Negation as failure is realized to be combined with SLD resolution for general logic programs, where the combined resolution is called an SLDNF resolution. In this paper, we introduce narrowing and infinite failure to SLDNF resolution for general logic programs with equations. The combination of SLDNF resolution with narrowing and infinite failure is called an SLDNFN resolution. In Shepherdson (1992), equation theory is combined with SLDNF resolution so that the soundness may be guaranteed with respect to Clark's completion. Generalizing the method of Yamamoto (1987) for definite clause sets with equations, we formally define a least fixpoint semantics, which is an extension of Fitting (1985) and Kunen (1987) semantics, and which includes the pair of success and failure sets defined by the SLDNFN resolution. The relationship between the fixpoint semantics and the pair of sets is regarded as an extension of the relationships for general logic programs as in Marriott and et al. (1992) and in Yamasaki (1996). Instead of generalizing Clark's completion for SLDNFN resolution, we establish, as a model for general logic programs with equations, an extended well-founded model so that the SLDNFN resolution is sound and complete for non-floundering queries with respect to the extended well-founded model.

Dynamical theory of cellular automata on groups is developed. Main results are non-Euclidean extensions of Sato and Honda's results on the dynamics of Euclidean cellular automata. The notion of the period of a configuration is redefined in a more group theoretical way. The notion of a co-finite configuration substitutes the notion of a periodic configuration, where the new term is given to it to reflect and emphasize the importance of finiteness involved. With these extended or substituted notions, the relations among period preservablity, injectivity, and Poisson stability of parallel maps are established. Residually finite groups are shown to give a nice topological property that co-finite configurations are dense in the configuration space.

In this paper, we present simulation algorithms among enhanced mesh models. The enhanced mesh models here include reconfigurable mesh and mesh with multiple broadcasting. A reconfigurable mesh (RM) is a processor array that consists of processors arranged to a 2-dimensional grid with a reconfigurable bus system. The bus system can be used to dynamically obtain various interconnection patterns among the processors during the execution of programs. A horizontal-vertical RM (HV-RM) is obtained from the general RM model, by restricting the network topology it can take to the ones in which each bus segment must be along row or column. A mesh with multiple broadcasting (MWMB) is an enhanced mesh, which has additional broadcasting buses endowed to every row and column. We present two algorithms:1) an algorithm that simulates a HV-RM of size nn time-optimally in θ(n) time on a MWMB of size nn, and 2) an algorithm that simulates a RM of size nn in θ(log² n) time on a HV-RM of size nn. Both algorithms use a constant number of storage in each processor. Furthermore, we show that a RM of size nn can be simulated in θ((n/m)² log n log m) time on a HV-RM of size mm, in θ ((n/m)² m log n log m) time on a MWMB of size mm (m < n). These simulations use θ((n/m)²) storage in each processor, which is optimal.

In this paper, an instruction-cache scheme called Multi-Path Tracing is proposed to enhance the trace cache. Paths are classified to improve the trace cache hit ratio by reducing the path conflict and basic blocks are joined to reduce the hardware cost needed to implement the trace cache. Simulation results for various SPEC integer benchmarks show that the proposed scheme increases the hit ratio by more than 25% and the effective fetch size by 10%.

In this paper, we focus on regularity and set-valued functions. Regularity was first introduced by S. C. Kleene in the propositional operations of his ternary logic. Then, M. Mukaidono investigated some properties of ternary functions, which can be represented by regular operations. He called such ternary functions "regular ternary logic functions". Regular ternary logic functions are useful for representing and analyzing ambiguities such as transient states or initial states in binary logic circuits that Boolean functions cannot cope with. Furthermore, they are also applied to studies of fail-safe systems for binary logic circuits. In this paper, we will discuss an extension of regular ternary logic functions into r-valued set-valued functions, which are defined as mappings on a set of nonempty subsets of the r-valued set {0, 1, . . . , r-1}. First, the paper will show a method by which operations on the r-valued set {0, 1, . . . , r-1} can be expanded into operations on the set of nonempty subsets of {0, 1, . . . , r-1}. These operations will be called regular since this method is identical with the way that Kleene expanded operations of binary logic into his ternary logic. Finally, explicit expressions of set-valued functions monotonic in subset will be presented.

