We show how "self-awareness", through on-lineself-monitoring and measurement, coupled with intelligent adaptivebehaviour in response to observed data, can be used to offer quality of service to networkusers based on the "Cognitive Packet Network" (CPN) design.

This work deals with how the machine classifies human facial expressions, which has been a challenging problem for many researchers from the diverse areas. The facial expression recognition system mainly consists of two cascade stages: the representation method for the facial images at the front and the facial emotion classifier at the back. The Gabor-wavelets based method has shown promising performance because of its efficient representation and biological implication. Here we focus on the classification method to obtain high recognition rate of facial expressions. Results suggest that enhanced Fisher discrimination model, which had been used for the face recognition task, outperformed Principal Component Analysis (PCA) based classifier (or the neural network) with the 93% correction rate, when it is combined with the Gabor representation.

This paper presents a practical method for detecting the point in the monitor where a user gazes by moving his face and eyes. Previous gaze detection system uses a wide view camera, which can capture the whole face of user. In such case, the image resolution is so low that the fine movements of user's eye cannot be exactly detected and the accurate eye gaze position cannot be located consequently. So, we implement the gaze detection system with a wide view camera and a narrow view camera. Because the narrow view camera captures the eye image with high magnification, the eye position easily escapes from the narrow view camera by user's facial movements. For these reasons, we adopt the functionalities of auto focusing and auto panning/tilting into the narrow view camera and those are performed based on the information of the detected 3D facial feature positions by the wide view camera. As experimental results, our gaze detection system operates in real-time and the gaze detection accuracy between the computed positions and the real ones is about 3.57 cm of RMS error.

A license plate segmentation system is designed and developed. The system has preprocessing, approximate region finding, pure plate segmentation and character segmentation modules. For each module, alternative available methods are examined and proper sequence of operations is developed. In character segmentation module a novel method is devised. Finally, overall performance of the system is reported.

This paper presents a novel algorithm for license-plate detection and localization problems by using Gabor transform in detection and local vector quantization in segmentation. As of our knowledge this is the first application of Gabor filters to license plate segmentation problem. Even though much of the research efforts are devoted to the edge or global thresholding-based approaches, it is more practical and efficient to analyze the image in certain directions and scales utilizing Gabor transform instead of error-prone edge detection or thresholding. Gabor filter response only gives a rough estimate of the plate boundary. Then binary split tree is used for vector quantization in order to extract the exact boundary and segment the plate region into disjoint characters which become ready for the optical character recognition

Increasing popularity of wireless services has triggered the need for efficient wireless transport mechanisms. TCP, being the reliable transport level protocol widely used in wired network world, was not designed with heterogeneity in mind. The problem with the adaptation of TCP to the evolving wireless settings is because of the assumption that packet loss and unusual delays are mainly caused by congestion. TCP originally assumes that packet loss is very small. On the other hand, wireless links often suffer from high bit error rates and broken connectivity due to handoffs. A range of schemes, namely end-to-end, split-connection and link-layer protocols, has been proposed to improve the performance of transport mechanisms, in particular TCP, on wireless settings. In this study, we examine these mechanisms for wireless transport, and discuss our comparative simulation results of end-to-end TCP versions (Tahoe, Reno, NewReno and SACK) in various network settings including wireless LANs and wired-cum-wireless scenarios.

Abstract—The key problem that is expected in the future mobile wireless networks is related with the mobility effect on the network data traffic, which may cause the performance degradation and congestion. This paper explores the mobility effect on the delay and congestion when the future mobile wireless networks have become the whole IP based networks and introduces a network level congestion control based on the Explicit Congestion Notification (ECN) broadcasting in the wireless sub networks. The simulation results are provided to show the performance improvement when the proposed scheme’s adopted.

Abstract. In this paper, a new static traffic-based location area design scheme named ETB-LAD (enhanced TB-LAD) for wireless systems is proposed. This scheme is an enhancement on the previously published TB-LAD scheme. In the new method, the expected intercell movement patterns of mobiles are determined and then the cells are partitioned into location areas (LAs) by applying the new traffic-based cell grouping algorithm, which has two goals. First, the cell pairs with higher intercell mobile traffic are grouped into the same LA. Second, the LAs, in which the neighbour cells have higher intercell traffic, allowed to include more cells than the LAs, where the intercell traffic is low. The aim of the scheme is to decrement the inter-LA movements, which create registration traffic. Experimental results show that the new scheme performs better than the TB-LAD and proximity based location area design (PB-LAD) schemes in the sense that it can reduce the location updates.

In this paper, we consider the call admission problem in next generation wireless networks, which must handles multi-media traffics. We give an algorithm, which finds the optimal number of guard channels, which minimizes the overall blocking probability calls with lowest level of QoS in a multi-cell cellular network subject to the hard constraint on the blocking probabilities of other calls.

