In geographical information processing approximating the original surface from contour line data information is achievable through triangulation. Surface approximation by triangulation is treated as a combination problem of triangular elements in a polygon approximation of the original data. Conventional triangulation methods tend to be sequential, deterministic and triangulation decisions are made on localized evaluations disregarding the global picture. In this paper a novel and simple contour data triangulation scheme proposed by the authors is used to triangulate digitized contour data from a geographical map. The scheme is relatively technique independent and triangulates contour line data in a parallel manner.

抄録全体を表示

The purpose of our research is to image temperature distribution inside the human head by non-invasive measurement. To achieve this goal, we investigate a method to image inhomogeneous permittivity distribution by measuring capacitance, and in addition to measure the temperature distribution using a permittivitytemperature characteristic. This paper describes the first step in estimating inhomogeneous permittivity distribution using our proposed new capacitance measurement method. We present the results of experiments performed using the proposed method on a homogeneous phantom containing agar-agar. As a result, we estimate that the proposed method is functional for the imaging of inhomogeneous permittivity distribution, although some of the problems in this measurement procedure and image reconstruction system are remained to be solved.

抄録全体を表示

In this paper, we examined the safety during the wheelchair marathon. By monitoring the cardiac function and the metabolic function at the same time through the blood and biochemical labo-data analysis, electrocardiogram, Holter's electrocardiogram and RR time series analysis and echocardiogram in the subjects at the wheelchair marathon. Heart rate increased significantly from 66.0±12.6 to 99.0±10.8 beat/min. Systolic blood pressure increased significantly from 117.7±18.6 to 143±12.6mmHg. Diastolic blood pressure decreased significantly from 70.0±10.4 to 45.0±8.8mmHg. No changes in left ventricular function and size on the echocardiogram. The findings of dehydration were not recognized after the race. The increase in plasma TAT complex concentrations was significantly greater from 2.2±0.4 to 4.0±0.6ng/ml. WBC was significantly increased from 7.0±1.9 to 13.3±5.6×10³/μ1. No other measurement items significantly changed after the race. The findings of arrhythmias. (supraventricular premature count, ventricular premature count, or other block etc.) were not significantly recognized during the race. In this paper, it is possible that these subjects who have hardly atherosclerosis, if they enter the race after full training and with supplying water, do safely without the bad influence to the heart. But old age athlete with atherosclerosis should be done full medical check before the race in the future.

抄録全体を表示

Most recently, magnetic stimulation has been used for clinical application or rehabilitation instruments. Nobody has well known the effects of magnetic stimulation on the muscle fatigue. To suppress or ease the muscle fatigue, it is important to make clear a mechanism of fatigue and effects of magnetic stimulation.
In this paper, we measured an electromyogram, when the arm or the head is stimulated by magnetic stimulation. In the case of the peripheral fatigue, we measured the amplitude of EMG. For the central fatigue, we measured the duration time of muscle discharge. As a result, the change of EMG is related to the muscle fatigue. Magnetic stimulation is useful to suppress the muscle fatigue.

抄録全体を表示

The method which separates vessels of solid crossing and recognizes each of them on a coronary cineangiogram is proposed in this paper. This method is constructed with the two steps which are region extraction and edge tracing of vessels. On vessel region extraction, the regions including level crossing areas are extracted by the edge preserved smoothing, the dynamic thresholding and the region growing methods. The vessel edge tracing is applied for the selected level crossing areas, and each individual vessel region is separated. The experimental results show the good ability of vessel separation by the proposed method.

抄録全体を表示

The immediate change in heart rate after the change to the upright posture is an indicator of autonomic neuropathy. The subjects are loaded with orthostatic stress. In this paper, we analyzed the correlation between the impulse response function derived from the time series of the R-R interval before postural change to the upright posture and the immediate change in heart rate after a change to the upright posture. We analyzed the postural changes from the supine to the standing position and from the sitting to standing position of healthy men, the postural change from the sitting to the standing position of a healthy young pregnant woman and the postural change from the squat to the standing position of college athletes. The impulse response function derived from the R-R interval time series before the postural change was closely correlated to the decrease in R-R interval with standing. The impulse response function could indeed serve a measurement without orthostatic stress.

