This paper proposes a new error backpropagation method (DBP) for a three-layered neural network containing a nondifferentiable binary output unit. In contrast to the conventional simple perceptron, in which the teacher signal is given only to the output layer, in the DBP method the teacher signal is also given to the middle layer so that the output error is decreased. Consequently, it is possible in the DBP method to correct the coupling weights in both the lower layer and the upper layer. This makes it easy to construct a network composed only of binary output units, which results in high-speed operation and is suitable for hardware implementation. When the DBP method is applied to linearly inseparable tasks such as XORing, the learning performance is greatly improved compared to learning by the simple perceptron, and almost the same learning performance as the conventional BP is obtained. © 2007 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(5): 40– 49, 2007; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/ecjc.20318
With the dramatic advances in mobile communication technologies, the problem of estimating the direction of arrival (DOA) of signals generated by several sources becomes even more important. High-resolution estimation methods based on eigenvalue decomposition include the minimum-norm method and the MUSIC method. Since the peak of the angular spectrum must be scanned over all directions, the computational effort is enormous. In this paper, a DOA estimation method using a database is proposed. In this method, a database is constructed in which the eigenvector corresponding to the minimum eigenvalue is the key and the DOA is the associated value. The DOA is estimated by seeking in the database the data with the key closest to the eigenvector obtained from the observation signals. In the proposed method, DOA estimation can be achieved by only a regional search of the database, and hence the process is extremely fast. Finally, the effectiveness of the proposed method is demonstrated by simulation. © 2006 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(3): 48–56, 2007; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/ecjc.20251
This paper considers a vehicle-borne target detection system for short range in which an UWB (Ultra Wideband) impulse radar is used. The UWB impulse radar uses a pulse with widths of less than nanosecond order. Therefore, many multipath returns are observed on the receiving side. An IPCP (Inter-Period Correction Processing) receiver that resolves complexity of the receiver caused by multiple paths has been proposed. However, the integration time of the correlator in the IPCP receiver is limited by the period of the transmission signal (code length) and affects the range measurement accuracy. This paper considers vehicle mounting for applications. In order to assure safety, a receiver is constructed with several IPCP receivers in parallel, so that the range measurement accuracy is improved. In addition, a receiver which reduces the complexity of threshold setting in the conventional IPCP receiver when several targets are detected is proposed. The outputs of these receivers are derived analytically and improvement of the range measurement accuracy is confirmed. Further, based on the analytical results, the output of each receiver is characterized and the relationship of the threshold, detection probability, and false error probability is derived. Finally, the characteristics of use of the receiver as a vehicle-mounted radar are evaluated by simulation. © 2007 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(7): 1–14, 2007; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/ecjc.20303
Irradiation of ultrasonic waves induces cavitation in a solution, and a chemical action instigated by radicals occurs. This chemical action is reported to be most efficient in the region of several hundred kilohertz frequencies. Application of sonochemistry based on ultrasonic chemical action remains at the laboratory scale and has not yet reached the industrial practical scale. In order to realize practical applications of sonochemistry, there is a need to scale-up the sonochemical reactors. For this purpose, we used a 490-kHz cylindrical sonochemical reactor that is long in the irradiation direction, and used potassium iodide (KI) dosimetry, sonochemical luminescence, and calorimetry to evaluate the sonochemical reaction efficiency and reaction field of the reactor. We performed the evaluation with the ultrasonic irradiation direction in the horizontal and vertical directions. As a result, we observed sonochemical luminescence and a rise in temperature near the reflection surface, for both the horizontal and vertical irradiation. For the horizontal irradiation, the 13 concentration was evenly distributed in the irradiation direction. Moreover, we found that the sonochemical reaction efficiency was the same for both the horizontal irradiation and the vertical irradiation. (C) 2007 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(8): 1-8, 2007; Published online in Wiley InterScience (www.interscience.wiley.com).