Dynamic hashing, while surpassing other access methods for uniformly distributed data, usually performs badly for non-uniformly distributed data. We propose a robust scheme for multi-level extendible hashing, allowing efficient processing of skewed data as well as uniformly distributed data. In order to test our access method, we implemented it and compared it to several existing hashing schemes. The results of the experimental evaluation demonstrate the superiority of our approach in both index size and performance.

Term weighting is one of the most important aspects of modern Web retrieval systems. The weight associated with a given term in a document shows the importance of the term for the document, i.e. its usefulness for distinguishing documents in a document collection. In search engines operating in a dynamic environment such as the Internet, where many documents are deleted from and added to the database, the usual formula involving the inverse document frequency is too costly to be computed each time the document collection is updated. This paper proposes two new simple and effective weighting functions. These weighting functions have been tested and compared with results obtained for the PIVOT, SMART and INQUERY methods using the WT10g collection of documents.

Flexible data are a common concept in today's information society. Some data can be unknown, other data may be inaccurate or uncertain. Still, this flexible data must be accounted for in modern businesses and therefore must be stored. Flexible relational databases have been studied extensively over time, which resulted in numerous models and representation techniques, some of which have been implemented as software layers on top of database systems. Different query languages and end-user interfaces have been extended to perform flexible queries on both regular and flexible databases. In this paper, a framework is presented that not only enables flexible querying on the relational model, but on other database models as well, of which the most important are object-oriented database models. This framework, called FQF or Flexible Querying Framework, is built on the recently developed Java Data Objects (JDO) standard.

In this paper, we introduce a new lexicon free, probabilistic stemmer used in a developing Turkish Information Retrieval system. It has a linear computational complexity and its test success ratio is 95.8%. The main contribution of this paper is to give a thorough description of a probabilistic perspective for stemming which can also be generalized to apply to other agglutinative languages like Finnish, Hungarian, Estonian and Czech

InfiniBand technology is being accepted as the future system interconnect to serve as the high-end enterprise fabric for cluster computing. This paper presents the design and implementation of an InfiniBand host channel adapter (HCA) based on dual ARM9 processor cores. The HCA is an SoC called KinCA which connects a host node onto the InfiniBand network both in hardware and in software. Since the ARM9 processor core does not provide necessary features for multiprocessor configuration, novel inter-processor communication and interrupt mechanisms between the two processors were designed and embedded within the KinCA chip. KinCA was fabricated as a 564-pin enhanced BGA (Ball Grid Array) device using 0.18mm CMOS technology. Mounted on host nodes, it provides 10 Gbps outbound and inbound channels for transmit and receive, respectively, enabling a high-performance cluster system.

Multiplication of long integers is a cornerstone primitive in most cryptosystems. Multiplication for big numbers can be performed best using Karatsuba-Ofman's divide-and-conquer approach. Multiplying long integers using Karatsuba-Ofman's algorithm is fast but the algorithm is highly recursive. We propose a less recursive and efficient hardware architecture for this algorithm. We compare the proposed multiplier to other existing ones.

Modular multiplication is fundamental to several public-key cryptography systems such as the RSA encryption system. It is also the most dominant part of the computation performed in such systems. The operation is time consuming for large operands. This paper examines the characteristics of yet another architecture to implement modular multiplication. An experimental modular multiplier prototype is described in VHDL and simulated. The simulation results are presented.

In this paper, we have proposed the merging of the OpenCable reference model with the smart card interface defined by ISO/IEC 7816-3 and have implemented a new Conditional Access System (CAS) using this interface along with a Personal Computer Memory Card International Association (PCMCIA) interface. In addition, we have also designed the hardware architecture of the Point-of-Deployment (POD) security module, designed with the Verilog HDL, using the above two interfaces. The designed POD security module has an embedded 32-bit Reduced Instruction Set Computer (RISC) microprocessor that manages applications, such as an MPEG-2 filter, descrambler, and Data Encryption Standard (DES). We have tested this modeled and designed system using a prototype and show satisfactory simulation results.

The generation of motion and motion capturing of an articulated body for computer animation is an expensive and time consuming task. Conventionally, animators manually generate intermediate frames between keyframes, but this task is very time consuming. This paper presents a new model-based singularity-free automatic-initialization approach to motion capturing of human motion video based on widely-available, static-background monocular video sequences. A 3D human body model is built and projected on a 2D projection plane to find best fitting with the foreground image silhouette. We convert the human motion capture problem into two types of parameter optimization problem: static optimization and dynamic optimization. First, we determine each model body configuration using static optimizations for every input image. Then, to get better motion, the results from all static optimizations are fed into a dynamic optimization process, where the entire sequence of motion is considered for the user-specified motion. The user-specified motion is defined by each user and the final form of the motion they want. A cost function for static optimization is used to estimate the degree of overlapping between the foreground input image silhouette and a projected 3D model body silhouette. The overlapping is computed using computational geometry by converting a set of pixels from the image domain to a polygon in the real projection plane domain. A cost function for dynamic optimization is the user-specified motion based on the static optimization results as well as image fitting. Our method is used to capture various human motions.

