Information processing with only locally connected networks such as cellular neural networks is advantageous for integrated circuit implementations. Adding long range connections can often enhance considerably their performance. It is sufficient to activate these connections randomly from time to time (blinking connections). This can be realized by sending packets on a communication network underlying the information processing network that is needed anyway for bringing information in and out of the locally connected network. We prove for the case of multi-stable networks that if the long-range connections are switched on and off sufficiently fast, the behavior of the blinking network is with high probability the same as the behavior of the time-averaged network. In the averaged network the blinking connections are replaced by fixed connections with low (average) coupling strength.

In this paper, we present a learning approach, positive correlation learning (PCL), that creates a multilayer neural network with good generalization ability. A correlation function is added to the standard error function of back propagation learning, and the error function is minimized by a steepest-descent method. During training, all the unnecessary units in the hidden layer are correlated with necessary ones in a positive sense. PCL can therefore create positively correlated activities of hidden units in response to input patterns. We show that PCL can reduce the information on the input patterns and decay the weights, which lead to improved generalization ability. Here, the information is defined with respect to hidden unit activity since the hidden unit plays a crucial role in storing the information on the input patterns. That is, as previously proposed, the information is defined by the difference between the uncertainty of the hidden unit at the initial stage of learning and the uncertainty of the hidden unit at the final stage of learning. After deriving new weight update rules for the PCL, we applied this method to several standard benchmark classification problems such as breast cancer, diabetes and glass identification problems. Experimental results confirmed that the PCL produces positively correlated hidden units and reduces significantly the amount of information, resulting improved generalization ability.

This paper describes a pattern classifier for detecting frontal-view faces via learning a decision boundary. The proposed classifier consists of two major parts for improving classification accuracy: the implicit modeling of both the face and the near-face classes resulting in an extended discriminative feature set, and the subsequent composite Support Vector Machines (SVMs) for speeding up the classification. For the extended discriminative feature set, Principal Component Analysis (PCA) or Independent Component Analysis (ICA) is performed for the face and near-face classes separately. The projections and distances to the two different subspaces are complementary, which significantly enhances classification accuracy of SVM. Multiple nonlinear SVMs are trained for the local facial feature spaces considering the general multi-modal characteristic of the face space. Each component SVM has a simpler boundary than that of a single SVM for the whole face space. The most appropriate component SVM is selected by a gating mechanism based on clustering. The classification by utilizing one of the multiple SVMs guarantees good generalization performance and speeds up face detection. The proposed classifier is finally implemented to work in real-time by cascading a boosting based face detector.

An efficient hybrid image vector quantization (VQ) technique based on a classification in the DCT domain is presented in this letter. This algorithm combines two kinds of VQ, predictive VQ (PVQ) and discrete cosine transform domain VQ (DCTVQ), and adopts a simple classifier which employs only three DCT coefficients in the 88 block. For each image block, the classifier switches to the PVQ coder if the block is relatively complex, and otherwise switches to the DCTVQ coder. Experimental results show that the proposed algorithm can achieve higher PSNR values than ordinary VQ, PVQ, JPEG, and JPEG2000 at the same bit-rate.

The presence of the complex scaling behavior in network traffic makes accurate traffic prediction a challenging task. Some conventional prediction tools such as the recursive least square method are not appropriate for network traffic prediction. In this paper we propose a timescale decomposition approach to real time traffic prediction. The raw traffic data is first decomposed into multiple timescales using the à trous Haar wavelet transform. The wavelet coefficients and the scaling coefficients at each scale are predicted independently using the ARIMA model. The predicted wavelet coefficients and scaling coefficient are then combined to give the predicted traffic value. This timescale decomposition approach can better capture the correlation structure of the traffic caused by different network mechanisms, which may not be obvious when examining the raw data directly. The proposed prediction algorithm is applied to real network traffic. It is shown that the proposed algorithm outperforms traffic prediction algorithms in the literature and gives more accurate results.

This investigation proposes a fast wavelet-based color segmentation (FWCS) technique and a modified directional region-growing (DRG) technique for semantic image segmentation. The FWCS is a subsequent combination of progressive color truncation and histogram-based color extraction processes for segmenting color regions in images. By exploring specialized centroids of segmented fragments as initial growing seeds, the proposed DRG operates a directional 1-D region growing on pairs of color segmented regions based on those centroids. When the two examined regions are positively confirmed by DRG, the proposed framework subsequently computes the texture features extracted from these two regions to further check their relation using texture similarity testing (TST). If any pair of regions passes double checking with both DRG and TST, they are identified as associated regions. If two associated regions/areas are connective, they are unified to a union area enclosed by a single contour. On the contrary, the proposed framework merely acknowledges a linking relation between those associated regions/areas highlighted with any linking mark. Particularly, by the systematic integration of all proposed processes, the critical issue to decide the ending level of wavelet decomposition in various images can be efficiently solved in FWCS by a quasi-linear high-frequency analysis model newly proposed. The simulations conducted here demonstrate that the proposed segmentation framework can achieve a quasi-semantic segmentation without priori a high-level knowledge.

