We have designed and implemented the “Virtual Laboratory” systems on the computer systems. Some of the implemented systems are Web-based learning environments where a server/client model is adopted. The subject of electronic circuits is the most essential one in the departments of electrical and electronic engineering. However, it is a difficult subject for students to understand the basic therory, technical term and electronic circuit design. In this paper, we will describe the remote measurement system for the learning support of electronic circuits using RTLinux.
Electromagnetic disturbances for vehicle-mounted radios are well known to be caused mainly by conduction noise currents flowing out wire harnesses from printed circuit boards (PCBs) having a common ground layer with slits. In this study, in order to investigate how ground-layer slits affect the above conduction noise currents, we paid FM band induced voltages or crosstalks on the trace connected to the wire harnesss, and simulated with the FDTD method the crosstalk levels between two traces perpendicularly fabricated on three kinds of simple PCBs with different ground-layer slits, which were compared with measurement in the frequency range from 10 MHz to 1 GHz. As a result, we could confirm that the FDTD calculation approximately agrees with the measured results, and also that the crosstalk levels do not always increase with the slit number, which can be reduced by the slit layout.
Performance of a brain machine interface (BMI) critically depends on selection of input data because information embedded in the neural activities is highly redundant. In addition, properly selected input data with a reduced dimension leads to improvement of decoding generalization ability and decrease of computational efforts, both of which are significant advantages for the clinical applications. In the present paper, we propose an algorithm of sequential dimensionality reduction (SDR) that effectively extracts motor/sensory related spatio-temporal neural activities. The algorithm gradually reduces input data dimension by dropping neural data spatio-temporally so as not to undermine the decoding accuracy as far as possible. Support vector machine (SVM) was used as the decoder, and tone-induced neural activities in rat auditory cortices were decoded into the test tone frequencies. SDR reduced the input data dimension to a quarter and significantly improved the accuracy of decoding of novel data. Moreover, spatio-temporal neural activity patterns selected by SDR resulted in significantly higher accuracy than high spike rate patterns or conventionally used spatial patterns. These results suggest that the proposed algorithm can improve the generalization ability and decrease the computational effort of decoding.
This paper proposes a design scheme of bilateral teleoperation of the constrained system with time delay. The stability of bilateral teleoperation with PD control is proved by the proposed Lyapunov function. For the bilateral teleoperation with constraints, a command governor is provided using the model prediction. By using the command governor, the bilateral teleoperation can avoid constraints violations. In the command governor, the proposed algorithm can decrease the computational cost in comparison with the quadratic programming. In the simulation and experiment, the performance of the proposed bilateral teleoperation system is verified using a single-degree of freedom master/slave system.
In order to decrease human stress, Animal Assisted Therapy which applies pets to heal humans is attracted. However, since animals are insanitary and unsafe, it is difficult to practically apply animal pets in hospitals. For the reason, on behalf of animal pets, pet robots have been attracted. Since pet robots would have no problems in sanitation and safety, they are able to be applied as a substitute for animal pets in the therapy. In our previous study where pet robots distinguish their owners like an animal pet, we used a puppet type pet robot which has pressure type touch sensors. However, the accuracy of our method was not sufficient to practical use. In this paper, we propose a method to improve the accuracy of the distinction. The proposed method can be applied for capacitive touch sensors such as installed in AIBO in addition to pressure type touch sensors. Besides, this paper shows performance of the proposed method from experimental results and confirms the proposed method has improved performance of the distinction in the conventional method.
In this paper, a new H∞ controller reduction method which considers to maintain the H∞ property is proposed. The generalized H∞-controllers are able to described using free-parameters which are stabe and bounded. If a low order controller can be written with a free-parameter which satisfies the stability and the bounded condition, the low order controller become an H∞-controller. In order to satisfy the conditions, we employ an iteration method which contains the D-scaling matrix calculation and the balanced model reduction. Some numerical examples are illustrated to show the effectiveness of our method.