In this paper, real-time people detection and tracking using background modeling and maintenance technique for visual surveillanmce system in an indoor and an outdoor environments is described. The system operates on monocular grayscale video imagery from an static CCD camera. In order to detect foreground objects, firstly, background scene model is statistically learned using the redundancy of the pixel intensity values during learning stage, even the background is not completely stationary. This redundany information of the each pixel is separately stored in an history map shows how the pixel intensity values changes till now. Then the biggest ratio of the redundancy of the pixel intensity values in the history map during training sequence is determined to have initial background model of the scene. A background maintenance model is also proposed for preventing some kind of falses, such as, illumination changes (the sun being blocked by clouds causing changes in brightness), or physical changes (person detection while he is getting out or passing in front of the parked car). And then for people detection, candidate foreground regions are detected using thresholding, noise cleaning and their boundaries extracted using morphological filters. From these, a body posture is estimated depending on skeleton of the regions. Finaly, the trajectory of the people in motion is implemented for analysing the people actions tracked in the video sequences. Experimental results demonstrate robustness and real-time performance of the algorithm.

This paper presents a method for tracking multiple objects using multiple cameras that integrates spatial position, shape and color information to track object blobs. Given three known points on the ground, camera calibration are computed by solving a set of quaternion-based nonlinear functions rather than solving approximated linear functions. By using quaternion-based method, we can avoid singularity problem. Our method focuses on establishing correspondence between objects and templates as the objects come into view. We fuse the data from individual cameras using an Extended Kalman Filter (EKF) to resolve object occlusion. Our results show that integrating simple features makes the tracking effective, and that EKF improves the tracking accuracy when long term or temporary occlusion occurs. Experimental results are provided to demonstrate the effectiveness of the proposed method.

Image sequence stabilization aims to remove unwanted fluctuations from an image sequence. This paper proposes a novel stabilization system making use of a fuzzy filter constructed by dynamic system based estimator definition with a fuzzy corrector. The stabilization system selectively switches between a pre-defined set of membership functions so as to improve gross movement tracking and stabilization performance. The stabilization results of the proposed system are compared with results of the Super SteadyShot stabilization system incorporated into SONY® camcorders.

Recently, many micro-mobility-supporting protocols have been suggested, such as the regional registration protocol and Cellular IP. In these protocols, mobility is mainly managed by the mobile node itself when it moves within an access node. However, these protocols often have a long registration delay based on variations in the advertisement message broadcasting period and location management message transmission period. Furthermore, various problems, including radio resource wastage and an increased network load, can also occur with these protocols. Accordingly, the current paper proposes the ANMM (Access Network Mobility Management) protocol as an improved and efficient protocol for supporting micro-mobility in an access network. To realize the ANMM protocol, an effective adaptation method is proposed for link-layer handoff information along with a tunneling scheme between an access node and the gateway in an access network. As such, the ANMM protocol is able to minimize the registration delay, signaling overhead, and radio resource wastage.

Even though an agent system contributes largely to mobile computing on distributed network environment, it has a number of significant security problems. In this paper, we analyze security attacks to this system presented by NIST[3]. In order to protect it from them, we suggest a security protocol for a mobile agent system by employing Identity-based key distribution and digital multi-signature scheme. To solve the problems described on NIST, securities of mobile agent and agent platform should be accomplished. Comparing with other protocols, our protocol performs both of these securities, while other protocols mention only one of them. Also, it is designed to guarantee the liveness of agent, and to detect message modification immediately by verifying the execution of agent correctly.

We describe a secure group communication protocol for a fault-tolerant synchronous ring. Our protocol, named SSRP (Secure Synchronous Ring Protocol), integrates a secure group communication facility into an existing scalable, fault-tolerant ring protocol. SSRP is a hierarchical group communication protocol that employs Cliques GDH [4] contributory key man-agement protocol and Diffie-Hellman encryption algorithm. Security related functions and group communication functions are integrated in a single layer guaranteeing that not only the application messages of the group, but also the messages related to the group communication layer are protected. A novel fault-tolerant leader selection algorithm and a proof of Virtual Synchrony semantics on SSRP are also given.

Nowadays, threats of information security become a big issue in internet environments. Various security solutions are used as such problems' countermeasure; IDS, Firewall, ESM and VPN. But, TCP/IP protocol based Internet basically have much vulnerability of protocol itself. Specially, ICMP traffic is ubiquitous to almost every TCP/IP based network. As such, many firewalls and networks consider ICMP traffic to be benign and will allow it to pass through, unmolested. But, attackers can generate arbitrary information tunneling in the data portion of ICMP packets. This channel exists because network devices do not filter the contents of ICMP traffic. So, attackers can tunnel(covert channel) any information they want. To detect a ICMP covert channel, our approach use SVM which has excellent performance in pattern classification problems. Our experiments showed that the proposed method could detect a ICMP covert channel from normal ICMP traffic using Support Vector Machine.

