This paper describes a highly performance scalable video coder. Wavelet transform is employed to decompose the video frame into different resolutions. Novel features of this coder are 1) a highly efficient multi-resolution motion estimation that requires minimum compuation and overhead motion information is embedded in this scheme; 2) the wavelet coefficients are organized in an extended zero tree (EZT) which is much more efficient than the simple zerotree. We show with experimental results that this video coder achieves good performances both in processing time and compression ratio when applied to typical test video sequences.

Alternative wiring techniques have been shown to be very useful for many EDA problems. The currently used rewiring techniques are mainly ATPG based. In this paper, we study the approach of applying purely graph-based local pattern search methods in locating alternative wires. The method searches minimal graph patterns containing alternative wires that limited to 2 edges distant from the target wire. The experimental result shows that this scheme is very fast and has the advantage of searching both the nearby forward and backward alternative wires easily. The overall number of alternative wires searched is quite comparable (104%), compared to the forward search only RAMBO version, and the CPU time is 200 times faster. We also illustrate its usage, among many others, by a simple coupling with the SIS algebraic operations and let this rewiring tool serve as a netlist-perturbing engine for logic minimization. The coupling scheme shows a further reduction of 8.5% in area compared to applying algebraic script alone, with a nearly negligible CPU overhead spent in rewiring.

We have reviewed recent progress on arrayed waveguide gratings for DWDM applications. AWGs can be used to realize not only mux/demux filters with various channel spacings, but also highly integrated optical components.

This paper presents the results on IF/baseband up/down direct digital conversion and multiple channel analysis/synthesis software defined radio modules which are implemented using high speed ADC, DAC and FPGA, for IS-95 code division multiple access (CDMA) systems. The implemented system can directly down-convert multiple channel IS-95 CDMA IF signals to the baseband, and selectively analyze specific channel signals based on polyphase analysis filter bank techniques. Moreover, the analyzed baseband signals of multiple channels can be directly up-converted and synthesized in the same system. We have deployed the implemented system in IS-95 CDMA optical digital repeaters for PCS applications.

Photonic crystals have optical properties characterized by photonic bandgap, large anisotropy and high dispersion, which can be applied to various optical devices. We have proposed an autocloning method for fabricating 2D or 3D photonic crystals and are developing novel structures and functions in photonic crystals. The autocloning is an easy process based on the combination of sputter deposition and sputter etching and is suitable for industry. We have already demonstrated devices or functions such as polarization splitters and surface-normal waveguides. In this paper, we describe our latest work on photonic crystals utilizing the autocloning technology. Phase plates and polarization selective gratings for optical pick-ups are demonstrated utilizing TiO2/SiO2 photonic crystals. The technology to introduce CdS into 3D photonic crystals is also developed and photoluminescence from the introduced CdS is observed, which is the first step to realize luminescent devices with 3D confinement or high polarization controllability.

We studied theoretically and experimentally an InGaAs/InAlAs/InP polarization-insensitive multiple quantum well (MQW) electroabsorption (EA) modulator operating over a very wide wavelength range in 1.55 µm wavelength region. One of the simplest possible potential-tailored quantum well, "pre-biased" quantum well (PBQW) is used to achieve wide-wavelength polarization insensitivity. PBQW is basically a rectangular quantum well with a thin barrier inserted near one edge of well. This thin barrier effectively introduces "pre-bias" to a rectangular quantum well and the same amount of Stark shift is achieved for electron-heavy hole and electron-light hole transition energies. By incorporating tensile strain into PBQW, polarization-insensitive modulation is achieved over 60 nm wavelength range, from 1510 nm to 1570 nm. This MQW-EA modulator plays an important role in wavelength division multiplexing (WDM) transmission and switching systems.

A 160 GHz colliding-pulse mode-locked laser diode (CPM-LD) was stabilized by injection of a stable master laser pulse train repeated at a 16th-subharmonic-frequency (9.873 GHz) of the CPM-LD's mode-locking frequency. Synchroscan steak camera measurements revealed a clear pulse train with 16-times repetition frequency of the master laser pulse train for the stabilized CPM-LD output, indicating that CPM-LD output was synchronized to the master laser and that the timing jitter was also reduced. The timing jitter of the stabilized CPM-LD was quantitatively evaluated by an all-optical down converting technique using the nonlinearity of optical fiber. This technique is simple and has a wider bandwidth in comparison to a conventional technique, making it possible to accurately measure the phase noise of ultrafast optical pulse train when its repetition frequency exceeds 100 GHz. The electrical power spectra measurements indicated that the CPM-LD's mode-locking frequency was exactly locked by the injection of the master laser pulse train and that the timing jitter decreased as the injection power increased. The timing jitter was reduced from 2.2 ps in free running operation to 0.26 ps at an injection power of 57 mW, comparable to that of the master laser (0.21 ps).