抄録全体を表示

We previously reported a method for automated detection of ventricular contours in cardiac magnetic resonance (MR) image. However, in the case that the gray-level difference between the ventricles and surrounding tissues is extremely obscure, this technique may not work well. To overcome the shortcomings, in this paper we present an improved algorithm to enable us to extract the contours of ventricles from cardiac MR images having extremely low contrast. The technique is based on our previously reported algorithm by incorporating a 2-step preprocessing procedure. Namely, we first automatically determine the gray value of the boundary between ventricles and surrounding tissues, and then enhance the location of edges to make the boundary clearer. Experimental results confirmed the superiority of our improved method over the previous technique and other commonly used methods.

抄録全体を表示

There are two method for the purpose of diagnosing medically by using gamma-ray light irradiation image.
One is to use of the scintillation camera for gamma-ray, the other is to use of the photostimulable luminescence point by the secondary excitation of the image plate (IP) system for X-ray.
The standpoint of the spatial resolution at the total medical image, using gamma-ray, the first can get the image on a short time, but the first is a poor image quality, and the second is good image quality, but the second can the image on a long time, becaus of insenstive to gamma-ray.
We report on the improvement for IP's week point by our proposal method, and by our clinical and quantative analysis data, to use the highly effecient IP (ST-III).
We make the improvement on the imaging time (from 30 minutes to 20 minutes), and the inprocessing time (from 33_??_50 minutes to 27 minutes) for a former method on an organnism.
We strongly believe that our convenience improvement method, and our clinical quantative analysis data can contribute to the wide application as well as the quality up for the clinical diagnosis to use gamma-ray.

抄録全体を表示

Two types of studies for inverse systems (or inverse model) are developed by Silverman and by Sain. One is a type of studies to construct the inverse systems which make an identity transfer function matrix of tandem union with original system, the other is a type of studies to construct the inverse systems which make a transfer function matrix integral value of identified matrix. Both require differentiators to be realized. On the other hand, filtered inverse systems which are restricted pass band do not require differentiators. These inverse systems have been proposed by Yoshikawa using transfer function matrix and by Yamada using state space representation. In this paper, a design method of exact inverse model for a given model using norm criterion for differece between the model and plant is proposed. And applying this inverse model to Model Feedback Control System (MFCS), it was shown by the numerical simulations that the adverse effect of disturbances can be reduced and robust stability can be achieved.

抄録全体を表示

For unbounded Jordan domains the new scheme of approximations and the distribution of charge points for numerical conformal mappings in the charge simulation method have recently been proposed by a potentially theoretical consideration⁽¹⁾.⁽²⁾. Using the distribution of charge points and a modification of the scheme, the method will be applied to numerical Dirichlet problems of ‘unbounded’ Jordan domains. Typical examples will show that numerical results of high accuracy may be obtained by the method.

抄録全体を表示

We propose a new IGBT structure with a new N⁺buffer, and confirm by experiments and numerical simulations that the new IGBT is superior to the conventional IGBT.
The following results were obtained.
(1) According to our experiments, the new IGBT was able to decrease the total power loss, and the parallel operation became easier, compared with the conventional IGBT. Moreover, the short circuit ruggedness of the new IGBT was almost the same as that of the conventional IGBT by optimizing the ratio of the N⁺⁺buried layer.
(2) We clarified why the characteristics of the new IGBT was improved by numerical simulations.
(a) When the new IGBT is on, holes injected from the P⁺substrate flow through, keeping out of N⁺⁺buried layer. And the holes rapidly turn around the N⁺⁺buried layer when passing by, and hole concentration becomes even. Because the lifetime of the new IGBT is designed long, the hole concentration of the new IGBT increase in the N-layer. Therefore, the saturation voltage of the new IGBT is lower than that of the conventional IGBT
(b) Since the lifetime of the N^-layer of the new IGBT can be extended, as a result, the temperature dependence of the lifetime becomes small, and IZTC of the new IGBT is improved.
(c) In the turn-on state, the holes is injected through the N⁺buffer layer with lower concentration from the P⁺substrate, thus the turn-on speed of the new IGBT become quicker and the turn-on loss of the new IGBT is reduced.
(d) In the turn-off state, as the N^-layer is depleted completely, the carriers in N⁺buffer layer mainly influence on the tail current. There are little carriers in N⁺⁺buried layer of the new IGBT, so the turn-off loss of the new IGBT is reduced.
(e) Since the effect to prevent the holes being injected from P⁺substrate affects N^-layer, the amount of carriers in the N^-layer of the new IGBT is limited in the saturation current region. Therefore, the saturation current is also controlled, and the short circuit ruggedness of the new IGBT is not deteriorated.