The canonical operators of genetic algorithms, i.e., mutation and crossover, have nondeterministic effects on the population.They use information from only one or two fit individuals and risk deforming the chromosomes of fit individuals and cause an interruption in the progression. Estimation of Distribution Algorithms (EDAs) use probabilistic models rather than mutation and crossover, to guide the progression of genetic algorithms by placing a density over all fit individuals and sampling from this density. EDA therefore makes better use of the fitness information of the previous generation and promise faster convergence without losing any schemata. We consider parametric, nonparametric, and semiparametric models for density estimation in the EDA template with continuous genes. We compare these methods with standard backpropagation and GA proper in the problem of training a multilayer perceptron which is a complex nonlinear estimator. Our results indicate that our algorithms perform definitely better than the proper genetic algorithm (GA) on every problem and can find better solutions than those of backpropagation in training a MLP.

Not so many benchmark problems have been proposed in the area of Genetic Programming (GP). In this study, a new artificial benchmark problem is designed for GP. The different parameters that can be used to tune the difficulty of the problem are analyzed. Also, the initial experimental results obtained on different instances of the problem are presented.

We introduce the modular genetic algorithm (MGA). The modular genetic algorithm is a search algorithm designed for a class of problems pervasive throughout nature and engineering: problems with modularity and regularity in their solutions. We hypothesize that genetic search algorithms with explicit mechanisms to exploit regularity and modularity on the problem space would not only outperform conventional genetic search, but also scale better for this problem class. In this paper we present experimental evidence in support of our hypothesis. In our experiments, we compare a limited version of the modular genetic algorithm with a canonical genetic algorithm (GA) applied to the checkerboard-pattern discovery problem for search spaces of sizes 2^{32}, 2^{128}, and 2^{512}. We observe that the MGA significantly outperforms the GA for high complexities. More importantly, while the performance of the GA drops 22.50% when the complexity of the problem increases, the MGA performance drops only 11.38%. These results indicate that the MGA has a strong scalability property for problems with regularity and modularity in their solutions.

In this study, compared to the existing one, a more realistic evaluation method for discourse segmentation is introduced. It is believed that discourse segmentation is a fuzzy task [Pas 96]. Human subjects may agree on different discourse boundaries, with high agreement among them. In the existing method a threshold value is calculated and sentences that marked by that many subjects are decided as real boundaries and other marks are not been considered. Furthermore automatically discovered boundaries, in case of being misplaced, are treated as a strict failure, disregarding the proximity wrt to the human found boundaries. The proposed method overcomes these shortcomings, and credits the fuzziness of the human subjects' decisions as well as tolerates misplacements of the automated discovery. The proposed method is tunable from crisp/harsh to fuzzy/tolerant on human decision as well as automated discovery handling.

Architectural description languages (ADLs) are used to specify a high-level, compositional view of a software application, specifying how a system is to be composed from coarse-grain components. ADLs usually come equipped with a formal dynamic semantics, facilitating specification and analysis of distributed and event-based systems. In this paper, we describe TrustME, an ADL for modelling architectures systems based built from web services. We describe how its syntax and semantics meets concerns specific to Business-to-business (B2B). The ADL provides a framework that provides both a process and a structural view of web service-based systems. We use Petri-net descriptions to give a dynamic view of business workflow for web service collaboration, and the ADL to provide a static, compositional, organisational view of web service deployment. We adapt the approach Schmidt to define a form of design-by-contract (see Meyer) for configuring workflow architectures. This serves as a configuration-level means of constructing safer, more robust systems.

The complexity of Web applications is increasing and modeling is used to manage such complexity. In this work, we propose modeling the architecture of web applications using the Specification and Description Language (SDL) that is a formal description language. SDL is shown to be quite convenient for modeling features of web application such as hyperlinks, sending and receiving data and client-server communication. Hence, it facilitates our understanding of web applications. Further, SDL modeling allows us to extend existing implementation and verification techniques to web applications.

Web applications have unique characteristics that require suitable software engineering practices in the development process. In this way, software architectures and pattern-based approaches are suitable design techniques for modeling purposes. But if we want to build sets of similar systems, we need to represent the common and variable aspects of such systems under an architectural point of view. Therefore, representing and managing those variable issues is a goal to achieve when designing similar software applications. In this work we will try to deal with the variability problem from a pattern point of view as well as applying this to web software products.

Given the complexity of web-applications, their development process should focus on improving reusability and flexibility of the application. An approach to meet these concerns is proposed in this paper by employing enterprise frameworks. Enterprise Frameworks encapsulate reusable, tailorable software solutions as a collection of collaborative components, assuring integration with new or existing components, reducing in this way the complexity of the systems. The paper presents a development process of enterprise frameworks for web-applications targeting to improve their flexibility and reusability. A compositional design pattern targeting to increase framework flexibility is also proposed.

