Showing papers in "Electronics and Communications in Japan Part Iii-fundamental Electronic Science in 1990"

A computational algorithm for solving 0-1 goal programming with GUB structures and its application for optimization problems in system reliability

Abstract: Upon to now, system optimal allocation problems such as system reliability and system availability problems have been formulated as single-objective problems and solved through the use of various well-developed optimization techniques. However, in this field, there are many problems that cannot be solved without applying MODM (multiple-objective decision making) methods. These methods deal with multiple objectives that conflict with each other instead of formulating the problem as a single objective programming problem which optimizes only the reliability of the cost function, as is done in previous methods. GP (goal programming) is one of the most powerful MODM tools in this field. In practical MODM problems, many GP problems involve a large number of 0-1 decision variables and a special type of 0-1 variable, which arises during the transformation of non-linear integer programming into 0-1 linear programming. In this paper, we propose an efficient and specific algorithm for solving large-scale 0-1 GP problems in particular structures, which are termed GUB structures. Furthermore, to illustrate the effectiveness of the algorithm proposed here, we introduce two numerical examples from among the problems of system reliability, and compare the algorithm proposed with previous methods.

Approximate realization of identity mappings by three-layer neural networks

Abstract: Recently, applications of neural networks to information compression, e.g., image data compression, have been studied. This paper discusses these problems from a viewpoint of the approximate realization of identity mappings by the use of three-layer neural networks. The relationship between statistical properties of data, hidden units number and the infimum of approximation error also is clarified theoretically by the study of the relation with principal component analysis in the case in which the hidden layer generally has nonlinear units. As a result, it is proved that the capability of information compression by three-layer networks does not exceed the capability of the K-L transformation method. Moreover, the phenomenon of generalization in this case is explained theoretically.

A new method for discrete optimization problems

Abstract: A generalized optimization system with a discrete decision space, is described, and on an optimization problem is defined which is associated with the system. A new solution method, which is called modular approach (MA), is presented to solve the optimization problem. This method extends the Morin-Marsten hybrid idea to solve troublesome problems of dynamic programming. The present method is also an extension of the branch-and-bound method using breadth-first search.

Algorithms for finding maximum and selecting median on a processor array with separable global buses

Abstract: A problem has been considered in which the maximum among N data should be determined by providing a global bus in a processor array. As a solution to this problem, various kinds of global bus configurations have been proposed from the viewpoint of improving the computation time. This paper considers the processor array with global buses, which includes several switching units and can be separated by the switching control from the processors. Parallel algorithms are constructed which can solve efficiently the semigroup computation such as maximum determination and the median selection problems. The computational complexity of the proposed algorithm is evaluated to indicate that the proposed method has an asymptotically desirable property. Furthermore, to verify that the separable global buses can be embedded efficiently in a VLSI chip, the area-time complexity is evaluated as a performance measure for the VLSI circuit. Comparing the result with the case of the processor arrays with the traditional global bus configuration, the proposed algorithms are shown to be more VLSI-oriented.

Multiobjective fuzzy linear regression analysis and its application

Abstract: This paper tests three formulations of linear programming problems used to derive a fuzzy linear regression model, which are based on three different indices representing the necessity and of sufficiency equivalence relations between fuzzy numbers. It is shown that those formulations are extensions of three traditional formulations for fuzzy data. The trade-off between “fit” (between the model and the observed data) and the “fuzziness” (of the model) is considered to be a multiobjective decisionmaking problem. This leads to the formulation of a multiobjective programming problem for deriving the fuzzy linear regression model. Corresponding to this situation, an interactive decision-making procedure is proposed, which is applied to the problem of estimating the pork demand.

Complex linear prediction method based on positive frequency domain

Abstract: This paper proposes a linear prediction method using the analytic signal which occupies half the bandwidth of the real signal. Consequently, the order is halved in the proposed method, thereby reducing drastically the computational complexity. The proposed method gives a stable solution for a finite word-length operation, and the obtained reflection coefficient has a low coefficient sensitivity. These are both useful properties in the synthesis of speech.

