A multiple classifier system can be a powerful solution for robust pattern recognition. It is expected that the appropriate combination of multiple classifiers may reduce errors, provide robustness, and achieve higher performance. In this paper, high performance Chinese syllable recognition is presented using combinations of multiple classifiers. Chinese syllable recognition is divided into base syllable recognition (disregarding the tones) and recognition of 4 tones. For base syllable recognition, we used a combination of two multisegment vector quantization (MSVQ) classifiers based on different features (instantaneous and transitional features of speech). For tone recognition, vector quantization (VQ) classifier was first used, and was comparable to multilayer perceptron (MLP) classifier. To get robust or better performance, a combination of distortion-based classifier (VQ) and discriminant-based classifier (MLP) is proposed. The evaluations have been carried out using standard syllable database CRDB in China, and experimental results have shown that combination of multiple classifiers with different features or different methodologies can improve recognition performance. Recognition accuracy for base syllable, tone, and tonal syllable is 96.79%, 99.82% and 96.24% respectively. Since these results were evaluated on a standard database, they can be used as a benchmark that allows direct comparison against other approaches.

We found a novel connecting orbit in the averaged Duffing-Rayleigh equation. The orbit starts from an unstable manifold of a saddle type equilibrium point and reaches to a stable manifold of a node type equilibrium. Although the connecting orbit is structurally stable in terms of the conventional definition of structural stability, it is structually unstable since a one-deimensional manifold into which the connecting orbit flows is unstable. We can consider the orbit is one of global bifurcations governing the differentiability of the closed orbit.

Various applications of cellular neural network (CNN) are reported such as a feature extraction of the patterns, an extraction of the edges or corners of a figure, noise exclusion, searching in maze and so forth. In this paper, we propose a cellular neural network whose each cell has more than two output levels. By using the output function which has several saturated levels, each cell turns to have several output states. The multiple-valued CNN enhances its associative memory function so as to express various kinds of aspects. We report an application of the enhanced asscociative memory function to a diagnosis of the liver troubles.

The capability of generalization is the most desirable property of a learning system. It is well known that to achieve good generalization, the complexity of the system should match the intrinsic complexity of the problem to be learned. In this work, introduction of fractal connection structure in nonlinear learning systems like multilayer perceptrons as a means of improving its generalization capability in classification problems has been investigated via simulation on sonar data set in underwater target classification problem. It has been found that fractally connected net has better generalization capability compared to the fully connected net and a randomly connected net of same average connectivity for proper choice of fractal dimension which controlls the average connectivity of the net.

C1 class smooth interpolation by a fuzzy reasoning for a small data set is proposed. The drafting technique of a human expert is implemented by using a set of fuzzy rules. The effectiveness of the present method is verified by computer simulations and by applications to the practical interpolation problem in a power system.

We propose a new QOS routing algorithm for finding a path that guarantees several quality of service (QOS) parameters requested by users, for ATM networks. It is known that a routing problem is NP-complete, if the number of additive QOS parameters, such as delay and cost, are more than or equal to two. Although a number of heuristic algorithms have been proposed recently to solve this problem, the appropriate choice of routing algorithms is still an open issue. In this paper, we propose a new heuristic routing algorithm, while being compliant with PNNI routing and signaling specification in the ATM Forum. The performance of algorithms is evaluated by simulation with a various network topologies and loading scenarios. This simulation results demonstrate that the proposed scheme improves the performance while reducing computational complexity.

Nonuniform traffic can degrade the overall performance of multistage interconnection networks substantially. In this paper, this performance degradation is traced back to blocking effects that are not present under uniform traffic patterns within a network. This blocking phenomenon is not mentioned in the literature and is termed higher order Head-of-Line-blocking (HOLk-blocking) in this paper. Methods to determine the HOL-blocking order of multistage networks in order to classify the networks are presented. The performance of networks under hot-spot traffic as a function of their HOL-blocking characteristics is studied by simulation. It is shown that network bandwidth and packet delay improve under nonuniform traffics with increasing HOL-blocking order of a network.

In this paper, we show that the conventional BCH codes can be better than the AG codes when the number of check symbols is relatively small. More precisely, we consider an AG code on Cab whose number of check symbols is less than min {g+a, n-g}, where n and g denote the code length and the genus of the curve, respectively. It is shown that there always exists an extended BCH code, (i) which has the same designed distance as the Feng-Rao designed distance of the AG code and the code length and the rate greater than those of the AG code, or (ii) which has the same number of check symbols as that of the AG code, the designed distance not less than that of the AG code and the code length longer than that of the AG code.