This paper presents a postprocessing technique based on the theory of projections onto convex sets (POCS) to reduce the blocking artifacts in low bit rate BDCT-coded images. In the block discrete cosine transform (BDCT), the image is divided into a grid of non-overlapped 8 X 8 blocks, and then each block is coded separately. Thus, a block, which is shifted one pixel diagonally from the original grid of BDCT, will include the boundary of the original 8 X 8 block. If blocking artifacts are introduced along the block boundary, the frequency characteristics of this shifted block will be different from those of the original block. Therefore, a comparison of frequency characteristics of these two overlapping blocks can detect the undesired high-frequency components, mainly caused by the blocking artifacts. By eliminating these undesired high-frequency components adaptively considering the local statistical properties of the image, robust smoothing projection operator can be obtained. Computer simulation results indicate that the proposed scheme provides better visual performance than conventional postprocessing algorithms.

To convert an interlaced video into a progressive one effectively, de-interlacing techniques have been developed. However, existing de-interlacing techniques can not perfectly remove high frequency alias components inherently existing in the interlaced video. In this paper, based on the theory of POCS, we propose a video enhancement technique to improve the visual quality of de-interlaced video. We introduce novel constraint sets that are suitable for the characteristics of de-interlaced video and derive the projection operators for those sets. These projection operators using multiframes can reduce high frequency alias components of the de-interlaced video effectively. Extensive computer simulations show that the proposed method significantly improve the quality of the de-interlaced video by restoring the meaningful frequency information of the original progressive video.

MPEG video traffic is expected to represent most of the load in the future high-speed networks. Adequate traffic models for MPEG Variable Bit-Rate (VBR) video are important for network design, performance evaluation, admission control and resource allocation. It is well-known that VBR video exhibits long-range correlations over arbitrarily long time-scales, a phenomenon usually referred to as Long-Range Dependence(LRD). Many models for VBR video traffic have been proposed in the literature. However, while the correlation between Groups of Pictures (GOPs) has been widely analyzed in literature, little effort has been devoted up to now to the generation of the different frame-types within a MPEG GOP. Due to the difficulty and complexity involved in this issue, this is often neglected even though it is a fundamental characteristic of MPEG traffic and it might have an important impact on queueing performance. In fact, many works avoid to descend to the frame level proposing models for the GOP process. In this paper, we analyze the impact on queueing performance of different solutions proposed in the literature for generating the different types of MPEG frames within a GOP (these solutions have never been compared and analyzed jointly). In this way, we can be in a good position to use the best model for MPEG in terms of simplicity, computational efficiency and queueing performance, depending on the performance metric to study (loss rate, mean delay and jitter).

In this paper, we introduce the composition problem associated with object-based video coding and propose a solution to this problem. Although the object-based rate control algorithm can provide the overall coding gain, it may create the composition problem due to different temporal resolutions in object encoding. By checking shape changes of video objects, we can minimize the possibility of the composition problem at the encoder. At the decoder, we can also apply hole detection and recovery algorithms to eliminate the effect of the composition problem to the human visual system.

In this paper we make a formal analysis of one-time password protocols using BAN logic and provide some guidelines to integrate securely one-time passwords with smartcard based authentication. We also propose some extensions to the BAN logic to facilitate analyzing authentication protocols using hash chains.

In this paper we present a complete policy-based framework, composed of a policy specification language and its deployment and management models for policy distribution. The Police policy language provides a common means of specifying every kind of policies that may be involved in management of distributed systems. It supports obligations and permissions. The Police policies contain event triggered action sets for policy-based management of distributed systems. Other key concepts of the framework include domains used to group objects, and distributed conflict handling.

Published work on computer network dependability frequently uses availability as a performance measure. However, although several ways of defining availability have been proposed, none capture the overall level of service obtained by the client hosts in a modern n-tier client/server architecture. We propose such a measure by calculating the fraction of client hosts receiving complete services from the network. We also extend a published, efficient heuristic method for calculating availability to take into account our new proposed measure. The end result is a procedure of polynomial complexity O(nt4) where nt is the total number of components (hosts, links and inteconnection equipment) in the network. Numerical results of applying the method to several networks are given.

Multicast is an evolving technology for efficient transmission for one-to-many and many-to-many communications. The successful deployment of secure multicast model needs to be more distributed rather than the current centralized approach. In this work, a distributed approach in multicast routing is proposed which enhances the group key management schemes over a large-scale network. Existing work prove that group key management is very much vulnerable due to latency and complexity involved in multicast transmission. This model is designed to reduce latency and distribute the complexity. The assumptions based on which this work was implemented are discussed.

Multiagent simulation is gaining popularity due to its intuitiveness and ability in coping with domain complexity. HLA, being a distributed simulation architecture standard, is a good candidate for implementing a multiagent simulation infrastructure on, provided that agent communication can be implemented. In this paper, we present a layered approach to the agent communication, followed by a discussion of implementing the lowest two layers of this approach using HLA and the agent communication language KQML in the context of a multiagent simulation.