There has recently been much discussion of MPLS multicast approaches. MMT and BLAST-Cast have been proposed as means of establishing multicast LSP using unicast LSP. These techniques are useful not only as transfer techniques in introducing the MPLS multicast technique, but also for the deployment of MPLS multicasting in established MPLS networks, since the scale can be expanded stepwise, and the cost of deployment and operation can be reduced in the deployment of MPLS multicast in established MPLS networks. However, there have been few studies of the detailed signaling protocol to be used in such an approach, or of problems accompanying these techniques. Consequently, we investigated a signaling protocol that can be applied to these approaches, and devised an MPLS multicast mechanism using RSVP-TE. The proposed mechanism was implemented and evaluate. This paper discusses the design and implementation of the proposed mechanism, together with its evaluation. (C) 2007 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(11): 40-49, 2007; Published online in Wiley InterScience (www.interscience.wiley.com).
Median filters and other nonlinear filters have been investigated for restoration of degraded images with Poisson noise. Recently, subband image restoration using the wavelet transform has been attracting much attention. This method is effective for small-amplitude noise, but in the case of Poisson noise, large-amplitude noise exceeds the preset threshold and is not removed. In this study, we propose a new method of noise removal from degraded images with Poisson noise by using a combination of the M-transformation [5] and the wavelet BayesShrink method. (C) 2007 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3. 90(11): 11-20, 2007; Published online in Wiley InterScience (www.interscience.wiley.com).
Face detection from images is a complex and non-linear problem due to the various kinds of face images. This problem is solved by conversion of the original feature vectors extracted from images into high-dimension feature vectors using nonlinear mapping, and then finding face/nonface discriminant functions in the mapping space. If such discriminant functions are based on the inner products of high-dimension vectors, such inner products can be easily obtained by substitute calculations of kernel functions in the original feature space. However, in conventional recognition algorithms using kernel functions, numerous features are required to improve recognition accuracy. This paper proposes a new face detection on the basis of Kullback-Leibler divergence (KLD). KLD refers to a distance between the distributions of face and nonface data for certain features. Features with large KLD are used for face detection. Moreover, by evaluating the features based on their KLDs, we can generate new features, and deal with different kinds of features concurrently. In experiments, a classifier designed by the proposed method achieved high recognition performance, while using few features. (C) 2007 Wiley Periodicals, Inc.
Coordinated checkpointing techniques ensure that a consistent global state is maintained by means of coordination between processes. The approach requires that application messages temporarily cease to be exchanged but the rollback procedure when recovering from a fault is consequently simplified and the recovery costs are small. With current reductions in communications costs, the importance of coordinated techniques may be seen to be growing. However, in large-scale systems there is a possibility that performance will be seriously impaired due to the frequent halting of the exchange of messages. In this paper we propose a method whereby coordination is performed at only a subset of the checkpoint generation points that are periodically visited while at the remaining points each process independently generates an incremental snapshot. This method aims to both alleviate the performance degradation incurred from coordination and to realize relatively high-speed recovery. In evaluating the effectiveness of this method we estimate the checkpointing overheads and recovery costs using a probabilistic model and simulations and compare it with existing coordination methods. The results show that the proposed method is more effective than existing coordination methods from the perspective of both performance and reliability in environments with a relatively low frequency of messages. In addition, we perform comparisons of two different delta schemes for representing the incremental snapshots and discuss which environments they are each respectively suited to. © 2007 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(8): 39– 53, 2007; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.20296