A algorithm for designing a pattern classifier, which uses MDL criterion and a binary data structure, is proposed. The algorithm gives a partitioning of the range of the multi-dimensional attribute and gives an estimated probability model for this partitioning. The volume of bins in this partitioning is upper bounded by ο((log N/N)K/(K+2)) almost surely, where N is the length of training sequence and K is the dimension of the attribute. The convergence rates of the code length and the divergence of the estimated model are asymptotically upper bounded by ο((log N/N)2/(K+2)). The classification error is asymptotically upper bounded by ο((log N/N)1/(K+2)). Simulation results for 1-dimensional and 2-dimensional attribute cases show that the algorithm is practically efficient.

A GaAs multilayer microwave integrated circuit(MuMIC) power amplifier with a harmonic rejection filter has been developed for 1.9-GHz digital European cordless telecommunication system. Adoption of the MuMIC structure has auccessfully ended up with Q-factor of 462 harmonic rejection filter. As a result, power-added efficiency of 62.2% and P1dB of 27 dBm have been obtained at drain supply voltage of 3.6V.

An effect of source harmonic tuning on linearity of power GaAs FET's under class AB operation is demonstrated. To improve efficiency of the power amplifiers, GaAs FET's are often poerated under class AB condition. Due to lower bias current, a class AB amplifier begins to show nonlinearity at lower input power comparing with class A operation, and as the power level of the input signal increases, however, an output power sometimes increases abruptly. From nonlinear circuit simulation, we have found this phenomenon is occurred by the distortion in gate RF voltage, and by suppressing even-order harmonics in the input circuit of GaAs FET, class AB amplifiers can be effectively linearized. In this paper, we show the condition for improving the linearity of power CaAs FET's under class AB operation by the source harmonic tuning technique.

A learning algorithm is presented for nearest neighbor pattern classifiers for the cases where mixed supervised and unsupervised training data are given. The classification rule includes rejection of outlier patterns and fuzzy classification. This partially supervised learning problem is formulated as a multiobjective program which reduces to purely super-vised case when all training data are supervised or to the other extreme of fully unsupervised one when all data are unsupervised. The learning, i. e. the solution process of this program is performed with a gradient method for searching a saddle point of the Lagrange function of the program.

Access control has been an important security issue in information systems. Multilevel hierarchical information access widely exists in present-day government, military, and business applications. Extending access control design to work in a hierarchical environment is natural and necessary but rarely addressed so far in the literature. In this paper, a dynamic group-oriented cryptographic scheme to access a multilevel data hierarchy is proposed. In the proposed scheme, a trusted central authority is in charge of the administrative activities among the organization hierarchy. At the beginning, each user class submits its associated information and a cryptographic key of its preference to the central authority. Next the central authority generates a public information for each class according to their location in the organization hierarchy. The cryptographic key held by each class can be used directly as an encryption key to encipher data. These keys need not be modified when adding/deleting a class to/from the system. Compare with other existing schemes, ours has the advantages of flexibility in choosing user preferred cryptographic keys, cryptographic keys not exceeding a fixed length, reduced storage space in publishing pubic information, and protection from conspiracy attack.

Up to now, the only useful and well-known structural or initial-marking-based necessary and sufficient liveness conditions of Petri nets have only been those of an extended free-choice (EFC) net and its subclasses such as a free-choice (FC) net, a forward conflict free (FCF) net, a marked graph (MG), and a state machine (SM). All the above subclasses are activated only by deadlock-trap properties (i.e., real d-t properties in this paper), which mean that every minimal structural deadlock (MSDL ND=(SD, TD, FD, MoD)) in a net contains at least one live minimal structural trap (MSTR NT=(ST, TT, FT, MoT)) which is initially marked. However, the necessary and sufficient liveness conditions for EFCF, EBCF, EMGEFCFEBCF, AC (EFCFC), and the net with kindling traps NKT have recently been determined, in which each MSDL without real d-t properties was also activated by a new type of trap of trap, i.e., behavioral traps (BTRs), which are defined by introducing a virtual MSTR, a virtual maximal structural trap (virtual STR), a virtual MSDL, and a virtual maximal structural deadlock (virtual SDL) into a target MSDL. In this paper, a structural or initial-marking-based necessary and sufficient condition for local liveness (i.e., virtual deadlock-trap properties) of each MSDL ND s.t. SDST, SDST, SDST (but ND s.t. SDST is dead owing to real deadlock-trap properties) in a general Petri net N is presented by extending that in NKT. Specifically, live minimal behavioral traps (MBTRs) as well as live maximal behavioral traps (BTRs), i.e., virtual deadlock-trap properties, in a general Petri net N are characterized using the real d-t properties of each MSDL ND s.t. SDST for a general Petri net N, which were also obtained by extending the concept of return paths in NKT in connection with an MSDL which contains at least one MSTR and by using the concepts of T-cornucopias and absolute T-cornucopias in a subclass Ñ of N. In other words, BTRs are defined by introducing a virtual MSTR, a virtual STR, a virtual MSDL, and a virtual SDL into a target MSDL without real d-t properties. Additionally, a structural or initial-marking-based necessary and sufficient condition for liveness of a new subclass Nn of a general Petri net N (i.e., a general Petri net without time) is derived, and the usefulness of the obtained results is also discussed.