Orthogonal projection algorithm for block adaptive signal processing and some of its properties

Abstract: With the progress of hardware technology, the development of real-time adaptive algorithms is becoming more and more important. The essential questions concerning adaptive algorithms are: 1) increasing the speed of the algorithm; and 2) speeding up the convergence. It seems that the block signal processing and the orthogonal projection on the subspace of state-vectors are effective methods to solve the forementioned questions 1) and 2), respectively. This paper proposes a block orthogonal projection algorithm (BOPA) which combines the block processing with the orthogonal projection operation. Next, an analysis of this algorithm is presented and the fundamental properties of the relation between the block length and the convergence characteristics are shown. Also, the efficiency of the proposed algorithm is verified by computer simulation. This algorithm is based on a block-by-block adaptation where the operation of orthogonal projection on the state-vector space is performed repetitively for arbitrary blocks.

A realization scheme for the identity-based cryptosystem

Abstract: At Crypto ′84, A. Shamir [1] presented the new concept of identity-based cryptosystems. Yet no realization scheme has yet been proposed. This paper proposes a scheme for realizing an identity-based cryptosystem which modifies Shamir's original concept without losing its essential characteristics. The fundamental idea of this realization scheme is based on the well-known one-way functions, “the difficulty of calculating a discrete logarithm” and “the difficulty of factoring the composite of two large primes.” As long as all the users in the system keep their own secret keys secure, this cryptosystem is perfectly secure. Even if there is a conspiracy among users in the system, this system remains secure when the number of conspirators is less than a certain number which has been determined during the design phase. In this way, a threshold scheme for a secret information sharing system is realized. This paper presents a detailed analysis of this threshold scheme and shows how to evaluate the secret key of the center when the number of conspiring users is over the threshold. In addition, a scheme which facilitates a reduction in the amount of computation needed to generate a common key is also presented.

A new class of 2‐D digital filters composed of all‐pass subfilters

Abstract: This paper proposes a new construction method for 2-D IIR digital filters with all-pass subfilters as components. Recently, numerous studies have been made to construct digital filters by combining all-pass characteristics. This paper is based on the construction by a parallel connection of an all-pass subfilter and delay element, which can realize approximately the linear phase characteristic. First, the basic design for 1-D digital filters as well as its properties are described. Then the theory for the 1-D case is extended to the 2-D case, and a new construction method for 2-D digital filters with low-pass characteristics with circular symmetry is shown. The proposed method features the following. The phase characteristic is approximately linear in the passband. The filter can be realized with a smaller number of multipliers than in the traditional method. The coefficient sensitivity is low.

SC immittance simulation circuits using ugbs and their applications to filters

Abstract: As switched capacitor (SC) immittance simulation circuits based on the bilinear z transformation, circuits are proposed that realize the inductor (L) as the linear system of s, and the FDNC (D) and FDNR (M) as the quadratic systems. The circuit configuration is separated into the feedback circuit and the immittance realization unit and their realizations are constructed. Since the UGB (Unity Gain Buffer) with a simple circuit is used in the active circuit for these units, one can expect low power consumption, small chip areas and high-speed operation. All the proposed circuits can be constructed with a binary phase clock. The element values can be determined only by one capacitance value comprising the feedback circuit and the sampling period. With respect to the inductor and the FDNC, the parallel resistance and capacitance generated by the choice of the capacitance value are investigated. With respect to FDNR, it is shown that an M-L-C-R parallel circuit with arbitrary values of M and L can be obtained by simply adding one set of feedback circuits. Finally, as an application of the proposed circuit, a low-pass filter, a bandpass filter and a high-pass filter are constructed. Their operations are confirmed by the experiments using individual components.

Analysis of a single server model with two queues having different service disciplines

Abstract: This paper analyzes a single server model with two queues using the generating-function method. Two service-discipline models are used: a model in which the service discipline of one queue is “exhaustive” and that of the other queue is “gated”; and a model in which the service discipline of one queue is “exhaustive” and that of the other queue is “1-limited.” For each model, the Laplace-Stieltjes transform (LST) of waiting time distributions and mean waiting times are derived. By using the results, the difference of the mean waiting times in the two models is shown with numerical examples. It is possible to control the characteristics (e.g., the mean waiting times in each queue and their ratio, which depend on parameters such as the traffic density) of models with mixed service disciplines by changing the combination. Therefore, it is important to analyze these models for designing an integrated system which handles different traffic characteristics and required qualities.

