This paper proposes Emergent Behavior Based Architecture (EBBA) that fusions heterogeneous sensor information at the level of behavior modules. The characteristics of EBBA are as follows. i) sensor based architecture, ii) constructed by a set of concurrently executable behavior modules, iii) to have multiple methods to achieve given tasks by utilizing behavior modules, iv) a planner can control emergent behaviors. We also have developed mobile robot navigation system based on EBBA and confirmed the efficiency by experiments in the various situations.

Complex rule conditions are commonly required to describe complicated business semantics. In these cases, efficient condition evaluation is crucial for high performance of active database systems. Most previous works used the incremental evaluation techniques, whose operations are relatively expensive due to the processing based on the exact calculation of the condition expression. In this paper we propose a new filtering technique that effectively identifies false condition in an early stage of condition monitoring. Since the results of condition evaluation tend to be false in most practical cases, an efficient filtering method can highly facilitate fast condition evaluation. The proposed filtering technique is developed based on the new perspective of database state and database operations, i. e. , a vector space model. We first present vector representations of database states, database operations, and complex condition expressions. Then, we propose a filtering method based on the properties of a vector space, called the sphere containment test. Our proposed method determines the truth value of the rule conditions only with the delta vectors maintained in main memory. We compare our method with a typical incremental evaluation method and show that the proposed method can give a significant performance enhancement.

This paper proposes a simple learning algorithm that can realize any boolean function using the three-layer binary neural networks. The algorithm has flexible learning functions. 1) moving "core" for the inputs separations,2) "don't care" settings of the separated inputs. The "don't care" inputs do not affect the successive separations. Performing numerical simulations on some typical examples, we have verified that our algorithm can give less number of hidden layer neurons than those by conventional ones.

Computational expensiveness of the training techniques, due to the extensiveness of the data set, is among the most important factors in machine learning and neural networks. Oversized data set may cause rank-deficiencies of Jacobean matrix which plays essential role in training techniques. Then the training becomes not only computationally expensive but also ineffective. In [1] the authors introduced the theoretical grounds for dynamic sample selection having a potential of eliminating rank-deficiencies. This study addresses the implementation issues of the dynamic sample selection based on the theoretical material presented in [1]. The authors propose a sample selection algorithm implementable into an arbitrary optimization technique. An ability of the algorithm to select a proper set of samples at each iteration of the training has been observed to be very beneficial as indicated by several experiments. Recently proposed approaches to sample selection work reasonably well if pattern-weight ratio is close to 1. Small improvements can be detected also at the values of the pattern-weight ratio equal to 2 or 3. The dynamic sample selection approach, presented in this article, can increase the convergence speed of first order optimization techniques, used for training MLP networks, even at the value of the pattern-weight ratio (E-FP) as high as 15 and possibly even more.

In the design of nonlinear reliable controllers, one major issue is to solve for the solutions of the Hamilton-Jacobi inequality. In general, it is hard to obtain a closed form solutions due to the nonlinear nature of the inequality. In this paper, we seek for the existence conditions of quadratic type positive semidefinite solutions of Hamilton-Jacobi inequality. This is achieved by taking Taylor's series expansion of system dynamics and investigating the negative definiteness of the associated Hamilton up to fourth order. An algorithm is proposed to seek for possible solutions. The candidate of solution is firstly determined from the associated algebraic Riccati inequality. The solution is then obtained from the candidate which makes the truncated fourth order polynomial of the inequality to be locally negative definite. Existence conditions of the solution are explicitly attained for the cases of which system linearization possesses one uncontrollable zero eigenvalue and a pair of pure imaginary uncontrollable eigenvalues. An example is given to demonstrate the application to reliable control design problem.

In this paper, a novel method is proposed for designing two channel biorthogonal filter banks with general IIR filters, which satisfy both the perfect reconstruction and causal stable conditions. Since the proposed filter banks are structurally perfect reconstruction implementation, the perfect reconstruction property is still preserved even when all filter coefficients are quantized. The proposed design method is based on the formulation of a generalized eigenvalue problem by using Remez multiple exchange algorithm. Then, the filter coefficients can be computed by solving the eigenvalue problem, and the optimal solution is easily obtained through a few iterations. One design example is presented to demonstrate the effectiveness of the proposed method.

To improve speech coding quality, in particular, the long-term dependency prediction characteristics, we propose a new nonlinear predictor, i. e. , a fully connected recurrent neural network (FCRNN) where the hidden units have feedbacks not only from themselves but also from the output unit. The comparison of the capabilities of the FCRNN with conventional predictors shows that the former has less prediction error than the latter. We apply this FCRNN instead of the previously proposed recurrent neural networks in the code-excited predictive speech coding system (i. e. , CELP) and shows that our system (FCRNN) requires less bit rate/frame and improves the performance for speech coding.