A structure is proposed for an adaptive system with an adaptive filter located before the unknown system (pre-inverse adaptive system) for estimation of the inverse of the transfer function (inverse transfer function) of the unknown system. In general, when an adaptive transversal filter is used as an adaptive filter, the delay signal of the output of the unknown system is needed in the adaptive algorithm for the weights. Since an adaptive filter is inserted in the front stage, this signal cannot be observed, so that a replica of the unknown system is needed. In this paper, an adaptive system that does not require this replica is discussed. Estimation of the inverse transfer function of the minimum phase of the unknown system is performed by an adaptive exponential filter and an inverse copy of the weights of the exponential filter placed in front of the unknown system. The signal within the adaptive algorithm consists of the observable input signal to the adaptive exponential filter and the estimation error. Estimation of the inverse transfer function for the allpass transfer function of the unknown system is performed by the adaptive transversal filter and the reversing copy of the weight to the transversal filter located before the unknown system. The signal in the adaptive system consists of the observable input signal to the exponential filter and the estimation error. Convergence of the weight is studied from the point of view of monotonic increase of the gradient. The unique feature of the approach is that the algorithm of the two adaptive filters consists of a gradient algorithm with guaranteed convergence for the weights and of copies of the weights after updating. Finally, a performance evaluation of the adaptive system and a comparison with conventional systems are performed by numerical simulation. © 2007 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(9): 10– 17, 2007; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.20305
Existing data-driven processors offer the advantages of being able to naturally resolve the problems inherent in a wide variety of granularities of parallelism, and of being able to multiprocess without overhead. However, because an instruction is not issued until the instruction that generates the source operand finishes executing, a delay equal to the number of stages in the pipeline occurs before an instruction that has a data-dependency relationship can be executed. As a result, portions of sequential processing in a program become a bottleneck. The authors are therefore in the process of developing an LSI processor that is able to retain the advantages of conventional data-driven processors while being able to efficiently execute blocks of sequential processing. This paper presents and evaluates the performance of the proposed execution model and pipeline structure of this processor. This processor uses the same pipeline to perform instruction level multiprocessing of data-driven programs that issue instructions based on the data-dependency relationships and control-driven programs that issue instructions sequentially based on a program counter. The effectiveness of parallel processing by data-driven programs, improvements in the efficiency of sequential processing by introducing control-driven programs, and the ability to evenly share the pipeline when data-driven and control-driven programs are multiprocessing are evaluated. © 2007 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(12): 36–49, 2007; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.20358
This paper proposes a novel data hiding method for JPEG 2000 coded images that embed multi-level information into quantized discrete wavelet transformed coefficients. Since the proposed method hides information represented as an integer to a transformed coefficient rounded to an integer, a JPEG 2000 code-stream conveying data keeps its standard JPEG 2000 code-stream structure. The proposed method is able to extract hidden data Without memorizing embedding positions; it thus Chooses transformed coefficients to which data is hidden freely. This characteristic makes the proposed method suitable for hiding data to a JPEG 2000 coded image with consideration of Region of Interest (ROI) coding that is a major feature. Simulation results show the effectiveness of the proposed method. (c) 2007 Wiley Periodicals, Inc.
This paper is intended to develop an LSI that restores in real time an image pattern with defects and contamination. To this end, based on the principle of shape formation in biological bodies, a reaction-diffusion model that can generate and repair Turing patterns, such as stripes and spots seen in fingerprint and body marks, is proposed. A parallel architecture of the reaction-diffusion chips incorporating this model is presented. By means of circuit simulation, it is shown that this LSI can restore stripe (spot) patterns. (C) 2006 Wiley Periodicals, Inc.
