirmed by applying it to the actual acoustic data.

View full abstract

A novel blind source separation of a mixture of two or more voice signals has been proposed in the present paper. The proposed BSS has been focused based on the spatio-temporal gradient analysis which utilizes the linearity among, 1:sound pressure of source signals, 2:particle velocity vector and 3:instanteneous strain tensor of the medium of the observation point.

View full abstract

A particle filtering algorithm using the parameters in the EM (Expectation-Maximization) algorithm is proposed for tracking multiple sound sources. Differently from the conventional EM based algorithms, the proposed algorithm can track multiple sound sources without knowing their starting points. Moreover, an idea of the group tracking is applied to the particle filtering algorithm so that better tracking performances can be obtained. Experimental results show the validity of the proposed algorithm.

View full abstract

[in Japanese]

View full abstract

In this paper, simplified calculation method of kurtosis is introduced for the analysis of impact vibration with one sided affiliated impact vibration which occurs towards the progress of time. One sided affiliated impact vibration is approximated by one sided triangle towards the progress of time and simplified calculation method is introduced. Furthermore absolute deterioration factor is introduced.

View full abstract

In order to make machine diagnosis, the method of calculating Kurtosis or Bicoherence was utilized. Calculating system parameter distance was also utilized applying time series data to AR(Auto Regressive) model or ARMA(Auto Regressive Moving Average) model. In this paper, simplified calculation method of machine diagnosis techniques is introduced utilizing system parameter distance under the application for AR model. Furthermore, absolute deterioration factor such as Bicoherence is introduced.

View full abstract

So far, many methods for machine diagnosis techniques were proposed. In this paper, principal component analysis is adopted to evaluate the machine deterioration condition totally based on the plural deterioration indices. Furthermore, multiple correspondence analysis is applied and simplified diagnosis method is proposed based on this. Data are well grouped for each deterioration condition which would be utilized as a total index of deterioration.

View full abstract

[in Japanese]

View full abstract

When images have complex features, the single neural network model can't often process them with enough accuracy. This report proposes image processing methods (compression and segmentation) by using multiple multi-layered neural networks. Moreover, we show the characteristics of the proposed methods from simulation reults.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

The obejective of this paper is to exprail structural learning for neural networks: a back-propagation type hierarchical neural network and a meta-control bolzman machine. These strauctural learning neural networks are employed im problems of financial engineering.

View full abstract

As one of the most important technologies to support the coming aging society, the development of human care robots is attracted by many researchers. In this presentation, we study the robot's control problem for the safe human-robot interaction. We propose full body impedance control so as to distribute the interaction force between the robot and human during care tasks.

View full abstract

This study deals with human locomotion based on the planar musculoskeletal model and physical kinematics. The skeletal model consists of a trunk, a pair of arms composed of two links, and a pair of legs composed of three links. Each leg has nine principal muscles modeled as massless wires whose contraction forces and activations are determined using lambda model. Numerical simulations establish stable bipedal locomotion compared with physiological data of human locomotion.

View full abstract

The foot rolling motion, including heel strike and toe off, plays an important role in generating bipedal locomotion. This study investigates dynamical influences of this motion based on a simple walking model in which a circular arc is attached to the leg tip. In particular, we achieve approximate periodic solutions of the walking motion and examine its stability using Poincare map.

View full abstract

Ankle joint compliance plays important role in humans' biped walk. On the other hand, a few researches have referred about the compliance of the biped robot. This paper reports that the physical bipedal walker with the compliant ankle joint is developed and that the different joint compliance provides different cyclic locomotion. In order to obtain the joint compliance, antagonistic pneumatic muscles around the joint is employed.

View full abstract

There are many researches on biped passive dynamic walking. As the results, many interesting results have been discovered. For example, the stability or the bifurcation phenomenons of P.D are discussed. But nobody has shown the existence of the 4-legged passive dynamic walking. In this paper, we developed a 4-legged passive dynamic walking machine and carried out the experiments. As the result, we show the existence of the phenomenon of 4-legged passive dynamic walking. Also, 6-legged passive dynamic walking will be considered.

View full abstract

It reports on the result of examining the composition of the operation screen in the mobile robot that uses the step-on-interface (SOI) that is newly proposed to operate the robot.

View full abstract

It reports on the result of achieving various operations in the mobile robot that uses the step-on-interface (SOI) that is newly proposed to operate the robot.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

It reports on the result of experimenting and evaluating the effect and the feature of four mobile robots equipped with the function to display the forthcoming operation in an interpersonal simulation environment.

View full abstract

It reports on the result of examining the content to display on the omni-directional display in the mobile robot that preliminary-announces the forthcoming operation.

View full abstract

Recently, lots of people have interests on the stock trading because on-line stock environments has been developed. On the other hand, almost every research of forecast of stock prices have been based on the long term investment, there are few researches which focus on the day trade. In this paper, authors proposes strategy evolution of day trade agent by means of the day trade agent framework. Authors also proposed the system which can replay the real market. The adaptation ability of the proposed method is confirmed by applying the agent to real market.

View full abstract

[in Japanese]

View full abstract

We have proposed a generation alternation model for co-evolutionary optimization of strategies in one-on-one competitive contests. In this paper, we discuss the potentials of the model, through the application to several contests, including Nim and the two-dimensional hockey games. The experimental results show that the proposed model performs better over the conventional models.

View full abstract

Evolutionary Computations(ECs) are the optimization methods based on the mechanism of natural evolution. Although several studied have been reported on analysis of ECs, there are few researches which study an operator based analysis. In this paper, the authors proposed the novel analysis method which based on operators. Computer simulation is carried out to show the advantage of proposed method. We also proposed a novel GUI tool to support GA analysis.

View full abstract

[in Japanese]

View full abstract

This paper proposes rank scoring of relay persons in social network. Relay persons are attractive in open innovation environment. In order to find relay persons, the method from the change of average shortest length is available. However it takes a very long time to find them in massive network. Therefore, we proposed a new scoring from consideration that relay persons have less links than not relay actors. Finally, We discuss the feasibility of our methods from experiments.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

Branch and Bound method is one of the optimization techniques of the BMI problem. In this research, we proposed the new method about the method of dividing the area and inspected the calculation efficiency. Moreover, we implement the algorithm of parallelized Branch and Bound method by C++ language. And numerical experiments performed on Beowulf cluster system.

View full abstract

We present a method for designing optimal smoothing splines with equality and/or inequality constraints. The method has a wide variety of applications such as path or trajectory planning problems. The splines are constructed by using normalized uniform B-splines as the basis functions, enabling us to obtain the numerical solution relatively easily. We examine the performance by numerical examples.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

This paper deals with model following control of input/output linearizable nonlinear sampled-data systems.When a digital computer is used as a controller,an emulation technique is usually applied.However,it sometimes leads to problems of instability or poor control performance.This study shows a design method of a model following controller by a digital redesign approach with Euler approximate model and a Lyapunov function on the basis of the idea of the sampled-data controller design by Nesic et al.

View full abstract

This paper describes a design scheme of performance adaptive PID controllers for nonlinear systems. This control system is constructed by data-driven system identification and model predictive-based PID controller. The user-specified parameter included in the criterion of the model predictive controller is adjusted in consideration of the control performance assessment. The behavior of the newly proposed control scheme is examined by a numerical simulation example.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract