In this paper, an application of genetic algorithms for modeling of an overhead traveling crane and generation of a reference speed pattern are described. The system consists of the overhead traveling crane with a fixed gain feedback controller, a modeling part of the crane and a generation part of the reference input pattern. To realize accurate modeling, the parameters of the crane are estimated by the genetic algorithm based on multiple objective optimization method. The estimated model is used to generate the reference speed pattern. The reference speed pattern is tuned by the genetic algorithm so that both the quick responses and the sufficient damping are achieved. Experimental results demonstrate that the proposed scheme is effective for modeling of the crane and the generation of suitable reference speed pattern.
In this paper, an application of genetic algorithm for parameters estimation in Hodgkin-Huxley equation under anesthetizing. To achieve adequate estimation, five indexes which characterize action potential waveforms are introduced. The set of parameters are regarded as a chromosome of gene and are tuned by the genetic algorithm so that the calculated indexes are correspond to the measured indexes. By using the estimated parameters, the effect of consistency of anesthesia and elapsed time on the ion currents are discussed.
In this paper, an application of neural networks and fuzzy rules for an autonomous mobile robot is described. The neural networks are used to select a suitable action from tracking and avoidance actions under operating environment and are adjusted by the back propagation algorithm so that the selected action corresponds to the action obtained by operator's decision skill. To realize the selected action, the steering angle and the speed are determined by using the fuzzy rules which are also automatically tuned so as to simulate the operator's control skill. Experimental results demonstrate the effectiveness of the proposed scheme for tracking and avoidance motions of the autonomous mobile robot.
In this paper, we propose a neuro-classification method of the new and used bills using time-series acoustic data. The technique used here is based on an extension of an adaptive digital filter (ADF) by Widrow and the error back-propagation method. Two-stage ADFs are used to detect the desired acoustic data of bill from noisy input data. In the first stage, superfluous signals are eliminated from input signals and in the next stage, only the desired acoustic data is detected from output signal of the two-stage ADFs. The output signal of two-stage ADFs is transformed into spectral data to produce an input pattern to a neural network (NN). The NN is used to discriminate the new and used bills. It is shown that the experimental result using two-stage ADFs is better than that obtained by using original observation data.
Magnetic resonance imaging provides a highly efficacity for the diagnosis of brain's disease. There are various methods to extract the required portions from MRI data. In this paper we propose a simple algorithm to carry out it by using vector quantization. Our porposed algorithm has three steps. First step is pixel classification using vector quantization. We classify image's pixel background, CSF (cerebrospinal fluid), gray matter and white matter. Second step, we obtain initial brain region using region mergering. Brain region consists of gray matter and white matter. Final step, we extracted brain region using location information and region size. We compare brain volume using proposed method in this paper with manually obtained brain volume by a doctor, and the difference for two methods is 0.1∼7.2%.
This paper reports on a signal processing using maximum value, that is, we consider maximum value, when we calculate the probability density (PD) and the exceedance probability. We purpose that maximum value of many samples is useful to emphasize large amplitude of rare phenomena. This signal processing gives us a lot of information. For example, when the ratio generating an impulsive interference is examined on communication channel, the impulsive component is emphasized in PD or exceedance probability. Finally we measured noises and signals on the band (940MHz) of PDC (Personal Digital Cellular) in Japan. Then the experimental results show that we are able to recognize impulsive-type noise well.
The recent advance of the technology of geographic information systems has led to the need for developing efficient and reliable methods of extracting various information from the huge stock of conventional maps. In this paper, an image processing method is thus proposed for obtaining road segments with high precision using multiple maps. Applying several techniques based on elastic matching of line drawings, the proposed method restores incomplete road segments on a map by referring to other maps containing the corresponding complete segments. Successful experimental results are presented for cases where two maps covering almost the same region are available.
In this paper, an extraction method of the handwritten character string superimposed on the delivery slip is proposed. First of all, the HSL transform is given to the delivery slip image input from the TV camera. The chroma color aspect distribution chart is made by using the saturation and the hue obtained by the HSL transform. The cluster is extracted from this chroma distribution chart and a binary code character area is extracted by using the adjustment threshold method. In this case, void elements exist in the obtained binary image, so the element is restored from a binary image by lightness information. The code character is extracted by using structural elements in the morphological filter, whose diameter is equal to the width of the average line estimated from a binary image. All characters were extracted from 200 characters of 50 images as a result of the extraction experiment.
In recent years, growing multimedia systems require more efficient signal separation methods to preserve quality of voice or music recording under noisy environment. Some of signal separation methods are based on minimizing the dependence measure among input signals to separate a noise component since a noise component is usually independent on the other signals. Under such circumstances, we have developed a new method to separate independent signal components which directly minimizes the Kullback-Leibler divergence by a genetic algorithm. The simulation results show that the proposed method is effective to separate the independent signals.
In this paper, we propose a method to classify new and used bills using acoustic data of a banking machine. The proposed method employs an acoustic energy pattern as the time-series data of acoustic energy of the banking machine. By using the energy patterns, we have obtained better classification performance than that of the conventional method using spectral patterns of the acoustic data of the banking machine.