In this paper, we propose the majority algorithm to choose the connection weights for the neural networks with quantized connection weights of 1 and 0. We also obtained the layered network to solve the parity problem with the input of arbitrary number N through an application of this algorithm. The network can be expected to have the same ability of generalization as the network trained with learning rules. This is because it is possible to decide the connection weights, regardless of the size of the training set. One can decide connection weights without learning according to our case study. Thus, we expect that the proposed algorithm may be applied for a real-time processing.

In this paper, a new network structure called generalized Hierarchical Completely-Connected networks (HCCs) is proposed, and its properties and features are evaluated. Simple routing strategies for HCCs are also developed for shortest-paths routing algorithms. A set of HCCs constructed by the proposed method includes some conventional hierarchical networks, then it is called generalized one. The construction of an HCC starts from a basic block (a level-1 block) which consists of n nodes of constant degree. Then a level-h block for h 2 is constructed recursively by interconnecting any pair of macro nodes (n level-(h-1) blocks) completely. An HCC has a constant node-degree regardless of an increase in its size (the number of nodes). Furthermore, since an HCC has a hierarchically structured topology and the feature of uniformity, a wide variety of inter-cluster connections is possible. Evaluation results show that an HCC is suitable for very large computer systems.

Taking traveling salesman problems (TSPs) as examples of combinatorial optimization problems, an "optimal" Hopfield network for ("optimal" neural representation of) TSPs is presented, where a vertex of state hypercube of the network is asymptotically stable if and only if it is an optimal solution. Of all the Hopfield networks for TSPs, this network most sharply distinguishes an optimal solution from other nonoptimal solutions and infeasible solutions. In this sense, we call this network "optimal" for TSPs. Whenever the network converges to a vertex, we can always obtain an optimal solution. However, we can not design such network without knowing an optimal solution to the problem. So, its approximate realization, which can be designed without a-priori knowledge of an optimal solution, is proposed. Simulations show that the "optimal" network and its approximate realization obtain optimal or good feasible solutions more frequently than familiar Hopfield networks. We can also design such "optimal" Hopfield networks for many combinatorial optimization problems as well as for TSPs.

A large-scale primarily public system requires in addition to high reliability, a broad range of applications from control to information services. As construction is phased-in this system must be flexible, changeable and able to grows as the needs arise. However, a changing a system may lead to loss of reliability. A system that is able to change and grow in a reliable and stable manner is called an assurance system and the technology it uses is called assurance technology. This paper describes the basic technology, phased-in system construction and so on of assurance technology based on an autonomous decentralized system. It further discusses application of assurance technology to ATOS as an example of a large-scale transport operation control system. Note: ATOS; Autonomous Decentralized Transport Operation Control System

In supervised learning, one of the major learning methods is memorization learning (ML). Since it reduces only the training error, ML does not guarantee good generalization capability in general. When ML is used, however, acquiring good generalization capability is expected. This usage of ML was interpreted by one of the present authors, H. Ogawa, as a means of realizing 'true objective learning' which directly takes generalization capability into account, and introduced the concept of admissibility. If a learning method can provide the same generalization capability as a true objective learning, it is said that the objective learning admits the learning method. Hence, if admissibility does not hold, making it hold becomes important. In this paper, we introduce the concept of realization of admissibility, and devise a realization method of admissibility of ML with respect to projection learning which directly takes generalization capability into account.

Localization properties of mode waves in an off-diagonally disordered waveguide system are presented. The disorder is introduced by taking spacings between cores to be random variables. Coupled mode equations are transformed into a matrix eigenvalue problem and eigenvalues and eigenvectors are numerically obtained. Correspondences between the natures of modes and the modal density of states are discussed. The system is divided into several sections which behave effectively as isolated systems. Modes in the entire system are a superposition of modes associated with the sections. A section is divided into several elements, which do not only behave apparently as isolated systems but also couple with each other. When an element includes two cores coupled strongly with each other due to a narrow spacing, modes are strongly localized there. The extent of the modes is almost independent of the disorder of the system. In a system with small disorder strongly localized modes can exist. The modes appear outside the propagation constant band of the ordered system composed of identical cores of equal spacing. Modes near the center of the band are extended over a number of elements and have the relatively large extent. Many modes appear near the center of the band and the modal density of states has a sharp peak there.