This paper presents a systems view of the Autonomous avigation and Sensing Experimental Research (ANSER) project. ANSER is one of the most advanced UAV projects of its kind, which is aimed at demonstrating Decentralised Data Fusion (DDF) and Simultaneous Localisation and Map (SLAM) building on multiple cooperative UAVs. The demonstration of DDF and SLAM require both navigation and terrain sensors to be carried onboard by the UAVs. These include an INS/GPS navigation system, millimeter wave (MMW) radar, and both a single vision node and a vision node augmented with a laser system. This paper presents the implementation of the algorithms, sensors and platforms, along with their interaction to address this demonstration. Furthemore, the paper will present the high fidelity real time simulator which is implemented to thoroughly test all algorithms and sensors before actual implementation in the environment.

This work proposes a novel approach for introducing market-driven strategy to robot soccer domain in order to solve vital issues related to multi-agent coordination. In robot soccer two groups of robots namely “teams”, compete with each other to win the match. For the benefit of the team, robots should work collaboratively, whenever possible. Market-driven approach applies the basic properties of free market economy to a team of robots, to increase the profit of team as much as possible. This approach lies on the assumption that maximizing individual profit will approximate the global profit maximization. In several works, this method was applied to some open issues in multi-agent systems like multi-robot exploration and coordination, but these implementations were limited. In this work, for the first time this approach is applied to robot soccer domain, which is one of the prominent topics of multi-agent researches, as being a complex, dynamic and real-time event. In this paper, a market-driven collaborative task allocation algorithm for robot soccer domain which differs from many other multi-agent problems with its highly dynamic and complex behavior is proposed and experimental results obtained are discussed.

In this study, the application of the single input single output (SISO) neural generalized predictive control (NGPC) and SISO generalized predictive control (GPC) to a three joint robotic manipulator are presented. The sinusoidal and cubic trajectory principles were used for position reference and velocity reference trajectories. NGPC-SISO algorithm performs better results than GPC-SISO algorithm for both trajectories. The GPC-SISO robotic manipulator control results have better values in the case of the sinusoidal trajectory, but the NGPC-SISO robotic manipulator control results for both the cubic and sinusoidal trajectory are almost similar.

ABSTRACT The primary features of the Object-Oriented paradigm lead to develop a complex and compositional testing framework for object-oriented software. The increased size and complexity of software systems has led to the current focus on developing distributed applications that are constructed from component-based systems. A component-based system is composed of both data and functionality and is configurable through parameters at run-time. For an object-oriented application, class definitions provide the basic components of a program. Low-level component testing is nothing but the unit testing, which is to be performed on the independent units without any interface and interactions. This is referred as Autonomous Unit Level (AUL) testing and it is true for conventional system testing but not for object oriented systems, since an object (class) is to be considered as an unit. Testing a class for its functionality means testing its member. The general members of a class are data members, constructors and methods. In the object-oriented environment, generally the interaction between the methods can be classified as follows. Case-I Interaction between the methods of same class. Interaction between the methods of same class and program functions. Case-II Interaction between the methods of different classes. Interaction between the methods of different classes and program functions. In case-II, interaction between different class members can be referred as unit?to?unit interaction. Normally the unit-to-unit interaction is to be examined in the integration-testing phase. But in case-I, the program functions can’t be considered as separate units. So, the interaction between the members of a class and program functions must be examined during the unit (object) testing. This is referred as Inter-Procedure Level testing (IPL) and it is not a part of the integration testing. Considerable research has gone into the development of methods for evaluating how ‘good’ a test set is for a given program against a given specification. Such criteria are also known as ‘Test Adequacy Criteria’ or simply ‘Adequacy Criteria’. A testing technique may use one or more of these criteria to assess the adequacy of a test set for a given unit and then decide whether to stop testing this unit or to enhance the test set by constructing additional test cases needed to satisfy a criterion. In principle, such criteria and the procedures may also be applied for the assessment of tests for subsystems. Agent technologies facilitate software testing by virtue of their high-level abstraction for interactions. In this paper, we propose a Multi-Agent based approach to enhance the definition for class testing in object-oriented paradigm. The Integrated framework has been built on two testing techniques namely Mutation Testing and Capability Testing. In both the cases, testing is carried out at AUL and IPL. This model provides a variety of adequacy criteria by developing separate agents and defining their responsibilities in an efficient way. This development is motivated by the need to assess test sets for subsystems, in which come about due to the complications during the unit testing. This proposed intra?class testing results in higher degree of control over integration testing costs on object-oriented software.

One of the important tasks during software testing is the generation of appropriate test cases. Various approaches have been proposed to automate this task. The approaches available, however, often have problems limiting their use. A problem of dynamic test case generation approaches, for instance, is that a large number of iterations can be necessary to obtain test cases. This article proposes a novel algorithm called binary search-based test case generation (BINTEST) algorithm. Binary search conducts searching tasks in logarithmic time, as long as its assumptions are fulfilled. This article shows that these assumptions can also be fulfilled in the case of path-oriented test case generation and presents an algorithm which can be used to generate test cases covering certain paths in control flow graphs in methods.

