There are many instances in which character shape of a class changes smoothly to that of another class. Although there are many ways of the change, the most delicate change is curvature feature. The paper treat this problem systematically in both theoretically and experimentally. Specifically some confusing pairs were constructed which are well known in the field of OCR, such as 2 Z and 4 9. A series of samples generated using each model which change subtly were provided to conduct a psychological experiment. The results exhibit a monotone change of recognition rates from nearly 100% to 0% as the shape changes continuously. To imitate the humans' performance, feature of curvature was extracted based on continuous function representation based on Bezier's spline curve. Specifically two methods were tried from theoretical and engineering points of view and very successful results were obtained.

AlGaAs/GaAs power HBTs for digital cellular phones have been developed. A three-dimensional thermal analysis taking the local-temperature dependence of the collector current into account was applied to the thermal design of the HBTs. The HBTs were fabricated using the hetero-guardring fully selfaligned transistor technique. The HBT with 220µm2 60 emitters produced a 31.7 dBm CW-output power and 46% poweradded efficiency with an adjacent channel leakage power of -49 dBc at the 50kHz offset bands for a 948 MHz π/4-shifted QPSK modulated signal at a low collector-emitter voltage of 3V. Through comparison with the conventional GaAs power FETs, it has been shown that AlGaAs/GaAs power HBTs have a great advantage in reducing the chip size.

In the traditional note symbol extraction processes, extracted candidates of note elements were identified using complex if-then rules based on the note formation rules and they needed subtle adjustment of parameters through many experiments. The purpose of our system is to avoid the tedious tasks and to present an accurate and high-speed extraction of note heads, stems and flags according to the following procedure. (1) We extract head and flag candidates based on the stem positions. (2) To identify heads and flags from the candidates, we use a couple of three-layer neural networks. To make the networks learn, we give the position informations and reliability factors of candidates to the input units. (3) With the weights learned by the net, the head and flag candidates are recognized. As an experimental result, we obtained a high extraction rate of more than 99% for thirteen printed piano scores on A4 sheet which have various difficulties. Using a workstation (SPARC Station 10), it took about 90 seconds to do on the average. It means that our system can analyze piano scores 5 times or more as fast as the manual work. Therefore, our system can execute the task without the traditional tedious works, and can recognize them quickly and accurately.

The problem of genetic algorithm's efficiency has been attracting the attention of genetic algorithm community. Over the last decade, considerable researches have focused on improving genetic algorithm's performance. However, they are generally under the framework of natural evolutionary mechanism and the major genetic operators, crossover and mutation, are activated by the prior probabilities. An operator based on a prior probability possesses randomness, that is, the unexpected individuals are frequently operated, but the expected individuals are sometimes not operated. Moreover, as the evaluation function is the link between the genetic algorithm and the problem to be solved, the evaluation function provides the heuristic information for evolutionary search. Therefore, how to use this kind of heuristic information (present and past) is influential in the efficiency of evolutionary search. This paper, as an attempt, presents a eugenics-based genetic algorithm (EGA) -- a genetic algorithm that reflects the human's decision will (eugenics), and fully utilizes the heuristic information provided by the evaluation function for the decisions. In other words, EGA = evolutionary mechanisms + human's decision will + heuristic information. In EGA, the ideas of the positive eugenics and the negative eugenics are applied as the principle of selections and the selections are not activated by the prior probabilities but by the evaluation values of individuals. A method of genealogical chain-based selection for mutation is proposed, which avoids the blindness of stochastic mutation and the disruptive problem of mutation. A control strategy of reasonable competitions is proposed, which brings the effects of crossover and mutation into full play. Three examples, the minimum problem of a standard optimizing function--De Jong's test function F2, a typical combinatorial optimization problem--the traveling salesman problem, and a problem of identifying nonlinear system, are given to show the good performance of EGA.

This paper describes advances in the study of handwritten Kanji character recognition mainly performed in Japan. The research focus has shifted from the investigation of the possibility of recognition by the stroke structure analysis method to the study of the feasibility of recognition by the feature matching methods. A great number of features and their extraction methods have been proposed according to this approach. On the other hand, studies on pattern matching methods of recognizing Kanji characters using the character pattern itself have been made. The research efforts based on these two approaches have led to the empirical fact that handwritten Kanji character recognition would become more effective by paying greater attention to the feature of directionality. Furthermore, in an effort to achieve recognition with higher precision, active research work has been carried out on pre-processing techniques, such as the forced reshaping of input pattern, the development of more effective features, and nonlinear flexible matching algorithms. In spite of these efforts, the current character recognition techniques represent only a skill of guessing characters" and are still on an insufficient technical level. Subsequent studies on character recognition must address the question of how to understand characters".