This paper studies mutually coupled integrate-and-fire type chaotic oscillators. The coupling is realized by impulsive switchings and the system exhibits various synchronous and asynchronous phenomena. We give a basic classification of the chaos synchronization phenomena and their breakdown patterns. The stability of the synchronous states can be confirmed using the piecewise exact solutions, and the basic mechanism of the phenomena can be elucidated by a simple geometric consideration. The typical phenomena are confirmed in the laboratory.

We determine the optimum time TOPT to order a spare part for a system before the part in operation has failed. TOPT is a function of the part's failure-time distribution, the lead (delivery) time of the part, its inventory cost, and the cost of downtime while waiting delivery. The probabilities of the system's up and down states are obtained from a non-regenerative stochastic Petri net. TOPT is found by minimizing E[cost], the expected cost of inventory and downtime. Three cases are compared: 1) Exponential order and lead times, 2) Deterministic order time and exponential lead time, and 3) Deterministic order and lead times. In Case 1, it is shown analytically that, depending on the ratio of inventory to downtime costs, the optimum policy is one of three: order a spare part immediately at t = 0, wait until the part in operation fails, or order before failure at TOPT > 0. Numerical examples illustrate the three cases.

Out-of-head sound localization achieved via binaural earphones is indispensable for a virtual sound system. It is necessary to measure the two transfer functions of each subject, Spatial Sound Transfer Function (SSTF) and Ear Canal Transfer Function (ECTF), for achieving sound localization. It is well known that the quality of sound localization may be poor if the individual transfer functions are not accurately reproduced. This is because each subject has his/her own transfer functions. It is very important to clarify which function includes more individual information, SSTF or ECTF, in order to implement a simpler model. Therefore, we introduce the quantity of "Personal differences" for investigating the subject's transfer functions included in SSTF and ECTF. We measure the transfer functions SSTF and ECTF of 60 subjects in a soundproofed room, and analysis of the data using Principal Component Analysis (PCA) and three subjective assessment tests. This study finds that ECTF differs more widely from person to person than SSTF.

This paper presents a heuristic algorithm called CLASSIC for the minimization of the control memory width in microprogrammed processors or the instruction memory width of application-specific VLIW (Very Long Instruction Word) processors. CLASSIC results in nearly optimal solutions with the time complexity of O(n2), where n denotes the number of microoperations. In this paper, we also propose the so-called incompleteness relations which are exploited for the minimization of the control memory width. Experiments using various examples have shown that CLASSIC always achieves smaller microprogram widths compared to the earlier techniques based on the maximal compatibility class or the minimal AND/OR set. The results show that CLASSIC can reduce the control memory width by 34.2% on average compared with a heuristic compatibility class algorithm.

This paper proposes a novel cache architecture suitable for merged DRAM/logic LSIs, which is called "dynamically variable line-size cache (D-VLS cache). " The D-VLS cache can optimize its line-size according to the characteristic of programs, and attempts to improve the performance by exploiting the high on-chip memory bandwidth on merged DRAM/logic LSIs appropriately. In our evaluation, it is observed that an average memory-access time improvement achieved by a direct-mapped D-VLS cache is about 20% compared to a conventional direct-mapped cache with fixed 32-byte lines. This performance improvement is better than that of a doubled-size conventional direct-mapped cache.

In this paper, we develop an effective smoothing technique to estimate the optimal software release schedule which minimizes the total software cost. The optimal software release problem is essentially reduced to a statistical estimation problem for the software failure rate, but the resulting estimator based on both the fault-detection time data observed in testing phase and its estimate in future is discontinuous and does not always function well for determining the optimal release schedule. We estimate the smoothed software failure rate using the usual quadratic programming approach and generate the optimal software release schedule with higher accuracy.

High accuracy, high reliability, and high performance have to be simultaneously satisfied to achieve high productivity of the latest processing equipment. High flexibility is also required because many options are available and processing equipment is modified frequently. A high-assurance-system (HAS) model for processing equipment has been developed according to the concept of an Autonomous Decentralized System (ADS). Heterogeneous devices, that have same function and diverse qualities, are utilized to assure the different requirements of high accuracy, high reliability, and high performance simultaneously. The Data Property (DP) and Assurance Manager (AM) are proposed in this model. Different accuracy, reliability, and performance indices characterize each device, and the DP describes the differences of the properties of the data transmitted from these heterogeneous devices. The AM assures not only high reliability but also high performance and high accuracy by utilizing the heterogeneity of data described by the DP. The HAS model was applied to a device-level system used in processing equipment, and its effectiveness was verified by simulating a pressure-control system.