This paper proposes an on-chip memory-path architecture employing the dynamically variable line-size (D-VLS) cache for high performance and low energy consumption. The D-VLS cache exploits the high on-chip memory bandwidth attainable on merged DRAM/logic LSIs by replacing a whole large cache line in one cycle. At the same time, it attempts to avoid frequent evictions by decreasing the cache-line size when programs have poor spatial locality. Activating only on-chip DRAM subarrays corresponding to a replaced cache-line size produces a significant energy reduction. In our simulation, it is observed that our proposed on-chip memory-path architecture, which employs a direct-mapped D-VLS cache, improves the ED (Energy Delay) product by more than 75% over a conventional memory-path model.

This paper describes a method to normalize the lip position for improving the performance of a visual-information-based speech recognition system. Basically, there are two types of information useful in speech recognition processes; the first one is the speech signal itself and the second one is the visual information from the lips in motion. This paper tries to solve some problems caused by using images from the lips in motion such as the effect produced by the variation of the lip location. The proposed lip location normalization method is based on a search algorithm of the lip position in which the location normalization is integrated into the model training. Experiments of speaker-independent isolated word recognition were carried out on the Tulips1 and M2VTS databases. Experiments showed a recognition rate of 74.5% and an error reduction rate of 35.7% for the ten digits word recognition M2VTS database.

In this paper, visualized sound retrieval and categorization methods using a feature-based image search engine were evaluated aiming at efficient video scene query. Color-coded patterns of the sound spectrogram are adopted as the visualized sound index. Sound categorization experiments were conducted using visualized sound databases including speech, bird song, musical sounds, insect chirping, and the sound-track of sports video. The results of the retrieval experiments show that the simple feature-based image search engine can be effectively used for visualized sound retrieval and categorization. The results of categorization experiments involving humans show that after brief training humans can at least do rough categorization. These results suggest that using visualized sound can be effective method for an efficient video scene query.

The aim of this study is to examine the effectiveness of various open-ended resonators. According to the required filter responses, the application to microwave filters based on presented open-ended resonators is systematically examined as well. First, the resonance property of the basic open-ended resonator is discussed based on even-and odd-mode analysis. The intrinsic property of a tapped open-ended resonator is also discussed here. Second, the basic properties of a stepped impedance resonator (SIR) and a loaded-element resonator are examined theoretically for improvement of spurious responses and the dual-passband response. The basic operations of these resonators are also explained based on even- and odd-mode analysis. Examples for filter applications based on presented resonators are also provided. We found that the intrinsic properties of the open-ended resonators are very useful for practical filter responses.

This letter describes two adaptive timing synchronization schemes for a short-ranged Bluetooth system in the partial-band noise environments. One estimates the variance of the partial-band interference, which is utilized for the trigger threshold value of the inquiry scan and page scan states, while second is designed using the scaled partial correlation value for the connection state. Numerical results show the proposed synchronization algorithms are robust to the partial-band noise interference and of low complexity, which is suitable for a low-cost personal area network (PAN).

To enhance fault tolerance ability of the feedforward neural networks (NNs for short) implemented in hardware, we discuss the learning algorithm that converges without adding extra neurons and a large amount of extra learning time and cycles. Our algorithm modified from the standard backpropagation algorithm (SBPA for short) limits synaptic weights of neurons in range during learning phase. The upper and lower bounds of the weights are calculated according to the average and standard deviation of them. Then our algorithm reupdates any weight beyond the calculated range to the upper or lower bound. Since the above enables us to decrease the standard deviation of the weights, it is useful in enhancing fault tolerance. We apply NNs trained with other algorithms and our one to a character recognition problem. It is shown that our one is superior to other ones in reliability, extra learning time and/or extra learning cycles. Besides we clarify that our algorithm never degrades the generalization ability of NNs although it coerces the weights within the calculated range.