In this paper, brain functional mapping method by hierarchical decomposition analysis (HDA) is proposed. HDA is one of the multi-dimensional AR modeling methods and well-known for its validity to detect temporal lobe seizures. The author transforms the estimated AR model in the form of transfer function from the inner blood flow signal to the cerebral cortex. The signal for HDA is oxidized hemoglobin density HbO, which is measured by near infrared spectroscopy (NIRS). Comparing the 2 tasks which use arithmetic sense, the difference of brain activity becomes clear by proposed technique.
This paper focus on design of control schemes for multi-color printing press which is a typical web handling system. In the printing press consisting of multi printing units, the dynamics between two neighboring printing units are coupled, and the coupling effects propagate to downstream units with delay. This paper presents a model-based nonlinear control strategy for printing positioning control in a three-color printing press. First, a nonlinear coordinate change is introduced to a state space model. Then, based on the new dynamical model, two feedback control schemes are proposed. A state feedback controller is designed such that tension error and printing error converge to zero. The state feedback controller requests the tension of the web to be measurable. In order to avoid the constraint, an observer-based feedback controller is proposed with stability analysis. Furthermore, the effectiveness of the proposed control scheme is demonstrated through simulation.
In this paper, we propose a self-organized learning model that can generate behaviors for successfully performing various tasks. The model memorizes various relationships between changes in a state pattern and a motor command through learning. After the learning, the model can perform various tasks by generating the various behaviors automatically. We confirmed the performance of the model by applying it to a mobile robot simulation. The results indicate that suitable behaviors for all the tasks generated spontaneously. Additionally, we propose a sequential learning method which modifies the memorized various relationships while the model executes the task. And we confirmed the effectiveness of the sequential learning by the simulation.
Three dimensional medical image registration is a fundamental technique which applied various medical treatments such as image diagnosis, treatment planning, image guided surgery, etc. In radiation therapy for cancer treatment, quick alignment method around target lesion is required. To align CT images, similarity measurement like normalized cross correlation or mutual information should be calculated. The computational cost must reduce for achieving the online image registration. In this paper, we propose a novel quick rigid registration method which makes it possible to align the three dimensional images using the parametric eigenspace method. By projecting each CT slice image into the eigenspace as a set of low dimensional vectors, image similarity can be calculated very rapidly. The experiments using CT images of the same patient, it is found that the alignment accuracy is almost the same as the method using normalized cross correlation, and the computation time is less than one second.
We have developed an eye-gaze input system for people with severe physical disabilities, such as amyotrophic lateral sclerosis (ALS) patients. This system utilizes a personal computer and a home video camera to detect eye-gaze under natural light. The system detects both vertical and horizontal eye-gaze by simple image analysis, and does not require special image processing units or sensors. Our conventional eye-gaze input system can detect horizontal eye-gaze with a high degree of accuracy. However, it can only classify vertical eye-gaze into 3 directions (up, middle and down). In this paper, we propose a new method for vertical eye-gaze detection. This method utilizes the limbus tracking method for vertical eye-gaze detection. Therefore our new eye-gaze input system can detect the two-dimension coordinates of user's gazing point. By using this method, we develop a new support system for mouse operation. This system can move the mouse cursor to user's gazing point.
In this paper, we propose a method to detect the waving drive from a single camera in real-time. The requirement for the proposed method is computation time, accuracy and the dependency on the lane marker. The previous waving drive detection systems are based on the detection of the lane marker. Thus, the previous systems can't work on a road with no lane marker, like a snow road. The proposed method uses the planar region in front of the vehicle by converting it to a top-view image. The system is able to work on a road without a lane marker. To resolve the problem of the computation time, we apply the particle filter to speed up the proposed method. The proposed method is tested on both synthesized and real image to check the accuracy and computation speed. The experiment shows that the proposed method works on both roads with or without lane marker. Also, an online experiment is done to demonstrate that proposed method works in real-time.