We present a new method to detect weak linear frequency modulated (LFM) signals in strong noise using the chaos oscillator. Chaotic systems are sensitive to specific signals yet immune to noise. With our new method we firstly use the Radon-Wigner transform to dechirp the LFM signal. Secondly, we set up a chaotic oscillator sensitive to weak signals based on the Duffing equation, and poising the system at its critical state. Finally, we input the dechirped sequence into the system as a perturbation of the driving force. A weak signal with the same frequency will lead to a qualitative transition in the system state. The weak signal in the presence of strong noise can then be detected from the phase transition of the phase plane trajectory of the chaotic system. Computer simulation results show that LFM signals with an SNR lower than -27 dB can be detected by this method.

We research on an importance sampling (IS) simulation to estimate a low error probability of turbo codes. The simulation time reduction in IS depends on another probability density function (p.d.f.) called simulation p.d.f. The previous IS simulation method can not evaluate the error probability on the low SNR and waterfall region. We derive the optimal simulation p.d.f. which gives the perfect estimator. A new simulation p.d.f. design, which is related to the optimal one, is proposed to overcome the problem of the previous IS method. The proposed IS simulation can evaluate all possible error patterns. Finally, some computer simulations show that the proposed method can evaluate the error probability on the low SNR, waterfall, and error floor regions. At the evaluation of the BER of 10-7, the simulation time of the proposed method is about 1/350 times as short as that of the Monte-Carlo simulation. When the BER is less than 710-8, the proposed method requires shorter simulation time than the conventional IS method.

This paper deals with the use of neural network rule extraction techniques based on the Genetic Programming (GP) to build intelligent and explanatory evaluation systems. Recent development in algorithms that extract rules from trained neural networks enable us to generate classification rules in spite of their intrinsically black-box nature. However, in the original decompositional method looking at the internal structure of the networks, the comprehensive methods combining the output to the inputs using parameters are complicated. Then, in our paper, we utilized the GP to automatize the rule extraction process in the trained neural networks where the statements changed into a binary classification. Even though the production (classification) rule generation based on the GP alone are applicable straightforward to the underlying problems for decision making, but in the original GP method production rules include many statements described by arithmetic expressions as well as basic logical expressions, and it makes the rule generation process very complicated. Therefore, we utilize the neural network and binary classification to obtain simple and relevant classification rules in real applications by avoiding straightforward applications of the GP procedure to the arithmetic expressions. At first, the pruning process of weight among neurons is applied to obtain simple but substantial binary expressions which are used as statements is classification rules. Then, the GP is applied to generate ultimate rules. As applications, we generate rules to prediction of bankruptcy and creditworthiness for binary classifications, and the apply the method to multi-level classification of corporate bonds (rating) by using the financial indicators.

In the future network, optical technology will play a stronger role not only for transmission but also for switching. Optical burst switching (OBS) emerged as a promising switching paradigm. It brings together the complementary strengths of optics and electronics. This paper presents the design and implementation of an overlay mode burst-switched photonic network testbed, including its architecture, protocols, algorithms and experiments. We propose a flexible "transceiver + forwarding" OBS node architecture to perform both electronic burst assembly/disassembly and optical burst forwarding. It has been designed to provide class of service (CoS), wavelength selection for local bursts, and transparency to cut-through bursts. The functional modules of OBS control plane and its key design issues are presented, including signaling, routing, and a novel scheduling mechanism with combined contention resolution in space and wavelength domains. Finally, we report the experimental results on functional verification, performance analysis and service demonstration.

This paper describes a sophisticated method to estimate visual direction using iris contours. This method requires only one monocular image taken by a camera with unknown focal length. In order to estimate the visual direction, we assume the visual directions of both eyes are parallel and iris boundaries are circles in 3D space. In this case, the two planes where the iris boundaries reside are also parallel. We estimate the normal vector of the two planes from the iris contours extracted from an input image by using an extended "two-circle" algorithm. Unlike most existing gaze estimation algorithms that require information about eye corners and heuristic knowledge about 3D structure of the eye in addition to the iris contours, our method uses two iris contours only. Another contribution of our method is the ability of estimating the focal length of the camera. It allows one to use a zoom lens to take images and the focal length can be adjusted at any time. The extensive experiments over simulated images and real images demonstrate the robustness and the effectiveness of our method.

Seasonal ARIMA model is a good traffic model capable of capturing the behavior of a network traffic stream. In this paper, we give a general expression of seasonal ARIMA models with two periodicities and provide procedures to model and to predict traffic using seasonal ARIMA models. The experiments conducted in our feasibility study showed that seasonal ARIMA models can be used to model and predict actual wireless traffic such as GSM traffic in China.

In this study, we present a way to choose route selection metric while discovering a new route in ad hoc mobile networks. We have used link expiration time and busy rate to calculate the route cost. The route cost is compared to a threshold value to decide whether the traffic of the route is high or low. If it is high then the system chooses busy rate as a route selection metric to avoid traffic congestion and if it is low the link expiration time is used to select the longlasting route. We have examined the characteristics of the routing protocol by computer simulation and found that it over performs the conventional protocols.