Feature interaction detection determines whether interactions occur or not between the new and existing telecommunication services. Most of conventional detection methods on state transition model utilize an exhaustive search. The exhaustive search is fundamentally very powerful in the sense that all interactions are exactly detected. However, it may suffer from the state explosion problem due to the exponential growth of the number of states in the model when the number of users and the number of features increase. In order to cope with this problem, we propose a new detection method using a state reduction technique. By means of a symmetric relation, called permutation symmetry, we succeed in reducing the size of the model while preserving the necessary information for the interaction detection. Experimental evaluation shows that, for practical interaction detection with three users, the proposed method achieves about 80% reduction in space and time, and is more scalable than the conventional ones especially for the increase of the number of users in the service.

In 1995, 8 kb/s CS-ACELP coder of G.729 is standardized by ITU-T SG15 and it has been reported that the speech quality of G.729 is better than or equal to that of 32 kb/s ADPCM (G.726). However G.729 is the fixed rate speech coder, and it does not consider the property of voice activity in mutual conversation. If we use the voice activity, we can reduce the average bit rate in half without any degradations of the speech quality. In this paper, we propose an efficient variable rate algorithm for G.729. The variable rate algorithm consists of two main subjects, the rate determination algorithm and the design of sub rate coders. For the robust VAD algorithm, we combine the energy-thresholding method, the phonetic segmentation method by integration of various feature parameters obtained through the analysis procedure, and the variable hangover period method. Through the analysis of noise features, the 1 kb/s sub rate coder is designed for coding the background noise signal. Also, we design the 4 kb/s sub rate coder for the unvoiced parts. The performance of the variable rate algorithm is evaluated by the comparison of speech quality and average bit rate with G.729. Subjective quality test is also done by MOS test. Conclusively, it is verified that the proposed variable rate CS-ACELP coder produces the same speech quality as G.729, at the average bit rate of 4.4 kb/s.

In order to improve the efficiency of the feature extraction of backpropagation (BP) learning in layered neural networks, model switching for changing the function model without altering the map is proposed. Model switching involves map preserving reduction of units by channel fusion, or addition of units by channel installation. For reducing the model size by channel fusion, two criteria for detection of the redundant channels are addressed, and the local link weight compensations for map preservation are formulated. The upper limits of the discrepancies between the maps of the switched models are derived for use as the unified criterion in selecting the switching model candidate. In the experiments, model switching is used during the BP training of a layered network model for image texture classification, to aid its inefficiency of feature extraction. The results showed that fusion and re-installation of redundant channels, weight compensations on channel fusion for map preservation, and the use of the unified criterion for model selection are all effective for improved generalization ability and quick learning. Further, the possibility of using model switching for concurrent optimization of the model and the map will be discussed.

In pattern recognition using neural networks, it is very difficult for researchers or users to design optimal neural network architecture for a specific task. It is possible for any kinds of neural network architectures to obtain a certain measure of recognition ratio. It is, however, difficult to get an optimal neural network architecture for a specific task analytically in the recognition ratio and effectiveness of training. In this paper, an evolutional method of training and designing feedforward neural networks is proposed. In the proposed method, a neural network is defined as one individual and neural networks whose architectures are same as one species. These networks are evaluated by normalized M. S. E. (Mean Square Error) which presents a performance of a network for training patterns. Then, their architectures evolve according to an evolution rule proposed here. Architectures of neural networks, in other words, species, are evaluated by another measurement of criteria compared with the criteria of individuals. The criteria assess the most superior individual in the species and the speed of evolution of the species. The species are increased or decreased in population size according to the criteria. The evolution rule generates a little bit different architectures of neural network from superior species. The proposed method, therefore, can generate variety of architectures of neural networks. The designing and training neural networks which performs simple 3 3 and 4 4 pixels which include vertical, horizontal and oblique lines classifications and Handwritten KATAKANA recognitions are presented. The efficiency of proposed method is also discussed.

In the domains involving environmental changes, some knowledge and heuristics which were useful for solving problems in the previous environment often become unsuitable for problems in the new environment. This paper describes two approaches to solve such problems in the context of case-based reasoning systems. The first one is maintaining descriptions of applicable scopes of cases through generalization and specialization. The generalization is performed to expand problem descriptions, i. e. descriptions of applicable scopes of cases. On the other hand, the specialization is performed to narrow problem descriptions of cases which failed to be applied to given problems with the aim of dealing with environmental changes. The second approach is forgetting, that is deleting obsolete cases from the case-base. However, the domain-dependent knowledge is necessary for testing obsolescence of cases and that causes the problem of knowledge acquisition. We adopt the strategies used by conventional learning systems and extend them using the least domain-dependent knowledge. These two approaches for adapting the case-base to the environment are evaluated through simulations in the domain of electric power systems.