Generalized Minkowski metrics for mixed features

Abstract: This paper presents simple and convenient generalized Minkowski metrics on multidimensional feature space in which each coordinate axis is associated not only with a quantitative feature, but also a qualitative feature. The metrics are defined on a new mathematical model (U(d), +, × constructed by the authors called simply the Cartesian space model. In this model, U(d) is the feature space, + is the Cartesian join operator which abstracts a generalized description from two given descriptions on U(d), and × is the Cartesian meet which extracts a common description from two given descriptions on U(d). In our approach, intervals and finite sets are also permitted as possible feature values. Two simple numerical examples are presented to illustrate the usefulness of our metrics.

The dead‐zone conductor hyperchaos generator

Abstract: This paper discusses a simple four-dimensional autonomous hyperchaos generator. This circuit includes only one nonlinear element: a dead-zone conductor (serially connected Zener diodes) and the circuit dynamics is described by a piecewise-linear canonical equation. Regarding the dead-zone conductor as a switch, the circuit equation can be simplified and hence a two-dimensional Poincare map can be defined. This simplification is proved to be reasonable for an n-dimensional canonical form. The Poincare map and its Lyapunov exponents can be calculated by using exact solutions and they explicate the generation of hyperchaos and related phenomena.

Identification of fuzzy parameters by interval regression models

Abstract: Tanaka et al. formulated the problem of determining the linear regression expression with fuzzy parameters as a fuzzy regression analysis problem, using the data for the input-output relation. This paper discusses such a fuzzy regression analysis, and proposes a method which identifies the fuzzy parameters using the interval regression expression. First, a method is proposed which identifies the fuzzy parameters with symmetrical triangular form from the interval regression expression. Then a method is proposed which identifies the fuzzy parameters with asymmetrical trapezoid form from two interval regression expressions. Using numerical examples, a method is illustrated which identifies the asymmetrical trapezoid fuzzy parameters from the usual point data and the fuzzy data. The method proposed herein employs the interval regression expression based on the interval operations, which leads to the merit that the procedure is simple and easy to understand compared with the conventional method employing the operations of fuzzy numbers. The fuzzy regression expression with asymmetrical trapezoid fuzzy parameters is a generalization of the fuzzy regression expression with symmetrical fuzzy parameters obtained by the conventional method.

An efficient algorithm for the linear assignment problem

Abstract: Although numerous exact algorithms have been proposed so far for the linear assignment problem, they are not necessarily satisfactory from the viewpoint of computation time when considered for application to practical large-scale problems. This paper proposes a new algorithm which can reduce the expected computation time needed to find an exact solution for the linear assignment problem by limiting the search space as much as possible within the scope in which no optimal solution is missed. Here, first, an equation is derived to give an upper bound of the search space so as not to miss an optimal solution in the process of finding an alternating path. Next, data structure is described for developing an operation to limit the search space based on this equation and an assignment algorithm is derived. As the results of experiments, the expected computation time of the algorithm O(n2) for preprocessing and 0(nl.6) for the process of finding an assignment of points were found. In addition, it was shown that the algorithm can be applied to large-scale problems since in the planar assignment problem for which an assignment cost function is defined in the plane, the expected computation time is 0(n1.6) including preprocessing.

Linear structure and error performance analysis of block PSK modulation codes

Abstract: This paper investigates the linear structure and error performance analysis of block PSK modulation codes. The integer group (0, 1, 2, 2l – 1) under the modulo-2l addition, denoted as Sl, is chosen to represent a 2-dimensional 2l-PSK signal set. A necessary and sufficient condition is derived for a block PSK code over Sl to be closed under the component-wise modulo-2l addition. Error performance of block PSK codes also are analyzed for an AWGN channel with soft-decision maximum likelihood decoding. Upper bounds and simulation results on the error probabilities for some specific codes are shown.