For building intranet systems as one form of Client Server Systems (CSS), new technologies are receiving much attention. However it is difficult to keep abreast of those technological advances for System Engineers (SE). Therefore, we have developed an exercise system for building the CSS especially using intranet technology for novice system engineers. The exercise is a part of a seminar about techniques of building CSS. The exercise system contains a material and manuals. As the material application of the exercise, an order management system is selected. A problem for the exercise is to program a part of the material system, which contains many techniques of developing a CSS. Manuals for the exercise are provided as HTML documents. In order to evaluate the exercise system, six subjects used the exercise system experimentally. For examinations, logs of references to manuals, programming logs, and compiling logs were recorded. By the record, the exercise was divided into three processes; understanding, programming and debugging. The durations of each process was calculated. By the result, we confirmed that learners can exercise in the limited time efficiently and easily.
The fault detection by stochastic qualitative reasoning is an effective way for complex systems such as a building air conditioning system. In this framework, the fault part of a system can be identified by comparing the behavior derived by stochastic qualitative reasoning with the measured behavior. The measured behavior is represented as a series of qualitative values that are obtained by classifying quantitative measurements into several qualitative categories based on the definition of the qualitative regions. The fault detection often fails under the inappropriate definition. This paper proposes a method for defining the normal qualitative regions from the field data of a building air conditioning system. Measurement data in the normal condition must be converted into the stable qualitative value so that the behavior can be distinguished from fault conditions. Therefore, by the probability of occurrence of qualitative value in the reasoning behavior, qualitative regions are determined. This method is applied to a real building air conditioning system. According to the definition of qualitative regions determined from the field data, the fault parts can be successfully identified.
For fault detection of a complex system such as an air-conditioning system, we have developed an effective stochastic qualitative reasoning method. In this method, a complex system is expressed by its simple qualitative models. By comparing the states derived from the reasoning with the observations, the “agreement rate”, which is a parameter that shows how the reasoning reflects the real world, can be calculated. The malfunctioning parts of a system can be identified by the highest “agreement rate”. In the stochastic qualitative reasoning, we have used a landmark separating definition for changing quantitative data into qualitative values. According to this definition, a little change in the quantitative data often gives rise to the radical fluctuation in the corresponding qualitative values. To tackle this problem, we propose a flexible definition using fuzzy set. In this flexible definition, the quantitative data near the landmark have double-defined qualitative values, and the grade of the quantitative data belonging to each qualitative value is expressed by the membership function. This flexible definition was applied to a heat source system of a heat source system, and its effectiveness has been confirmed.
This paper describes a method to detect particular speech segments as a keyword from speech database. The Generalized Hough Transform (GHT) has been used to find arbitrary complex shapes on the image plane. We extended the GHT to be able to deal with three-dimensional shapes and applied it to the keyword spotting in which a speech signal is represented as a spectral sequence. Based on experimental results on the keyword spotting in which 25 keywords were tried to extract from 750 utterances produced by 30 male speakers, we propose an approach for improving the keyword spotting performance. This approach combines GHT and a word spotting method based on DP matching.
For successful application of Genetic Algorithm (GA) to combinatorial optimization problems, a suitable distance between two solutions, or phenotypes is useful to estimate the problem landscape. This paper presents a general distance function between two phenotypes. The phenotypic distance is defined by the least Hamming distance between isomorphic genotypes. Therefore, it is convenient to analyze and control the behavior of genotypes in the search space. By using the phenotypic distance, this paper proposes a new crossover technique named Harmonic Crossover. Because a new child is located between two parents in the problem space, the character of parents is preserved with the Harmonic Crossover. Furthermore, this paper presents a hybrid algorithm which combines GA with a conventional local search algorithm. The effectiveness of the proposed techniques are also confirmed on the traveling salesman problems.
Most of process-centered software engineering environments and those languages focus on process-oriented software development process. However, recent software development tends to require to focus on product-oriented software development, because of emergence of various types of software developments; e.g., software reuse, component-based composition, and so on. In this paper, we propose a new process modeling method named“Mono Process”to support the new software development approach with the idea of software process. Mono Process consists of a set of objects which represent artifacts and resources in the software development. An object has attributes and methods, which represent characteristics and operations of the object, respectively. Mono Process illustrates software development environment as it is, and provides a framework for software process description and software process management.
An optical velocimeter using spatial filtering of the scintillation of laser beam is described. In this approach a sensor array made of photo diodes has been used as a spatial filter for one-component flow-velocity vector measurements. The velocimeter has been developed in order to apply it to an open atmosphere which is probed by a quick scanning tunable diode laser absorption spectrometry (TDLAS) for measuring the column density of the greenhouse effect gases. Thus the velocimetry avails to monitor the gas flux with the eddy correlation method. A feasibility test for the velocimetry has measured the flow-velocity vector of the convection indoors. Measured flow-velocity vector of ascending currents agreed with results from an ultrasonic anemometer.
Calibration techniques of power meters are studied in the radio frequency (RF) of 10MHz-18GHz for the 7mm coaxial waveguide with APC-7 type connector. The measurement principle of RF power is based on an isothermal control type microcalorimeter. It measures the effective efficiency of the standard thermistor mount used as the RF load. Calibration of a power meter is performed by comparing it with the standard thermistor mount of which effective efficiency is known. The signal source is composed of a leveling loop with a broadband synthesizer, a power splitter and a monitor power meter. An automated measurement system was fabricated for both calorimetric measurement and comparison measurement. Measurement examples and the evaluation of the uncertainty of measurements are described.