抄録全体を表示

Waveform distortion at signal edge generally falls into two categories: hysteresis type and non hysteresis type. Non hysteresis type distortion has a flat portion in signal transition, but it will not give any serious problem except for timing oriented circuits. While, hysteresis type distortion which makes round trip between two threshold levels will affect logic judgment. Incorrect information will be transmitted if signal is seriously distorted. It is important to understand relationship between spectrum modification and waveform distortion.
In this paper, an occurrence mechanism for the output waveform distortion of CMOS logic ICs are analyzed and improving method for them are proposed. Waveform distortion occurs when series resonance circuits are composed at load side of driving device. Relationship between rising time/period [%] and limit frequency at which waveform distortion begin to appear will be made clear. There are two method to improve waveform distortion. One is to change spectrum distribution of rectangular wave by modifying rising time. The other is to change resonance frequency by modifying such parasitic elements as inductance (pattern length) and capacitance (amount of ICs).

抄録全体を表示

A design of low-frequency current-mode continuous-time filters for low-voltage and low-power appli-cations using complementary current mirrors is presented. When we realize low-frequency current-mode active filters, the time constant RC becomes quite large which can not be realized by passive elements in a monolithic integrated circuit. In this paper, negative feedback technique is used to realize the low-frequency current-mode filters. Since the filters consist of integrators as active elements, the proposed method can be used for a wide range of application. The filters can operate with a 1.5V power supply, thus they can be applied for portable equipments which are driven by a single battery. A third-order leapfrog filter with the castoff frequency of 8kHz is designed as an example and simulated by SPICE using a standard bipolar parameter set. The results show good performance in frequency characteristics.

抄録全体を表示

Today, information obtained a medical image is important for a diagnosis, such as X-ray C. T., M. R. I., ultrasonic image. Especially, ultrasonic image has the merits of the simplicity of the machine, low cost, and safety. On the other hand, ultrasonic image is sometimes poor in quality because of much noise. Therefore, the diagnosis by ultrasonic image has been depended on the judgement of skilled doctors. It is necessary to develope a method of an automatic diagnosis that is based on a qualitative evaluation. In this paper, we present a novel method of the regional extraction by using a statistical character of the echo speckle pattern in ultrasonic image. The discriminating function is defined to recognize the similarity of the statiscical character. It is based on a multi-dimensional autoregressive model. Using the proposed regional extraction method, it is made possible to discriminate and extract plural number of regions in one image. We also propose the developed method to extract the border of the tissue region, that is an application of the regional extraction method. Finally, we have evaluated those methods by analysing the ultrasonic image including the region of the tissue in human body and the simulated data.

抄録全体を表示

In recent years, a very high-speed divider is required in real-time applications of digital signal processing and robot control and so on. In this paper, a high-speed cellular array divider with selection function is proposed, which is based on the non-restoring algorithm and can deal with both fixed-point and negative operands in two's complement form. This divider uses new techniques which can generate in parallel both a quotient bit of one row and a partial remainder and CLA bit of its next row. Moreover the delay time of the proposed divider is calculated in terms of a delay of one unit such as NAND gate. Finally, by using PARTHENON, a CAD (Computer Aided Design) system for VLSI, this divider is designed and evaluated. As a result, elimination of delay time for even rows becomes possible. Thus, the delay time can decrease by approximately one-half of the high-speed divider proposed by Cappa and Hamacher which uses the most general high-speed techniques of carry-save and CLA.

抄録全体を表示

This paper proposes a new construction method of neural networks. The construction method consists of two fundamental ideas, which are a parallel selection-style evaluation and rules evolution.
This paper first explains these two ideas and combines these two ideas into one design method for neural networks. The neural network constructed by this proposed method is called emergent neural network.
An emergent neural network is applied to moving obstacles avoidance problem and its performance is demonstrated.

抄録全体を表示

A blur correction method of license plate images on running vehicles taken with traffic monitoring cameras is proposed. Since the exposure time of the traffic monitoring cameras is generally 1/30-1/60 s, the taken images are degraded by the motion blur of running vehicles. In the proposed method, the parameters of the motion blur are estimated from a taken image sequence, and then the motion blur is corrected with Wiener filter. The images which are restored by the proposed method demonstrate the efficacy of the proposed method.