Application of numerical schemes admitting symmetries

Abstract: This paper shows that when a single-step numerical method to integrate ordinary differential equations admits some symmetries, an infinite number of numerical solutions can be obtained from a single one without a further numerical experiment. When this property is obtained, trajectories to be integrated numerically can be confined in advance, which may lead to a reduction in the number of experiments. Here, by symmetry is meant invariance of difference equations under an action of a Lie group. However, it is not necessarily easv to judge if a scheme admits a symmetry. This paper illustrates a combination of a backward Euler method and a forward one, whereby a symmetry is attached automatically to any invariant (first integral). Kepler motion is treated as an example. For this system, the integration scheme introduced here conserves three angular momenta which produce a symmetry composed of all threedimensional rotations. By means of this symmetry group, one numerical solution produces a three-parameter solution by way of linear transformations. It is shown also that the numerical values calculated directly by the scheme and those obtained from another solution by means of a symmetry agree with each other with sufficient accuracy.

Design of separable-denominator two-dimensional digital filters based on the reduced-dimensional decomposition of frequency domain specifications

Abstract: This paper proposes a new technique for design of separable-denominator two-dimensional digital filters (2DDF·s) in the frequency domain. The technique is based on decomposing both the transfer function of a separable-denominator 2DDF and the given 2DDF frequency domain specification into a product of two one-dimensional ones. By these decompositions, the design problem of a separable-denominator 2DDF can be reduced to the design of two IDDF·s. The technique makes it possible to design separable-denominator 2DDF·s by utilizing IDDF design techniques directly. Thus stable separable-denominatro 2DD·s satisfying the given specifications can be designed efficiently.

Design of three‐dimensional IIR digital filters with complex coefficients using spectral transformations

Abstract: This paper describes a new method for designing two-dimensional (2-D) digital filters with complex coefficients having circularly symmetric and fan characteristics by applying variable transformation (spectral variable) to one-dimensional (1-D) IIR digital filters with complex coefficients. A 2-D digital filter with complex coefficients has an advantage in that it can increase greatly the degree of freedom for design, since once it satisfies the design specification partly in the frequency plane, its characteristics can be arbitrary in other regions. Here, a design method is proposed which makes positive use of this property of the complex coefficient digital filters. Since we have introduced the complex coefficient filters into spectral transformation functions and 1-D prototype filters, there is no constraint on cutoff frequency (radius) accomplishable in circularly symmetric filters. In fan filters even a characteristic that has been difficult to design unless one uses optimization techniques, has become easy to design by means of variable transformation without using any optimization techniques. Moreover, the stability of the resulting IIR filters is guaranteed in simple conditions.

A method to construct constant weight t-error correcting codes

Abstract: This paper proposes a construction method for constant weight t-error-correcting codes. These codes are nonsystematic t-error-correcting and all-unidirectional-error-detecting (t-EC/AUED) codes. The format of these codes is given by generalization of the Bose-Rao code and the Kundu-Reddy code, and a codeword of these codes is the concatenation of constant weight codes. Concrete methods to construct 3- and 4-EC/AUED codes also are given. For example, 3-EC/AUED codes of length 80 proposed so far have at most 246 codewords, whereas the proposed method gives a code having 251 codewords when Bose's method for encoding binary information words into constant weight words is used.

Analysis of frequency characteristics for flexural‐mode tuning‐fork type quartz crystal resonator, taking into account thickness, rotary inertia and shear force

Abstract: A study was conducted with a view to finding theoretically and experimentally the relationship between the resonant frequency and its temperature dependence of the thickness of the resonator. As the analytical means, the energy method was used for deriving the equation of motion in which the thickness and rotary inertia and shear force were taken into account. Since a tuning-fork type flexural-mode quartz crystal resonator is considered to have two cantilevers, the equation of motion was solved under the boundary condition of a clamped-free end. The frequency equation was derived as a function of the thickness zO' width xO and lenght yO. Next, from this equation, the following relationships are derived: the frequency constant (f·yO) versus the dimensional ratio Rxy = xO/yO, the resonant frequency versus the dimensional ratio Rzy = zO/yO, the turnover temperature point Tp versus the dimensional ratio Rxy and the relationship between the first- and second-order frequency temperature coefficients and Tp versus the dimensional ratio Rzy. It is shown that the apex temperature Tp changes significantly for the thickness. Further, these results were compared with the experimental data and good agreement between them is shown. Hence, it is concluded that the present analysis results are useful for the design of new frequency resonators.