This paper demonstrates the necessity of special handling mechanisms for type (or sort) information when learning logic programs on the basis of background knowledge that includes type hierarchy. We have developed a novel relational learner RHB, which incorporates special operations to handle the computing of the least general generalization (lgg) of examples and the code length of logic programs with types. It is possible for previous learners, such as FOIL, GOLEM and Progol, to generate logic programs that include type information represented as is_a relations. However, this expedient has two problems: one in the computation of the code length and the other in the performance. We will illustrate that simply adding is_a relations to background knowledge as ordinary literals causes a problem in computing the code length of logic programs with is_a literals. Experimental results on artificial data show that the learning speed of FOIL exponentially slows as the number of types in the background knowledge increases. The hypotheses generated by GOLEM are about 30% less accurate than those of RHB. Furthermore, Progol is two times slower than RHB. Compared to the three learners, RHB can efficiently handle about 3000 is_a relations while still achieving a high accuracy. This indicates that type information should be specially handled when learning logic programs with types.

In the present paper we shall examine the real-time restoration of biomedical signals under additive noises. Biomedical signals measured by instruments such as catheter manometers, ambulatory electrocardiographs and thermo-dilution sensors are susceptible to distortion and noise. Therefore, such signals must be restored to their original states. In the present study, nonstationary biomedical signals are observed and described using a mathematical model, and several restoration filters that are composed of a series of applications of this model are proposed. These filters restored band-limited approximations of the original signals in real-time. In addition, redundancy is introduced into these restoration filters in order to suppress additive noise. Finally, an optimum filter that accounts for restoration error and additive noise is proposed.

This paper reports on the development of an eyeglass-type infrared-controlled telephone communication interface for the disabled. This system is comprised of four major components: A) a headset; B) an infrared transmitting module; C) an infrared receiving/signal-processing module; and D) a main controller, the Intel-8951 microprocessor. The headset with a tongue-touch panel, a wireless earphone, and a wireless microphone. The infrared transmitting module utilizes a tongue touch panel via tongue-touch circuitry which is converted to an infrared beam and a low power laser (<0.1 mW) beam. The infrared receiving/signal-processing module, receives the infrared beam and fine tunes the unstable infrared beam into standard pulses which are used as control signals. The main controller is responsible for detecting the input signals from the infrared receiving/signal-processing module and verifying these signals with the mapping table in its memory. After the signal is verified, it is released to control the keys of the telephone interface. This design concept was mainly based on the idea that the use of an infrared remote module fastened to the eyeglasses could allow the convenient control of the dialing motion on the keys of a telephone's dialing-pad which are all modified with infrared receiving/signal-processing modules. The disabled are competent for some of work, such as a telephone operator. The increase of opportunity to do a job for the disabled would help them live independently.

The recording of electrocardiogram (ECG) signals for the purpose of finding arrhythmias takes 24 hours. Generally speaking, changes in R-R intervals are used to detect arrhythmias. Our purpose is to develop an algorithm which efficiently detects R-R intervals. This system uses the R-wave position to calculate R-R intervals and then detects any arrhythmias. The algorithm searches for only the short time duration estimated from the most recent R-wave position in order to detect the next R-wave efficiently. We call this duration a WINDOW. A WINDOW is decided according to a proposed search algorithm so that the next R-wave can be expected in the WINDOW. In a case in which an S-wave is enhanced for some reason such as the manner in which the electrodes are installed in the system, the S-wave positions are taken to calculate the peak intervals instead of the R-wave. However, baseline wander and noise contained in the ECG signal have a deterrent effect on the accuracy with which the R-wave or the S-wave position is determined. In order to improve detection, the ECG signal is preprocessed using a Band-Pass Filter (BPF) which is composed of simple Cascaded Integrator Comb (CIC) filters. The American Heart Association (AHA) database was used in the simulation with the proposed algorithm. Accurate detection of the R-wave position was achieved in 99% of cases and efficient extraction of R-R intervals was possible.

This paper presents an improved method of speaker verification using the sequential probability ratio test (SPRT), which can treat the correlation between successive feature vectors. The hidden Markov model with the mean field approximation enables us to consider the correlation in the SPRT, i. e. , using the mean field of previous state, probability computation can be carried out as if input samples were independent each other.