Recently, failure detection by acoustic data is applied to quality check such as new or used bill classification. However, it is difficult to reduce the noise in acoustic data of bill by conventional techniques since there are many types of noises in the environment. In this paper, we propose a method by using an adaptive digital filter and a neural network to reduce noises in acoustic data of bill. First, the acoustic data is divided into some frequency bands by the wavelet transform. Next, each data is passed through two types of filters to reduce the noise. One is an adaptive digital filter trained by the LMS algorithm, and the other is a nonlinear filter based on a neural network trained by the error back-propagation method. Then the processed data in each frequency band are composed to get an estimated signal. Finally, simulation results are illustrated to show the effectiveness of the proposed method compared with conventional approaches.
In general, meta-parameters in a reinforcement learning system such as learning rate are empirically determined and fixed during the learning. Therefore, when an external environment has changed, the sytem cannot adjust to the change. Meanwhile, it is suggested that the biological brain could conduct reinforcement learning and adjust to the external environment by controlling neuromodulators corresponding to meta-parameters. In the present paper, based on the above suggestion, a method to adjust meta-parameters using the TD-error is proposed. Through computer simulations using maze problem and inverted pendulum control problem, it is verified that meta-parameters are appropriately adjusted according to the amplitude of the TD-error.
Policy gradient methods are useful approaches to reinforcement learning. Applying the method to behavior learning, we can deal with each decision problem in different time-steps as a problem of minimizing an objective function. In this paper, we give the objective function consists of two types of parameters, which represent state-values and environmental dynamics. In order to separate the learning of the state-value from that of the environmental dynamics, we also give respective learning rules for each type of parameters. Furthermore, we show that the same set of state-values can be reused under different environmental dynamics.
The business in the enterprise is closely related with the information system to such an extent that the business activities are difficult without the information system. The system design technique that considers the business process well, and that enables a quick system development is requested. In addition, the demand for the development cost is also severe than before. To cope with the current situation, the modeling technology named BPM(Business Process Management/Modeling)is drawing attention and becoming important as a key technology. BPM is a technology to model business activities as business processes and visualize them to improve the business efficiency. However, a general methodology to develop the information system using the analysis result of BPM doesn't exist, and a few development cases are reported. This paper proposes an information system development method combining business process modeling with executable modeling. In this paper we describe a guideline to support consistency of development and development efficiency and the framework enabling to develop the information system from model. We have prototyped the information system with the proposed method and our experience has shown that the methodology is valuable.
Remotely sensed images observed by synthetic aperture radar (SAR) become to be more significant for a variety of purposes. It is necessary to process and focus SAR images at a user's side for required precision. This paper presents a concept, a system structure, and an implementation method for a web application system with a SAR image processing for education. The system employs Ajax technology in consideration of current computer and network systems. This system has three software components, a SAR server, an HTTP server, and a web browser in each terminal computer. The SAR server is the main component for managing the SAR database and processing all SAR data sets. The range-Doppler method is adopted as a focusing algorithm. Three displaying measures are newly introduced since the SAR images are sets of mass complex number pixels. Operation of SAR image, sample processed images are shown to exhibit actual examples of processing ALOS PALSAR (Advanced Land Observing Satellite Phased Array type L-band SAR) data. From step-by-step processing results, the proposed web application system was demonstrated to provide a useful graphical user interface (GUI) and to produce the SAR image with adjusted parameters by using dialogs.
In this paper, we present a new robust digital watermarking method by applying a conventional fragile watermarking method. Since the candidate pixels for embedding are selected from among the smooth image area having a good tamper recovery rate, a good detection rate of watermarked information is provided. Our method also produces a high PSNR of the recovered image by improving the accuracy of the prediction value for the recovery against the tampered pixel. We demonstrate the experimental results concerning the quality of the watermarked image and the watermark detection rate in the case of embedding 8000 bits.
Up to now, a speaker identification method using MFCC spectral ratio between air-transmitted speech sounds and bone-transmitted speech sounds, has been proposed. However, the recognition accuracy rate of this method which is applied to many speakers, has not been shown. This paper shows that the proposed method is useful to identify many speakers using MFCC spectral ratio by the experiment with 70 speakers.