In earlier works we proposed the Exclusive Learning neural NET work (ELNET), which can be utilized to construct large scale recognition system for Chinese characters. However, this did not resolve the problem of how to use training samples effectively to generate more suitable recognition boundaries. In this paper, we propose ELNET- wherein an attempt has been made to deal with this problem. In comparison with ELNET, selection method of training samples is improved. And the number of module size are variable according to the number of training samples for each module. In recognition experiment for ETL9B (3036 categories) using ELNET-, we obtained a recognition rate of 95.84% as maximum recognition rate. This is the first time that such a high recognition rate has been obtained by neural networks.

We have shown a non-invasive method for estimating transient changes in aortic flow and ventricular volume based on optimal control theory by using successful simulations of reported experimental data. The performance function to evaluate the optimality of the cardiovascular system was proposed based oh physical, fluid mechanical and pathophysiological considerations. It involved the work of the ventricle, the rate of changes in the aortic flow and the ventricular pressure. We determined that the cardiovascular system operates optimally when the performance function has been minimized. The relative magnitudes of the reductions of changes in these terms were expressed by the weighting coefficients. The arterial system was described by the Wind Kessel model using arterial resistance, aortic compliance and aortic valvular resistance. We set boundary conditions and transitional conditions derived from the systolic and diastolic phases of the aortic flow and the arterial pressure. The optimized system equations were converted to 6 linear simultaneous differential equations with 6 boundary conditions. The optimal ventricular pressure and aortic flow rate that minimize the performance function were obtained by solving these differential equations. By alternating the weighting coefficients of the work of ventricle and the rate of change in the ventricular ejection pressure, successful simulations of the ventricular pressures recorded from human subjects and those from isolated canine ventricle were obtained. Once the sets of weighting coefficients had been determined by successful simulations of ventricular pressures, the calculated aortic flow curves and pressure volume loops by the present method coincided with the reported experimental data. The changes in ventricular pressure and aortic flow produced by alternating the weighting coefficients to simulate the reported ventricular pressures and aortic flow curves under the different afterload conditions were consistent with biophysical experimental data. The present method is useful to estimata aortic flow curve and ventricular pressure volume loops non-invasively.

The benefits of ATM techniques have been widely recognized and many organizations envisage the introduction of ATM techniques into their telecommunication networks. The ATM benefits can, however, be fully exploited only after effective network resource management techniques have been developed. This paper focuses on CBR-VP management techniques. The ATM transport network architecture and VP roles are summarized. Next, the issues of VP accommodation design are described. The point is how to create a design that accommodates cell loss and cell delay jitter, both of which depend on various network parameters and conditions. For this purpose, analytical procedures based on an M/D/1 queueing model are adopted. The approximation method is shown to be very effective in practical use through computer analysis. The method guarantees conservative QOSs. Finally, the proposed method is applied to several design examples to illustrate VP management issues. The proposed method will enable ATM techniques to be introduced to our telecommunication networks by the mid-1990's.

This study is aimed to derive a new theoretical solution for blind equalizers. Undr the common assumptions for this framework, it is found that the condition for blind equalization is directly associated with an eigenproblem, i.e. the tap coefficients of the equalizer appear as an eigenvector of a higher order statistics matrix. Computer simulations show that very fast convergence can be achieved based on the approach.

A multi-terminal serial optical link(MSOL) achieves very simple and cost effective radio cell configurations because only one pair of fibers is needed. In addition, low cost Fabry-Perot laser diodes(FP-LDs) can be employed. MSOL has a substantial problem in that the beat noise degrades the C/N in the up-link. To reduce this noise, we propose using an automatic wavelength-offset control(AWOC) circuit. The AWOC circuit offsets the LD wavelength by controlling the laser bias current to minimise the RF band beat noise which is inherent in MSOL systems, and enables high C/N optical-microwave transmission. An experimental MSOL consisting of 5 radio access stations, each equipped with AWOC, is constructed to estimate the noise free dynamic range for 800-MHz 20-carrier signal transmission. The up-link comprises a single mode fiber connecting five 1.3-µm FP-LDs operating at 0.2 mW. The down-link consists of a single mode fiber and one 1.3-µm Distributed Feedback type Laser Diode(DFB-LD) emitting at 4.0 mW. The experimental device achieves over 15 dB noise reduction compared to MSOL without AWOC in the temperature range of 0 to 40. By using the proposed AWOC, MSOL can achive low cost optical fiber RF microcell systems that are easy to install. Additionally, when we install MSOL in the radio base station, the links become more cost effective than coaxial cable links; they offer a wide dynamic range and higher transmission quality.

We have developed the Epi-Base SATURN process as a silicon bipolar process technology which can be applied to optical transmission LSIs. This process technology, to which low temperature selective epitaxial growth technology is applied, is based on the SATURN process. By performing selective epitaxial growth for base formation in 2 steps, transistors with a 40GHz maximum cut-off frequency have been fabricated. In circuit simulation based on SPICE parameters of transistors, the target performance required for 2.4 Gbit/s optical interface LSIs has been achieved.