Software quality improvement process remains incomplete if it is not initiated and conducted through a wide improvement program that considers process quality improvement, product quality improvement, and evolution of human resources. But, small software organizations are not capable of bearing the cost of establishing software process improvement programs. In this work, we propose a new software quality improvement model for small organizations, SQIMSO, based on three major issues. The first issue is that every improvement program should be wide enough to include the three main trends. The second issue is that any process quality model should answer the question 'How to do' things. The third issue is that any suggested quality model should be practical enough to be implemented by small software organizations for saving cost and time without decreasing the quality level. SQIMSO also draws upon international quality standards, models, experiences, and on a local field survey.

As the Internet usage increases rapidly, an earnest desire for flexible software applications grows larger for reflecting diverse users¡¯ demands, for handling evolution of software applications, and for treating pressure of short-term development cycle. In this trend of flexible software, service-oriented software concept is introduced, in which a system is considered as a set of flexible or distributed services. In service-oriented software approach, it is necessary to provide an analysis environment for abnormal service interactions as well as to provide a service execution framework. In this paper, we provide an approach for modeling and analyzing service interactions by using Petri nets. Each service is described as an independent module by using a new modular Petri nets and interactions among services are also described and analyzed. As a case study, feature interactions in the telecommunication literature are discussed as an example of service-oriented software approach.

This paper provides a framework for information assurance within collaborative design, based on a technique we call role-based viewing. Such role-based viewing is achieved through integration of multi-resolution geometry and security models. 3D models are geometrically partitioned, and the partitioning is used to create multi-resolution mesh hierarchies. Extracting an appropriately simplified model suitable for access rights for individual designers within a collaborative design environment is driven by an elaborate access control mechanism.

An image-space-parallel, ray-casting-based direct volume rendering algorithm is developed for rendering of unstructured data grids on distributed-memory parallel architectures. For efficiency in screen workload calculations, a graph-partitioning-based tetrahedral cell clustering technique is used. The main contribution of the work is at the proposed model, which formulates screen partitioning problem as a hypergraph partitioning problem. It is experimentally verified on a PC cluster that, compared to the previously suggested jagged partitioning approach, the proposed approach results in both better load balancing in local rendering and less communication overhead in data migration phases.

An example based rendering (EBR) method based on generalization and localization that uses artificial neural networks (ANN) and k-Nearest Neighbor (k-NN) is proposed. The method involves learning phase and application phase, which means that once a transformation filter is learned, it can be applied to any other image. In learning phase, error back-propagation learning algorithm is used to learn general transformation filter using unfiltered source image and filtered output image. ANNs are usually unable to learn filtergenerated textures and brush strokes hence these localized features are stored in a feature instance table for using with k-NN during application phase. In application phase, for any given grayscale image, first ANN is applied then k-NN search is used to retrieve local features from feature instances considering texture continuity to produce desired image. Proposed method is applied up to 40 image filters that are collection of computer-generated and human-generated effects/styles. Good results are obtained when image is composed of localized texture/style features that are only dependent to intensity values of pixel itself and its neighbors.

Feature selection is very important for many computer vision applications. However, it is hard to find a good measure for the comparison. In this study, feature sets are compared using the translation model of object recognition which is motivated by the availablity of large annotated data sets. Image regions are linked to words using a model which is inspired from the machine translation. The word prediction performance is used to evaluate large number of images.

A public key infrastructure (PKI) is being deployed in a field of network security. PKCS#5 is one of the most popular standards in PKI framework, intended for the practical implementation of password-based cryptography. So, the PKCS#5 encryption must be useful for general software applications within multimedia systems. However, it has a critical weak point in terms of security such as being vulnerable to off-line attacks due to the password-derived encryption key. In this paper, we provide a practical and simple method to improve security of the PKCS#5 encryption without modifying the installed base. The idea is to hide a salt by exploiting several existing schemes.

In this paper, a new watermark algorithm that based on wavelet packet transform is proposed for watermark of the remote sensing images, which include many high frequency components. The proposed watermark algorithm is used in the high frequency component of image and applies watermark to the overall subband that include the lowest frequency band. And the watermark is embedded on original image after selecting the significant wavelet packet coefficient. For selecting significant coefficients that apply watermark, zerotree is applied to packet transform using CPSO (Coefficient Partitioning Scanning Order) for imperceptibly and robustness. From the experimental result, the proposed algorithm shows better invisibility and robustness performance compare with conventional watermark methods. Especially, it demonstrates better robustness for high image compression in the remote sensing images application.