There have been many developments on neural network research, and ability of a multi-layered network for classification of multi-spectral image data has been studied. We can classify non-Gaussian distributed data using the neural network trained by a back-propagation method (BPM) because it is independent of noise conditions. The BPM is a supervised classifier, so that we can get a high classification accuracy by using the method, so long as we can choose the good training data set. However, the multi-spectral data have many kinds of category information in a pixel because of its pixel resolution of the sensor. The data should be separated in many clusters even if they belong to a same class. Therefore, it is difficult to choose the good training data set which extract the characteristics of the class. Up to now, the researchers have chosen the training data set by random sampling from the input data. To overcome the problem, a hybrid pattern classification system using BPM and Kohonens feature mapping (KFM) has been proposed recently. The system performed choosing the training data set from the result of rough classification using KFM. However, how the remotely sensed data had been influenced by the KFM has not been demonstrated quantitatively. In this paper, we propose a new approach using the competitive weight vectors as the training data set, because we consider that a competitive unit represents a small cluster of the input patterns. The approach makes the training data set choice work easier than the usual one, because the KFM can automatically self-organize a topological relation among the target image patterns on a competitive plane. We demonstrate that the representative of the competitive units by principal component analysis (PCA). We also illustrate that the approach improves the classification accuracy by applying it on the classification of the real remotely sensed data.

In this paper we study the properties induced by the symmetrical properties of a system of hybridly coupled oscillators of the Rayleigh type on the bifurcations of its equilibria. We first discuss the symmetrical properties of the system. Then we classify the equilibria according to their symmetrical properties. Demonstrating the structural degeneracy of the system, we give the complete stability analysis of the equilibria.

In this paper we study the bifurcations of the periodic solutions induced by the symmetrical properties of a system of hybridly coupled oscillators of the Rayleigh type. By analogy with the results concerning with the equilibria, we classify the periodic solutions according to their spatial and temporal symmetries. We discuss the possible bifurcations of each type of periodic solution. Finally we analyze the phase portraits of the system when the parameters vary.

The wavelet transform provides information both in the spatial domain and in the frequency domain because of its inherent nature of space-frequency analysis. This paper presents a classification result of synthetic aperture radar image obtained by JERS-1 based on the discrete wavelet transform. This paper points out that the wavelet analysis has yielded a fine result in texture classification compared to a conventional method with less computation time.

The classification of remotely sensed multispectral data using classical statistical methods has been worked on for several decades. Recently there have been many new developments in neural network (NN) research, and many new applications have been studied. It is well known that NN approaches have the ability to classify without assuming a distribution. We have proposed an NN model to combine the spectral and spacial information of a LANDSAT TM image. In this paper, we apply the NN approach with a normalization method to classify multi-temporal LANDSAT TM images in order to investigate the robustness of our approach. From our experiments, we have confirmed that our approach is more effective for the classification of multi-temporal data than the original NN approach and maximum likelihood approach.

Object-oriented programming requires different skills from those of traditional structured programming. Thus, a good interactive environment for beginners of object-oriented programming should be provided. We have designed and implemented a visual environment of object-oriented programming for beginners. If a programmer draws a diagram of the tree of the hierarchy of classes visually by using our tool, the relationship between superclasses and subclasses are automatically established. Moreover, in order to prevent careless mistakes to override methods, the prototype environment in the Smalltalk language checks written methods. We conducted an experiment with our tool and evaluated its usefulness.