In this paper, we study decentralized supervisory control of discrete event systems where local disabling actions are fused by the OR rule. We generalize an on-line procedure for synthesizing decentralized supervisors proposed by Prosser. By using the generalized procedure, we can achieve a sublanguage of a specification which is not achieved by a class of decentralized supervisors synthesized by the Prosser's procedure.

This paper proposes aspiration controls which restrains aspiration branches and keeps the original tabu-based searching by considering past and/or (predicted) future searching profiles. For implementation of the aspiration control we employ not only the short-term and long-term memory but also future memory which is first introduced in this paper as a new concept in the tabu search field. The tabu search with the aspiration control is also parallelized. Moreover two types of parallel cooperative searching scheme are proposed. Through computational experiment, we observe efficiency of our approach comparing to the traditional ones. Especially, we find that cooperative searching has possibility to improve the solution quality very well.

The expected lengths of the parsed segments obtained by applying Lempel-Ziv incremental parsing algorithm for i.i.d. source satisfy simple recurrence relations. By extracting a combinatorial essence from the previous proof, we obtain a simpler derivation.

Based on genetic algorithm (GA) in this paper we present a simple method to extract distributed circuit parameters of a multiple coupled nonuniform microstrip transmission lines from it's measured or computed S-parameters. The lines may be lossless or lossy, with frequency dependent parameters. First a sufficient amount of information about the system is measured or computed over an specified frequency range. Then this information is used as an input for a GA to determine the inductance and capacitance matrices of the system. The theory used for fitness evaluation is based on the steplines approximation of the nonuniform transmission lines and quasi-TEM assumptions. Using steplines approximation the system of coupled nonuniform transmission lines is subdivided into arbitrary large number of coupled uniform lines (steplines) with different characteristics. Then using modal decomposition method the system of coupled partial differential equations for each step is decomposed to a number of uncoupled ordinary wave equations which are then solved in frequency-domain.

An iterative decoder of turbo code over an inter-symbol interference channel is proposed. A component decoder realizes decoding and equalization simultaneously with the soft-output Viterbi algorithm (SOVA). A decoding algorithm and simulation results are shown.

In this paper we study general complex eigenvalue problems in engineering fields. The eigenvalue problems can be transformed into the associated problems for solving large systems of nonlinear ordinary differential equations (dynamic equations) by optimization techniques. The known waveform relaxation method in circuit simulation can then be successfully applied to compute the resulting dynamic equations. The approach reported here, which is implemented on a message passing multi-processor system, can determine all eigenvalues and their eigenvectors for general complex matrices without any restriction on the matrices. The numerical results are provided to confirm the theoretical analysis.

This paper investigates noise enhancement factors of a zero-forcing detector and a decision feedback detector for synchronous Multiple Input Multiple Output (MIMO) channels. It is first shown that the zero-forcing and decision feedback detectors can be implemented in a vector digital filter form, and the noise enhancement factors with the detectors can easily be calculated by using the vector digital filter form. This paper then applies the zero-forcing and decision feedback detectors to the signal detection of a frequency-overlapped multicarrier signaling (FOMS) system. The normalized noise enhancement factor, which is given as a product of the noise enhancement and bandwidth reduction factors, is shown to be smaller with the decision feedback detector than the zero-forcing detector. Results of computer simulations conducted to evaluate bit error rate (BER) performances with the two detectors are also shown together with the BER performance with a conventional channel-by-channel detector.

In this paper, we propose an adaptive IIR equalizer based on prefiltering techniques. The proposed equalizer has a cascade structure of an ARMA prefilter and an adaptive FIR equalizer. The ARMA prefilter is designed based on the transfer function estimated by the gradient-type instrumental variable algorithm. Simulation results are shown to confirm the performance of the proposed adaptive IIR equalizer.

