We are considering the integration of power system control and teleprotection signal transmission in ATM networks. To transmit teleprotection signals in ATM networks, we need ATM switching systems and terminal adapters which convert teleprotection signals into ATM cells and vice versa. In ATM networks we have to reduce delay times and delay variations of these signals. This paper describes requirements of an ATM switching system and a terminal adapter to transmit teleprotection signals. The ATM switching system is required a delay priority control and a good shaping method. While a commercial terminal adapter sometimes is not satisfied transmission requirements of digital current diffirential teleprotection signals, we proposed configuration of a terminal adapter to be satisfied these requirements. We examined the number of repeaters in ATM networks when we transmit these signals applying above ATM switching system and a proposed terminal adapter.
A WDM transmission system using three wavelength bands of 1.3μm, 1.47μm, and 1.55μm are proposed. Selection of these wavelengths enables to use popular optical couplers and to apply known optical amplifiers. Among three kinds of optical amplifiers, Tm3+-doped fiber amplifiers for the 1.47-μm band are constructed and the basic properties are experimentally clarified. Crosstalk between the 1.48-μm and 1.55-μm bands originating from isolation of the couplers and Raman scattering is found to make negligibly small by consid-ering the combination of the couplers and suppressing input optical power to the single mode fiber (SMF) under 6 dBm. Experiments of repeaterless transmission over 50-km SMF showed error free performance in 10Gb/s×6 WDM case where two signal lights with different wavelength in each band are launched. Finally, utility of the 1.47-μm band is theoretically discussed.
WDM transmission suitable for the optical network of electric power companies is discussed considering its characteristics such as medium transmission length under 300km and relatively long span length over 100km. The channel spacing in optical wavelength and the input power to fibers are evaluated by bit error rate mainly for 10Gbit/s×8-channel WDM transmission where typical erbium doped fiber amplifiers are used. In the study on the transmission using dispersion shifted fibers, 100GHz spacing is too small and unequal spacing is effective. This is confirmed by a field experiment carried out using an optical ground wire which includes dispersion shifted fibers with a total length of 588km consisting of 6 spans. WDM on single mode fibers is discussed from the viewpoint of dispersion compensation. The channel spacing of 100GHz is found to be enough to transmit 20 channels through the 360-km line consisting of 3 spans if the dispersion compensation is made toeether at the receiver end.
To prevent the collision accident, it is necessary always to calculate a distance from the driver's vehicle to the one in front. In this paper we described a new technique for measuring the distance using stereoscopic images. First the disparity map is calculated from the stereoscopic images. Then this map is projected in the direction of the ordinate axis. This map is called projected disparity map. The vehicle's disparities should be the same whole area of the disparity map. Thus they are translated into horizontal lines segment on the projected disparity map. So we can detect the front vehicle by extracting the straight line. For verifying this method we applied it to the stereoscopic images obtained by a stereo camera mounted on the vehicle traveling on the Tokyo Expressway. And the effectiveness was shown by experiments.
The morphological skeleton defined by a conventional structuring element is not necessarily connected We propose a double structuring element method for producing morphological skeletons of a discrete binary image. The introduced main and sub-structuring elements control the size and orientation of the skeleton, where the substructuring element is a subset of the main structuring element. The resultant skeleton contains the entire skeleton by a single structuring element, which significantly improves connectivity. Furthermore, we show that the morpholgical skeleton by double structuring elements is the intersection of the two skeletons produced by the main structuring element and the divided two subsets of the original substructuring one. Connected skeletons can be obtained by successive application of the theorem starting with a small subset of the substructuring element. Original image can be reconstructed by the substructuring element as well as the main structuring one. We will also discuss another properties of the method and clarify the relation to conventional morphological skeleton.
Since an object is always accompanied with its shadow, each extracted region of the object and its shadow is often assimilated into one region. On an analysis of the object's shapes, the shadow region causes misunderstanding for detection of several moving objects as intruders. In order to detect moving objects by image processing, it is necessary to separate the objects from their shadows. This paper introduced a separation technique of objects from accompanying shadows and presented a monitoring system facilitating detection and tracking for several moving objects. By using proposed separation method of the object and its shadow, tracking of movement loci of several objects was performed and resulted in success of 80%.