抄録全体を表示

For constructing an intelligent multimedia database system, we have been focusing on an intelligent retriever using keywords which are defined as abstract descriptions of archived multimedia data. We proposed a front-end database preprocessor called “keyword-network” where multiple keywords are linked together beyond these semantic relationship and its structure will be updated by users' retrieval operations.
Although this model could be applied for a distributed database system, in this paper we propose an extended idea called “distributed keyword-networks”. This is an inter-networking model between multiple “keyword-networks”, which provides capability to grow not only for a distributed database system but also for a stand-alone database system. We also propose a CAD tool for developing this new type of database systems, and show one of our developed applications using that, a database of cardiovascular disorder cases. Although the evaluation of this application has not been done yet, we are considering various possibilities to extend and apply our introduced new models.

抄録全体を表示

Genetic Algorithm (GA) has been widely used to solve planning problems. However, it requires to determine the optimal values of many genetic parameters, such as population sizes, crossover probability, mutation probability and so on. To make matter worse, the mostt suitable combination of parameters for one problem is not always optimal for others. Therefore, these parameters should be tuned up whenever the problem changes.
In this paper, we propose an adaptable GA mechanism which has an autonomic parameter tuning for the composition of generic operators. This mechanism raises the probability of the genetic operators which acted effectively, that is, the operator created better individuals than the other operators. And, it adjusts the combinations of genetic parameters as the suitable one for the target problem successively. We applied the adaptable GA mechanism to a project scheduling model (PSM) acrd evaluated with the manual tuning methods.

抄録全体を表示

Robot system uses many kind of sensors to detect obstacles and recognize their environment. Ultra sonic sensor system is such a typical sensor, which is only valid in quiet space. Since, the received signal includes both of its own reflect signal, and reflect signals of other sensors. It is difficult to use sonic sensors multiplically. In this paper, we propose a sonic measurement method under multiple ultra sonic sensors environment. In proposed method, the transmittal signals have their own pattern by using chaos phenomenon. Generally, chaos phenomenon is known as generating unforcastable behavior. Thus, each own transmittal signal pattern can be distinguished from other ultra sonic sensor systems in any transmittal period. Each system calculate the cross correlation of their own transmittal signals and receive signals, and it extracts the right reflect signal. We show some experimental results to demonstrate our proposed strategy.

抄録全体を表示

We have investigated ways to improve pattern recognition ability by combining several small back-propagation neural networks (BPNNs) into a modular-net architecture. In this architecture several subfeatures are extracted from patterns, each subfeature is input into a separate BPNN, and the output vectors from the BPNNs are combined to obtain the recognition results. Using two subfeatures extracted from handwritten digits, we investigated how best to obtain the desired outputs for similar patterns in order to improve the generalization ability of the modular-net architecture. We found that the conventional all-or-nothing desired-output approach, “1” for the correct class and “0” for the other classes, prevents the BPNN outputs for similar classes from becoming sufficiently large, preventing the combined output for the correct class for patterns in which both subfeatures are very similar to those of the other classes from becoming maximal. We also found that modifying the desired outputs according to the similarity of the input patterns (i.e., increasing desired outputs to similar classes) increases the BPNN outputs for similar classes, which help maximize the combined output for the correct class, thus improving the generalization ability of the modular-net architecture. The effectiveness of our approach was shown by several experiments using two subfeatures extracted from handwritten digits.

抄録全体を表示

Spatial positioning of concepts on a plane or in 3-dimensional space is effective in understanding the relationship between them. In addition to the traditional principal component analysis, quantification theory and multidimensional scaling, various studies using neural networks have recently been carried out. Most of them aim at the positioning of concepts based on the similarity between them. However, resulting positioning does not necessarily coincide with that of a user.
Having applications to document retrieval in mind, the present paper determines the positioning of keywords on a plane by an optimization method called quantification theory 4. This positioning is modified reflecting an opinion of a user. This is followed by an inverse optimization method: the acquisition of the similarity between keywords which optimize the modified positioning. This is reduced to the maximization of the square of the cosine of the angle between an output vector and a target output vector. The learning of neural networks under constraints realizes this optimization. An application of this bidirectional positioning to small scale real data demonstrates its effectiveness.

抄録全体を表示