We have developed a multiple quantum well (MQW) electroabsorption (EA) modulator for wavelength-division multiplexing (WDM) switching systems. The fabricated MQW EA gate has low polarization and wavelength-dependent loss and high extinction ratio within the wavelength range of 1545 to 1560 nm. And by using this gate ultra-high-speed switching is achieved for WDM signals. Moreover, we optimize the EA gate for the full gain-band of an erbium-doped fiber amplifier (EDFA)(1535 to 1560 nm). This EA gate provides low polarization-dependent loss, higher extinction ratio, and high saturation input power in the wider wavelength range. These MQW EA gates will play an important role in future WDM switching systems.

A fiber Bragg grating, which has periodical perturbation of the refractive index in the fiber core, acts as a wavelength selective reflection filter and steep optical spectrum can be realized by forming more than ten thousand of gratings along the fiber core. Owing to capability of making steep optical spectrum, fiber Bragg gratings has been expected to be introduced practical use as multiplexing or demultiplexing filters in dense WDM transmission systems. On the other hand, radiation mode loss, reflection side mode and temperature dependence of Bragg wavelength, should be improved to put the fiber Bragg grating to practical use in dense WDM transmission systems. In this paper, an optimum design and performance of the fiber Bragg grating for dense WDM systems are described. The photosensitive cladding fiber realized less than 0. 2 dB insertion loss at transmitted signal channels and less than 0. 1 dB splicing loss with standard single-mode fibers. An adequate apodization technique in the refractive index distribution suppressed reflection side modes. A temperature compensating package, which gives longitudinal strain with negative temperature dependence to a fiber Bragg grating, minimized temperature dependence of Bragg wavelength less than 0. 001 nm/. Thermal decay of Bragg grating was also investigated and adequate annealing condition was estimated to obtain sufficient stability for practical use in dense WDM transmission.

An optical communications technology roadmap leading up to the second decade of the 21st century has been investigated to provide a future vision of the optoelectronic technology in 15 to 20 years. The process whereby technology may progress toward the realization of the vision is indicated. A transmission rate of 100 Mbps for homes and a rate of 5 Tbps for the backbone network will be required in the first decade of the 21 century. Two technology roadmaps for public and business communications networks are discussed. It is concluded both WDM and TDM technology will be required to realize such an ultra-high capacity transmission. Technical tasks for various optical devices are investigated in detail.

We introduce an image contour clustering method based on a multiscale image representation and its application to image compression. Multiscale gradient planes are obtained from the mean squared sum of 2D wavelet transform of an image. The decay on the multiscale gradient planes across scales depends on the Lipshitz exponent. Since the Lipshitz exponent indicates the spatial differentiability of an image, the multiscale gradient planes represent smoothness or sharpness around edges on image contours. We apply vector quatization to the multiscale gradient planes at contours, and cluster the contours in terms of represntative vectors in VQ. Since the multiscale gradient planes indicate the Lipshitz exponents, the image contours are clustered according to its gradients and Lipshitz exponents. Moreover, we present an image recovery algorithm to the multiscale gradient planes, and we achieve the skech-based image compression by the vector quantization on the multiscale gradient planes.

An efficient active noise control algorithm based on the lattice-transversal joint (LTJ) filter structure is presented, and applied to the active control of broadband noise in a 3-dimensional enclosure. The presented algorithm implements the filtered-x LMS within the LTJ structure obtained by cascading the lattice and transversal structures. Simulation results show that the LTJ-based noise control algorithm has fast convergence speed that is comparable to the lattice-based algorithm while its computational complexity is less demanding.

An experiment is reported comparing six pen input strategies for selecting a small target using five diffenent sized targets (1, 3, 5, 7 and 9 dot diameter circles respectively, 0. 36 mm per dot). The results showed that the best strategy, in terms of error rate, selection time and subjective preferences, was the "land-on2" strategy where the target is selected when the pen-tip touches the target for the first time after landing on the screen surface. Moreover, "the smallest maximum size" was determined to be 5 dots (1. 8 mm). This was the largest size among the targets which had a significant main effect on error rate in the six strategies. These results are important for both researchers and designers of pen-based systems.