In this paper the authors propose a program allocation method for the purpose of achieving real-time processing using a chip multiprocessor-type data-driven processor. The data-driven processor can perform parallel processing of executable processes without interference as long as the pipeline is not overloaded. The turnaround time for the individual processes being subjected to multiprocessing is independent of other processes being executed at the same time, and is maintained as is for the execution time when a single process is executed. Therefore, real time for a program can be estimated, and if the program can be allocated to a process such that overload does not occur, then real-time processing can be achieved without scheduling during execution. In this paper the authors first formalize a program allocation method in a data-driven processor. They then propose a combined method using genetic algorithms and a heuristic algorithm that uses lists as a program allocation method that is better at satisfying the time constraints and offers greater scalability for large-scale program and multiprocessor systems to be subjected to allocation. Finally, the authors use experiments in which a communications protocol process is allocated to verify the effectiveness of their program allocation method. © 2007 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(6): 48– 60, 2007; Published online in Wiley InterScience (www.interscience. wiley.com). DOI 10.1002/ecjc.20324
In this paper, the authors propose a new noise suppression technique that executes iterative processing and sets parameters that are suited for that processing in the spectral subtraction method, which is a noise reduction technique for noise-added speech. Iterative processing is a technique in which speech enhancement processing is executed by considering the estimated speech that is obtained when noise reduction processing is executed once as the input signal again so that a reduction of the residual noise is anticipated. A further reduction of residual noise for which speech degradation is controlled can be achieved by adjusting the parameters for each iteration. The authors also simultaneously propose a technique for maintaining the real-time nature of spectral subtraction when the proposed technique is executed. They use actual speech to which white noise, automobile noise, and crowd babble noise is added to compare the characteristics of the two proposed methods with the conventional spectral subtraction method and its improvements. The authors verified according to objective and subjective evaluations that each proposed technique showed superior results in all noisy environments. © 2006 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(4): 39 –51, 2007; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.20242
We used ultrasound exposure to perform disaggregation and surface modification of aggregated nano-diamond particles. We formed a standing-wave sound field using a Langevin transducer to drive a stainless-steel vibrating disk installed at the bottom of a water tank, and generated efficient acoustic cavitation. With 20 minutes of ultrasonic exposure, we were able to reduce aggregated nano-diamond particles with average particle diameters of 5µm to average particle diameters of 100 nm or less. We also confirmed that we had improved the zeta potential, an index of surface modification, by a factor of more than two. Furthermore, we used a hydrophone to measure the sound field inside the water tank, and calculated the average acoustic intensity at 800W/m2 or more to demonstrate that the nano-diamond surfaces had been modified. We confirmed that the zeta potential and average particle diameter of the nano-diamonds could be maintained for at least 150 days after ultrasound exposure. © 2007 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(12): 10– 18, 2007; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.20268
Changes are occurring in the environment Surrounding business organizations, such as the development of open networks and flexible migration of employees. In the context of such changes, it has recently been considered more important to prepare measures against information leaks from inside the organization. Among these, an urgent issue in ordinary organizations is to reinforce the existing business system so as to realize highly cost-effective security functions. This paper considers an actual case of information leakage as a model and systematically summarizes the conditions regarding security functions, based on the views of legal specialists. An actually introduced security system based on fingerprint identification is investigated, and it is shown that practical security measures can be established. A development configuration is then proposed in which part of the security system is placed in a data center and shared by multiple organizations. (C) 2007 Wiley Periodicals, Inc.
In the ubiquitous computing environment, the physical locations of persons and goods are extremely important information. In this paper, the design of a fully distributed localization system based on ultrasound, mainly for the indoor environment, is described. This system performs localization with as few positioning references as possible by an iterative technique. When such a localization method is used, deterioration of localization accuracy due to no-line-of-sight signals and to accumulated errors is a problem. In this paper, the resolution of these problems is discussed and system evaluation by implementation experiments is carried out. With 24 devices, localization with an accuracy of about 20 cm is demonstrated. © 2007 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(6): 17– 26, 2007; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.20249
We propose hybrid intra prediction, a new method for generating intra prediction images in intra coding. When generating intra prediction images in pixel domain, this method selects the intra prediction image from one generated using pixel-based intra prediction and another generated using block-based intra prediction, according to which is optimal based on a rate distortion criterion. This selection is performed for macroblock units and segmented macroblock units that we term sub-macroblocks. In addition, information indicating which intra prediction method was used within a macroblock is described within the bitstream using a tree structure. We evaluate the proposed method by implementing it in an H.264-based intra prediction coding scheme. Experimental results confirm that compared to conventional methods, our implementation of the proposed method is able to achieve a 6% reduction in bit rate while maintaining the same peak signal to noise ratio (PSNR). (C) 2006 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 90(2): 50-62, 2007;