Fundamental studies on combined coding for information source and channel

Abstract: From both theoretical and practical viewpoints, one of the most basic and important problems in the digital communication system is to minimize the distortion between the output signal from the information source and the destination signal on the receiving side, under the condition that the transmission rate in the channel is fixed. From such a viewpoint, this paper presents a detailed investigation of the trade-off relation between the vector quantization (source coding) and the error-correcting coding (channel coding), where the transmission rate of the channel is fixed as constant. The combined source and channel coding is shown, which is optimal in the sense that the average overall distortion is minimized in the communication system. The minimized average distortion is compared with the average overall distortion, where the ideal error-correcting code with the minimum distance given by Hamming's upper bound is applied. A detailed discussion is made. The memoryless Gaussian source, the squared-error distortion measure, the binary symmetric channel and the binary (n, k) block error-correcting code are assumed. Finally, an investigation is made using the actual speech communication system.

Ring code – code symmetry and uncorrectable errors

Abstract: This paper proposes a ring code as a new burst error-correcting code, where the check sequence forms a ring on a square code array. The code is a quasi-cyclic quasi-optimal code with a symmetrical structure. Because of its cyclic decoding process, it has a feature that the error-correcting code for a long burst can easily be constructed. This paper discusses also the properties of the code, as well as the correspondence between the code symmetry and the uncorrectable property. The relation is described clearly from the viewpoints of geometrical structure and the group theory. The computation for the uncorrectable probability is established. Finally, the two-dimensional ring code is considered, but the theory can be extended to the three-dimensional and the higher-dimensional ring code. Due to the symmetry of higher order, the hidden errors are reduced and the error-correcting power is improved. The exact evaluation of the performance of the code, however, is left to another paper.

An adaptive LMS algorithm with a quick tracking property for time-varying parameter estimation

Abstract: This paper proposes a method which estimates time-varying parameters of a nonstationary signal such as a biological one in real time. This method is based on a local quasi-stationary concept in the block processing and on the standard LMS algorithm in the recursive processing. That is, we consider a first-order approximation to the time-varying parameters for a short period of time, and estimate coefficients of the constant and the first-order variation terms by the standard LMS algorithm. At the end of each period the estimated values are renewed. This approach does not require calculating matrix inversion and It uses only input values at each time to estimate time-varying parameters adaptively. Also, it can be implemented because of its simplicity. The most attractive point of this approach is that the coefficients of the first-order variation terms are estimated at each time, and thus quick tracking of time-varying parameters can be achieved. This improves the tracking performance drastically. A convergence property of the proposed algorithm is discussed theoretically. Finally, simulation results are presented to illustrate the tracking performance of the proposed algorithm compared with that of the standard LMS algorithms.

Ordering properties for multiple‐choice questionnaire items

Abstract: The multiple-choice questionnaire is used frequently in psychological investigations of human behavior. Previously the author has proposed SS (semantic structure) analysis. The method is based on the order of the questionnaire items, and the structurization of the items is considered using the asymmetrical measure called ordering coefficient, which is a quantitative representation of the order of the items. The multiple-choice questionnaire can roughly be divided into the following three kinds: (1) Type I data, which are evaluation measure data where the score is high when the evaluation is high; (2) Type II data, which are evaluation measure data where the score is the highest at the central value of the evaluation and lower at both sides; (3) Type III data, which are a nominal measure data where the score does not have a particular meaning. Semantic structure analysis which was reported previously, is concerned with Type I data. With that method as the background, this paper newly proposes the ordering coefficient, which can be applied also to Type II and Type III data. Then the effective category number is proposed as an index representing the statistical property of the answer distribution. It is shown that the sequence reference of the ordered structure based on the ordering coefficient is the second-order moment around the central value in Type II data, and is the effective category number in Type III data. Finally, the usefulness of the structurization methods is investigated and possible applications are discussed.