In industrial motor drive systems used in industrial robots and machine tools, a torsional vibration is often generated by the elastic elements in torque transmission. Such a system can be modelled simply as a 2-mass-system. Many control methods have been useful to suppress the torsional vibration. The state feedback control using an observer, robust control using H∞ theory, and a resonance ratio controller are some of them. However, these systems and determining of control parameters are complex. In this paper, we propose a scheme using only PID control to suppress the torsional vibration. The control performance of the proposed scheme is checked by experimental results.
Graph partitioning problems (GPP) are the most important NP complete problems faced on the design of VLSI chips. Since one of the most powerful methods to solve such problems is simulated annealing (SA) algorithm, it wastes too much time in annealing process. In this paper, we propose a method of learning automata for GPP and construct two algorithms based on the learning automata with a fixed structure and the learning automata with a variable structure. The computer simulations show that our method has comparable accuracy to SA and faster convergent speed than SA. It also shows that our method has better accuracy than mean field algorithm (MFA) and comparable convergent speed to MFA.
This paper proposes a new method of detecting changes in object configurations by a mobile robot in changeable indoor environments. For mobile robots working in such an environment, there is a need for a global environment map that has to be updated timely. We are developing a map-updating robot system in which a mobile robot with an original map detects changes in the environment and modifies the map into a new version. The changes are identified by matching map features in the original map to data features extracted from range data acquired at some observation positions. To assure the accuracy of the matching, a precise self-localization is carried out prior to the map-and-data feature comparison. The method has the following characteristics: 1) The uncertainty due to changed objects can be tolerated to some degree by using a tree-based feature searching algorithm to select more reliable data features. 2) A wide range of object changes can be detected since no assumptions on the object configurations are made here.
A noise reduction method of a hearing aid is developed for continuous speech under noisy environment. We make use of a comb filter for the noise reduction. The comb filter requires a fundamental frequency of the voice. A property of continuity in a power spectrum envelope(PSE) is applied for noise-proof fundamental frequency detection. The comb filter for continuous speech with noise often makes strange sound because the harmonic structure at higher frequency is disturbed by the noise. The disturbed frequency band is detected by using two types of PSE, and an estimated noise level is subtracted from the spectrum at this band. Accuracy of the fundamental frequency needed for the comb filter is obtained from these processing. Efficiency of noise reduction was evaluated by normal listeners, and the results showed good performance in the noise reduction. In conclusion, the present noise reduction method is useful for the hearing aid in continuous speech including strange sound suppression.
This paper describes a development of tactile sensor with a high compliant sensing surface on a robot hand. A method of discrimination for a local surface shape of an object is represented with the differetial geometry for tactile images which are detected by tactual sensing. The tactile sensor has the function of obtaining a displacement distribution and a force distribution as tac-tile patterns. These distributions are taken as its pattern induced on the high compliant sensing mechanism by a CCD camera. A method for descriminating these patterns corresponding to the object's surface shape is given by using a method of the differetial geometry. Gaussian curvature and mean curvature of a curved object's surface are given from the first and second fundamental forms in the differetial geometry. Experimental results for estimating the radius of curvature of object's surfaces verify the effectiveness of the proposed methods.
The importance of swift and correct identification of individuals is increasing as a key technology to gain access various kinds of information. Specific features appeared in handwritten characters can be used to discriminate a person from the others without invading privacy. This paper deals with the individual identification from handwritten characters in off-line systems. An effective method to extract the specific features appeared in the straight and/or the curved part of handwritten characters is described in this paper. Our feature extraction method is based on calculation of the length and inclination of straight-lines and parameters of the circular arcs that are found in the binarized and thinned lines of the characters. Since the distribution of those specific features depends on the writer, the distribution pattern can be used for the purpose of individual identification. The effectiveness of this method is examined experimentally. The result of our identification experiments enabled us to obtain an 88.4% identification average.