A WDM (Wavelength Division Multiplexing) technology is a new optical technology, providing multiple wavelengths at the rate of 10 Gbps on the fiber. IP (Internet Protocol) over WDM networks where IP packets are directly carried on the WDM network is expected to offer an infrastructure for the next generation Internet. For IP over WDM networks, a WDM protection mechanism is expected to provide a highly reliable network (i.e., robustness against the link/node failures). However, conventional IP also provides a reliability mechanism by its routing function. In this paper, we first formulate an optimization problem for designing IP over WDM networks with protection functionalities of WDM networks, by which we can obtain IP over WDM networks with high reliability. Our formulation results in a mixed integer linear problem (MILP). However, it is known that MILP can be solved only for a small number of variables, in our case, nodes and/or wavelengths. We therefore propose two heuristic algorithms, min-hop-first and largest-traffic-first approaches in order to assign the wavelength for backup lightpath. Our results show that the min-hop-first approach takes fewer wavelengths to construct the reliable network, that is, all of lightpaths can be protected using the WDM protection mechanism. However, our largest-traffic-first approach is also a good choice in the sense that the approach can be saved the traffic volume increased at the IP router by the link failure.

The alignment of a nematic liquid crystal (LC) parallel to the polarization direction of the laser could be induced by three types of polyimide (PI) films, a PI based on aromatic dianhydride and two PIs on alicyclic dianhydride, exposed to polarized pulsed laser at 266 nm at high fluence in air. The UV-visible absorption spectra of the PI films showed that a remarkable chemical change occurred after exposure at the high fluence in air. In contrast, in argon atmosphere the PI based on aromatic dianhydride was radiation-resistant and the exposed PI film could induce alignment of the LC molecules parallel to laser polarization. We estimate that the mechanism of the parallel alignment observed in argon is not the photodegradation but the orientation of the PI molecules.

We present novel semiconductor technologies of ZnO epitaxial films with using laser molecular-beam epitaxy method. Exciting optical properties such as room temperature lasing in ZnO nanocrystalline films and quantum size effects in ZnO/MgxZn1-xO superlattices were observed. By developing crystalline quality with using lattice-matched substrates, we could control resistivity of the doped ZnO films from 10-3 Ωcm to 104 Ωcm. These results would provide us an opportunity to construct a monolithic array consisted of light emitting devices and field effect transistors towards a possible flat panel display.

We investigate possibility of fault-tolerant and self-stabilizing protocols (ftss protocols) using an unreliable failure detector. Our main contribution is (1) to newly introduce k-accuracy of an unreliable failure detector, (2) to show that k-accuracy of a failure detector is necessary for any ftss k-group consensus protocol, and (3) to present three ftss k-group consensus protocols using a k-accurate and weakly complete failure detector under the read/write daemon on complete networks and on (n-k+1)-connected networks, and under the central daemon on complete networks.

Loop optimization plays an important role in compiler optimization and program transformation. Many sophisticated techniques such as loop-invariance code motion, loop restructuring and loop fusion have been developed. This paper introduces a novel technique called loop quasi-invariance code motion. It is a generalization of standard loop-invariance code motion, but based on loop quasi-invariance analysis. Loop quasi-invariance is similar to standard loop-invariance but allows for a finite number of iterations before computations in a loop become invariant. In this paper we define the notion of loop quasi-invariance, present an algorithm for statically computing the optimal unfolding length in While-programs and give a transformation method. Our method can increase the accuracy of program analyses and improve the efficiency of programs by making loops smaller and faster. Our technique is well-suited as supporting transformation in compilers, partial evaluators, and other program transformers.

With the continuing growth of the World Wide Web (WWW) services over the Internet, the demands for rapid image transmission over a network link of limited bandwidth and economical image storage of a large image database are increasing rapidly. In this paper, a classified binary-tree-structured Self-Organizing Feature Map neural network is proposed to design image vector codebooks for quantizing images. Simulations show that the algorithm not only produces codebooks with lower distortion than the well-known CVQ algorithm but also can minimize the edge degradation. Because the adjacent codewords in the proposed algorithm are updated concurrently, the codewords in the obtained codebooks tend to be ordered according to their mutual similarity which means more compression can be achieved with this algorithm. It should also be noticed that the obtained codebook is particularly well suited for progressive image transmission because it always forms a binary tree in the input space.