A method of designing FIR fan filters based on McClellan transformations

Abstract: This paper describes a method of designing the fan filter with a steep cutoff characteristic, using McClellan transformation. The conventional design by McClellan transformation has a problem in that the cutoff region of the fan filter is broad due to the restriction to a lower order of the subfilter employed. From such a viewpoint, this paper proposes a systematic design method for the subfilter with arbitrary order. First, it is pointed out that the ideal response of the subfilter can be represented in a functional form. Then the error function is defined in regard to the ideal response, and the design is shown for the subfilter where the filter coefficients minimize the error function. Finally, a design example is discussed, and it is shown that a fan filter with a steeper cutoff characteristic than in the conventional method can be designed by the proposed method.

Saving viterbi decoder states for general rate 1/2, constraint length K, quick look‐in codes

Abstract: This paper discusses how to reduce the number of states in the main decoder placed in an SST Viterbi decoder for general rate 1/2, constraint length K, and quick look-in codes. Presented are three different cell configurations for which state elimination is possible if an MAP estimation method is applied for branch metric computation. It is shown that for each of these cell configurations, a considerable portion of decoder states can be eliminated without any accompanying performance degradation. It is clarified also that, by transforming encoder states to syndrome former states (i.e., dual states), these cell configurations can be mapped to those discussed in Schalkwijk's paper.

An approach to channel routing problem using neural networks

Abstract: A model for neural network was proposed by Hopfield et al., and when applied to a traveling salesperson problem, which is known to be an NP-complete problem, good results were obtained. This neural network is a minimal value of an energy function which has a kinetic stable point. When we express constraints and objective functions of objective combinatorial problems in forms of energy functions, neural networks having the energy functions provide solutions to the problems. Thus we applied a solution due to neural networks to the channel routing problem which is known as a method for automatic routing within LSI. This paper reports an outline of a neural network and the idea of an energy function for the channel routing problem, its application as a decision problem and a simple compaction method for optimization, and several results of computer simulation.

Bandpass magnitude and delay approximations using digital transfer functions

Abstract: This paper proposes a method to approximate transfer functions of magnitude and delay digital equalizers for transmission channels with bandpass characteristics. In this method, an extrapolation band is added to the equalization band on the unit circle on the z-plane, and the resulting extended approximation band is mapped on the overall unit circle on the auxiliary complex plane. In the magnitude equalization, the all-pole type transfer function is used for the approximation to have a loss from the extended approximated band. On the auxiliary complex plane, the desired characteristics are squared and inverted. Then the extrapolation characteristics are determined, and the inverted desired characteristics are expanded into Fourier cosine series. This series is transformed inversely into the z-plane to determine the denominator of the transfer function. If the extrapolation characteristics are determined improperly, the resulting transfer functions may become unrealizable. This paper proposes a method of determining the extrapolation characteristics so that it always gives realizable transfer functions. In the delay equalization using the auxiliary complex plane, the approximation cannot be performed in a straightforward manner. In the proposed method, the delay characteristics are treated graphically and the inverted characteristics are expanded into Fourier cosine series. After inverse transformation into the z-plane, the inverse of the series is expanded into the Fourier series and truncated in some length to give the transfer function.

A general method of deriving a discrete normalized system of orthogonal functions and its application to adaptive digital filters

Abstract: The system of orthogonal functions is very useful in science and engineering, and the theory on the orthogonal function system in the continuous time domain has been established. On the other hand, with the advance in digital signal processing, a need for processing the signal in the discrete time domain is acute. With this background, a general derivation of the orthogonal function system in the discrete time domain is desired. This paper proposes a general method for deriving the orthogonal function system in the discrete time domain from the already established orthogonal function system in the continuous time domain. First, the orthogonal function system in the continuous time domain is presented. Next, by the bilinear information of its transfer function, the system is transferred to the discrete time domain. However, in general, the orthogonality in the discrete time domain is destroyed by the bilinear transformation. It is shown that the orthogonality can be maintained in the discrete time domain and the orthonormal function system can be constructed by an appropriate processing. Further, the digital filter using this orthonormal function system is applied to the adaptive digital filter and the number of adaptive parameters is reduced. Finally, by computer simulation, the effectiveness of the adaptive filter configuration by means of the orthonormal function system is proven.