In this paper, we propose a network model of MST area in visual path. The network is composed of 3 layers. The first layer corresponds to the MT area, the second and third layers correspond to the MST area. The proposed network has special connections at each layer and consists of 2 types of neurons. In monkey's brain, neurophysiological researches found that response amplitude of most MSTd neurons decrease with the decreasing of stimulus size. This result indicates that MST-processings may contain an information about an region size of motion. The Ruye Wang's model using a fully-connected multi-layer network can learn and detect the motion type as well as the center of motion (e. g. rotation, size change). However, it could not give consideration to express region size of motion. The MST network model proposed in this paper can detect the motion type, the center of motion and the region size motion. Further, our model can also express relation between region size of motion and an output value of motion's center.
This paper proposes a method to display a halftone image with only binary pixels from a gray-leveled image. White pixels are allocated in the gray-levels order after the weights based on Bayer matrix are added to all pixels in a local image area. The additional matrix is composed by the decimal weights to determine the order to allocate white pixels easily even if a local area includes multiple pixels with the same gray-level. The proposed method can display an appropriate halftone image even if a gray-leveled image includes even brightness areas. Through the experiments for evaluating the reproducibility of gray-levels, the smoothness of halftone, the resolution and the texture, the equal or better results were obtained by the proposed method in comparison with the conventional half-toning methods.
A new scheme of speech synthesis by rule was presented which could be implemented in ROM less than 4M bytes. The features of our scheme are consist in the use of the vector quantization of LSP parameters for VCV instance and table look-up process. A few dozens of VCV instances were picked up from speech signal for each VCV unit in order to cope with different phonemic enviroments. As many as 128 representative vectors were verified to be sufficient experimentally by hearing test of synthsized voice. This number of 128 made our scheme feasible from view points of memory size and calculation time. The memory size of VCV unit dictionary could be less than 256K bytes. When VCV instances were concatenated by selecting appropriate VCV instances through DP, distance table between 128 representative vectors were looked up in order to speed up process.
A simple method based on the Finite-Difference Time-Domain(FD-TD) method is presented for analyz-ing nonlinear propagation of strong ultrasonic pulses in water. First, the present method is applied to the problem of plane-wave nonlinear propagation which is often used to explain nonlinear phenomena, and the results are compared with theoretical results. Then a model is prepared for the numerical simulation of the Reichenberger's experimental setup in which the ultrasonic pulse radiated from electromagnetic acoustic sources is focussed by a lens. The present method is applied to this model for simulating the shock wave propagation. The results show that the present method can simulate the real shock wave nonlinear propaga-tion in water with fairly high accuracy, provided that supersonic nonlinear parameters in water are assumed to be higher than those known thus far.
This paper describes a method for measuring earth impedance of a tree. As the tree grows, its root well advances in the ground. A grade of the advance may be monitored by measuring the earth impedance of the tree, defined in the fields of electrical engineering. However, the method directly measuring the value has not been developed. The method presently available for measuring the earth impedance of a metal conductor cannot apply to the tree. Because electrode impedance of an electrode using for connection between the measurement circuit and the tree, brings serious error in the measurement. To avoid the effect of the electrode impedance, a circuit connection is newly devised which makes it possible to measure earth impedance of the tree directly. Further, equipotential surface within the trunk is adjusted to be parallel to the ground level by planting the trunk with several current-electrodes. This transaction assures accurate measurement for the cylindrical and semi-conductive object such as trees. Materials used in the experiments were zelkova tree, cherry tree, and silk-tree. Their girths are 1.7-2.0 meters. Obtained results from the experiments were 58.5-103.6 Ω in absolute value, and they showed slight capacitive reactance. It is concluded that the presented method is available for the measurement of earth impedance of the tree.
We made a network graphics system for three dimensional particle motion of seismic waves using Virtual Reality Modeling Language (VRML). This graphics system is made on a personal computer in which par-ticular hardwares are not installed. A VRML viewer and a web browser are used to analyze and manipulate 3D image of particle motion.