This paper studies the routing algorithms for multi-destination connections where each destination may require different amount of data streams. This asymmetric feature can arise mostly in a large and/or heterogeneous network environment. There are mainly two reasons for this. One is that terminal equipments may have different capabilities. The other is that users may have various interests in the same set of information. We first define the asymmetric multicast problem and describe an original routing method for this type of multicast. The method is then employed in the presented routing algorithms, which can be run in multi-cluster environment. The multi-cluster architecture is considered to be effective for running routing in the networks, where a variety of operating methods might be applied in different clusters but global network performance is required. Our algorithms are designed based on some classical Steiner tree heuristics. The basic goal of our algorithms is to make routing decisions for the asymmetric multicast connections with minimum-cost purpose. In addition, we also consider delay constraint requirements in the multicast connections and propose correspondent algorithms. We compare the performance between SPT (Shortest Path Tree)-based algorithms and the presented algorithms by simulations. We show that performance difference exists among the different types of the algorithms.

It becomes essential in practice to improve a processing rate and to divide an image into small segments adjusting a limited memory, because image filing systems handle large images up to A1 size. This paper proposes a new method of an automatic skew normalization, comprising a high-speed skew detection and a distortion-free dividing rotation. We have evaluated the proposed method from the viewpoints of the processing rate and the accuracy for typed documents. As results, the processing rate is 2. 9 times faster than that of a conventional method. A practical processing rate for A1 size documents can be achieved under the condition that the accuracy of a normalized angle is controlled within 0. 3 degrees. Especially, the rotation with dividing can have no error angle, even when the A1 size documents is divided into 200 segments, whereas the conventional method cause the error angle of 1. 68 degrees.

The new technique for reducing the load latency is presented. This technique, named tunneling-load, utilizes the register specifier buffer in order to reduce the load latency without fetching the data cache speculatively, and thus eliminates the drawback of any load address prediction techniques. As a consequence of the trend toward increasing clock frequency, the internal cache is no longer able to fill the speed gap between the processor and the external memory, and the data cache latency degrades the processor performance. In order to hide this latency, several techniques predicting the load address have been proposed. These techniques carry out the speculative data cache fetching, which causes the explosion of the memory traffic and the pollution of the data cache. The tunneling-load solves these problems. We have evaluated the effects of the tunneling-load, and found that in an in-order-issue superscalar platform the instruction level parallelism is increased by approximately 10%.

An electroabsorption (EA) modulator array using a double optical-pass (DP) configuration has been developed to obtain high-speed modulation in parallel. Feeding electrical signals from the highly reflective side of the modulator eliminated component assembly problems with lenses and microwave feeder lines. Passive waveguide integration enabled wafers to be cleaved with very short absorbers. The degradation in frequency response was theoretically calculated to be <0. 2 dB compared to that of EA modulators without a passive waveguide. A common upper doping layer in the absorber and passive waveguide regions was introduced to attain high product throughput due to good epitaxial flatness and processing. The integrated 4-channels multiquantum well DP EA modulator array demonstrated high overall performance for a wavelength range from 1545 to 1558 nm. It features a drive voltage of 2 V for 10 dB attenuation, an insertion loss of 12 dB, and 4 channels17 GHz bandwidths for each channel, with low -20 dB crosstalk between adjacent waveguides.

This paper proposes a new approach to recover the sign of local surface curvature of object from three shading images using neural network. The RBF (Radial Basis Function) neural network is used to learn the mapping of three image irradiances to the position on a sphere. Then, the learned neural network maps the image irradiances at the neighbor pixels of the test object taken from three illuminating directions of light sources onto the sphere images taken under the same illuminating condition. Using the property that basic six kinds of surface curvature has the different relative locations of the local five points mapped on the sphere, not only the Gaussian curvature but also the kind of curvature is directly recovered locally from the relation of the locations on the mapped points on the sphere without knowing the values of surface gradient for each point. Further, two step neural networks which combines the forward mapping and its inverse mapping one can be used to get the local confidence estimate for the obtained results. The entire approach is non-parametric, empirical in that no explicit assumptions are made about light source directions or surface reflectance. Results are demonstrated by the experiments for real images.

Recently, a great deal of effort has been devoted to the design problem of "constrained least squares M-D FIR filter" because a significant improvement of the squared error is expected by a slight relaxation of the minimax error condition. Unfortunately, no design method has been reported, which has some theoretical guarantee of the convergence to the optimal solution. In this paper, we propose a class of novel design methods of "constrained least squares M-D FIR filter. " The most remarkable feature is that all of the proposed methods have theoretical guarantees of convergences to the unique optimal solution under any consistent set of prescribed maximal error conditions. The proposed methods are based on "convex projection techniques" that computes the metric projection onto the intersection of multiple closed convex sets in real Hilbert space. Moreover, some of the proposed methods can still be applied even for the problem with any inconsistent set of maximal error conditions. These lead to the unique optimal solution over the set of all filters that attain the least sum of squared distances to all constraint sets.

The paper obtains an algorithm to estimate the irregular sampling in wavelet subspaces. Compared to our former work on the problem, the new estimate is relaxed for some wavelet subspaces.