We propose a parallel parsing algorithm based on Earley's method, which works in O(log2n) time using O(n4.752) processors on CREW PRAM. This algorithm runs with less number of precessors compared with previously proposed W. Rytter's algorithm.

This paper proposes a mutual exclusion method that is unified for the parallel and distributed systems. The method partially serializes requests into partial queues of requests, which are next totally serialized into a main queue. A request in the main queue is authorized to enter the critical section (CS) when the request receives the privilege token from the previous request in the queue. In the distributed system of N sites that each is a parallel system, mutual exclusion is performed by cooperation of two algorithms based on the same method. The algorithm for the distributed system works on a logical network (that is a directed tree) of S ( N) sites. The algorithm for each site produces a local-main queue of requests. The chunk of requests in the local queue is concatenated at a time to the partial queue of the distributed system. The the cost of mutual exclusion -- the number of intersite messages required per CS entry -- is reduced to O(1) (between 0 and 3).

In this paper, we present some contributions to improve a previous work's approach presented for the segmentation of magnetic resonance images of the human brain, based on the unsupervised Hopfield neural network. We formulate the segmentation problem as a minimization of an energy function constructed with two terms, the cost-term as a sum of errors' squares, and the second term is a temporary noise added to the cost-term as an excitation to the network to escape from certain local minimums and be more close to the global minimum. Also, to ensure the convergence of the network and its utility in clinic with useful results, the minimization is achieved with a step function permitting the network to reach its stability corresponding to a local minimum close to the global minimum in a prespecified period of time. We present here our approach segmentations results of a patient data diagnosed with a metastatic tumor in the brain, and we compare them to those obtained based on, previous works using Hopfield neural networks, Boltzmann machine and the conventional ISODATA clustering technique.

This paper presents a parallel move generation of a Chess machine system for achieving the purpose of reducing the number of move generation cycles. The parallel system is composed of five move generation modules which share the move generating cycles to reduce the time of building a game tree. Simulation results show that the proposed parallel move generation architecture takes about half of the number of move generation cycles to build a game tree that is the same as the one built by a sequential move generation module.

A new superresolution technique is proposed for high-resolution estimation of the scattering analysis. For complicated multipath propagation environment, it is not enough to estimate only the delay-times of the signals. Some other information should be required to identify the signal path. The proposed method can estimate the frequency characteristic of each signal in addition to its delay-time. One method called modified (Root) MUSIC algorithm is known as a technique that can treat both of the parameters (frequency characteristic and delay-time). However, the method is based on some approximations in the signal decorrelation, that sometimes make problems. Therefore, further modification should be needed to apply the method to the complicated scattering analysis. In this paper, we propose to apply a time-domain null filtering scheme to reduce some of the dominant signal components. It can be shown by a simple experiment that the new technique can enhance estimation accuracy of the frequency characteristic in the Root-MUSIC algorithm.

In this paper, we give an algorithm which, given a set F of at most n-k faulty nodes, and two sets S={s1, , sk} and T = {t1,, tk}, 1kn, of non-faulty nodes in n-dimensional hypercubes Hn, finds k fault-tree node disjoint paths sitje, where (j1, , Jk) is a permutation of (1, , k), of length at most n + k in O(kn log k) time. The result of this paper implies that n disjoint paths of length at most 2n for set-to-set node disjoint path problem in Hn can be found in O(n2 log n) time.

Polynomials are called permutation polynomials if they induce bijective functions. This paper investigates algebraic properties of permutation polynomials over a finite field, especially properties associated with permutation cycles. A permutation polynomial has a simple structure but good randomness properties suitable for applications. The cycle structure of permutations are considered to be related to randomness. We investigate the algebraic structure from the viewpoint of randomness. First we show the relationship between polynomials and permutations using a matrix equation. Then, we give a general form of a permutation polynomial corresponding to a product C1C2Ck of pairwise disjoint cycles. Finally, permutation polynomials with fixed points -or with 2, 3 and 4-cycles -and their compositions are given together with distribution of degree of the permutation polynomials.

A calibrating analog-to digital (A/D) converter employing a T-Model neural network is described. The T-Model neural-based A/D converter architecure is presented with particular emphasis on the elimination of local minimum of the Hopfield neural network. Furthermore, a teacher forcing algorithm is presented and used to synthesize the A/D converter and correct errors of the converter due to offset and device mismatch. An experimental A/D converter using standard 5-µm CMOS discrete IC circuits demonstrates high-performance analog-to-digital conversion and calibrating.

A simple and convenient approach for a radial symmetrical point detection is proposed. In this paper the real part-only synthesis is utilized in order to make an origin symmetric pattern of the original image and to perform automatically the calculation of its autocorrelation for the detection of the symmetry center of the image.