This paper proposes a novel PN (Pseudo Noise) synchronization system using Cycle-and-Add property of M-sequence featuring fast acquisition in DS-CDMA (direct sequence-code division multiple access). Fast acquisition is carried out by generating a PN sequence in a simple multiplicative action of a received signal with its delayed one. This multiplicative action is similar to differentially coherent detection and realizes an anti-fading property. Easy implementation is materialized by the fact that the system is mostly composed of baseband devices. The principle, performance evaluation and the detection probability of synchronization for the proposed method are shown. Furthermore, detection probability of synchronization in a fast Rayleigh fading channel is shown for the performance evaluation.

Interactive multimedia applications (IMAs) require not only adequate bandwidth to support large volume data transmission but also bounded end-to-end transmission delay between end users. This study proposes a Delay and Degree constrained Multicasting Spanning Tree (D2MST) algorithm to build an any-to-any share tree for IMAs. D2MST comprises root selection and spanning tree generation. A weighting function is defined based on the novel concept of network center and gravity to choose the root of a share tree. From the root, a spanning tree is built by incrementally connecting nodes with larger "power" to the tree so that the degree constraint is satisfied. Simulation results show that D2MST can successfully generate a Δ-constraint MST in which a high percentage of nodes can interact within the bounded delay.

In reconstructing 3-D from images based on feature points, one usually defines a triangular mesh that has these feature points as vertices and displays the scene as a polyhedron. If the scene itself is a polyhedron, however, some of the displayed edges may be inconsistent with the true shape. This paper presents a new technique for automatically eliminating such inconsistencies by using a special template. We also present a technique for removing spurious occluding edges. All the procedures do not require any thresholds to be adjusted. Using real images, we demonstrate that our method has high capability to correct inconsistencies.

The mobile agent paradigm is a promising technology to structure distributed applications. Since mobile agents physically move to a remote host that is under the control of a different principal, they need to be protected from this environment which is responsible for execution. In this paper, we provide a new cryptographic methodology of protecting mobile agents from unauthorized modification for the program code by malicious hosts.

In this paper, the electrical constants of a biological tissue-equivalent agar-based solid phantom from 3.0 to 6.0 GHz are described. The developed phantom can reproduce the electrical constants of biological tissues from 3.0 to 6.0 GHz, and it is not necessary to change the phantom for each frequency band in the range of 3.0 to 6.0 GHz during the measurements. Moreover, adjustments to the dielectric constants of the phantom at 3.0, 3.8, 5.2, and 5.8 GHz are presented. The constants of this phantom can be adjusted mainly by using polyethylene powder and sodium chloride. The phantom can be used to evaluate the Specific Absorption Rate (SAR) as well as the antenna characteristics in the range of 3.0 to 6.0 GHz. Furthermore, the effect of the electrical constants of the phantom on the SAR is investigated. The investigation of SAR measurements is performed on the phantom at 5.2 GHz using the thermographic method. Calculations using the FD-TD method and the finite difference method based on the heat conduction equation are carried out in order to evaluate the thermal diffusion in the measurements using the thermographic method. The measured and calculated results are in good agreement. There is evidence that the thermal diffusion influences the SAR estimation at 5.2 GHz more than in a lower frequency range even though this method basically does not depend on the frequency.

Deflection routing is one of the promising approaches to resolve contention in the optical burst switching networks. In the conventional deflection routing scheme, optical bursts may be unable to traverse the route evaluated to select an outgoing link because of the contention at succeeding downstream transit nodes. As a result, the optical bursts may traverse a different route resulting in a long distance or decreased performance. This paper proposes a deflection routing scheme that considers the possibility of the contention at downstream nodes. This scheme utilizes the "expected route distance" instead of the static route distance toward a destination node. The expected route distance considers the possibility of contention at each downstream transit node and is calculated using measured link blocking probabilities at each downstream transit node. The selection priority of each outgoing link is given dynamically based on its expected route distance toward a destination node. By considering the possibility of contention at downstream nodes, a routing scheme with high performance can be realized. The loss rate of optical bursts is improved when an imbalanced load is applied to the network, and the loss rate of optical bursts is also improved when the network includes links with extremely different distances.

This letter proposes two very-low-complexity maximum-likelihood (ML) detection algorithms based on QR decomposition for the quasi-orthogonal space-time code (QSTBC) with four transmit antennas [3]-[5], called VLCMLDec1 and VLCMLDec2 decoders. The first decoder, VLCMLDec1, can be used to detect transmitted symbols being extracted from finite-size constellations such as phase-shift keying (PSK) or quadrature amplitude modulation (QAM). The second decoder, VLCMLDec2, is an enhanced version of the VLCMLDec1, developed mainly for QAM constellations. Simulation results show that both of the proposed decoders enable the QSTBC to achieve ML performance with significant reduction in computational load.