In this paper, we propose a new video watermarking algorithm based on the human visual system (HVS) properties to find effective locations in video sequences for robust and imperceptible watermarks. In particular, we define a new HVS-optimized global masking map for hiding watermark signals by combining frequency, spatial, and motion masking effects of HVS. In this paper, we generate the watermark by the bit-wise exclusive-OR operation of a logo image and a random sequence, and we embed the watermark in the uncompressed video sequence. The amount of inserted watermarks is controlled by the peak-signal-to-noise ratio of the watermarked frame. Experimental results show that the proposed method is imperceptible and robust against various attacks with a good watermark capacity.

In this paper, we develop a robust image data hiding scheme based on multiple description coding of the signature image. At the transmitter, the signature image is encoded by balanced two-description scalar quantizers in the wavelet transform domain. The information of the two descriptions are embedded in the host image in the spatial domain with a masking factor derived from the gradient of the image intensity values. At the receiver, the multiple description decoder combines the information of each description and reconstructs the original signature image. We experiment the proposed scheme for embedding a gray-scale signature image of 128×128 pixels size in the spatial domain of a gray-scale host image of 512×512 pixels. Results show that data embedding based on multiple description coding has low visible distortions in the host image and robustness to various signal processing and geometrical attacks, such as addition of noise, quantization, cropping and down-sampling.

In the last years the event-based communication paradigm has been largely studied and considered as a promising approach to develop the commu-nication infrastructure of distributed systems. It is particularly interesting when easy reconfiguration and decoupling among components is required. In fact, it allows events to be propagated in a way that is completely hidden to the com-ponent that has generated them as well as to the receivers. RUBDES (Rule Based Distributed Event System) is a distributed event system that allows the use of composite events in publish/subscribe computational model. In this sys-tem, an event is represented as an object and a rule is represented as an expres-sion or a function that is evaluated or executed depending on the occurrence of events. This paper presents the design details of RUBDES and usage of Rule Definition Language (RDL) to describe composite events and event filters

In this paper, we present a novel buffer management algorithm for interactive multimedia streaming workload. We carefully examine the workload traces obtained from several streaming servers in service. The analysis results show that most users exhibit non-sequential access pattern using VCR-like operations such as jump backward and jump forward. Moreover, short jump accesses are shown to be common. We exploit the workload characteristics of the VCR-like operation and develop a buffer caching algorithm called Virtual Interval Caching scheme. Experimental results show that the proposed buffer management scheme yields better performance than the legacy schemes.

The computer-based sharing and dissemination of knowledge and learning activities are known as 'E-learning'. In this paper we propose an E-learning model, focusing in particular on courses modeling, which aims at promoting the sharing and reuse of courses contents and teaching materials, allowing the construction of personalized learning paths.

In this paper, a comprehensive e-tutoring service framework is presented. It comprises of a set of individual applications implemented based on the use of existing open source tools. The significant point here is the flexibility of the implementation proposed, that allows it to use a combination of unicast and multicast without posing any requirements for the supporting network infrastructure. This is achieved through the use of an overlay network architecture that allows for dynamic configuration of the media transmission and relay points.

The agent controllability of the environment is investigated using simple deterministic interacting automata pair model. For this purpose, a general extended design approach for such couple is developed. In the experiments, we focused on simple binary state case and generated stability/cycle behavior characteristics map for the couple. Examinations on the map showed that the behavior of the couple shows some initial value sensitivity. Also, we observed some non-controllable agent/environment couple definition which may imply an inherent communication border between agent and the environment.

The need for explicit time representation in workflow environments has been recently identified. Although the concept of time is an intrinsic part of a workflow application, time management has been treated as an issue of ordering the constructs of the workflow specification. This traditional approach aids mainly in the monitoring of activity deadlines. There is limited support for modelling and representation of time constraints associated with the processes and their activities or cases. This paper presents a temporal formalism for representing workflow specifications. We further argue that integrating workflows with temporal databases can further enhance the workflow specification. We propose a way for encoding the history of processes as well as the history of activities and their cases-items with respect to the time dimension and temporal uncertainty. Overall this paper is proposing a framework for modelling time and temporal uncertainty as an intrinsic part of the process dimension or control flow dimension.

We consider a tool- and setup-constrained short-term capacity allocation problem that arises in operational level planning at a semiconductor wafer fabrication facility.We formulate this problem as a degree-constrained network flow problem on a bipartite graph. We show that the problem is NP-hard and propose the first constant factor (1/2) approximation algorithms. Experimental study demonstrates that, in practice, our algorithms give solutions that are on the average less than 1.5% away from the optimal solution in less than a second.

We show how the notion of relational diagram, introduced by Schmidt, can be used to give a single demonic de nition for a wide range of programming constructs. Our main result is Theorem 11, where we show that the input-output( I/O) relation of a compound diagram is equal to that of the diagram in which each sub-diagram has been replaced by its input-output relation. This process is repeated until we obtain elementary diagrams to which we apply the results given in previous work.This is achieved by using monotypes and residuals.

In this study, a fuzzy logic based algorithm is presented for the concentration estimation of the CCl4 and CHCl3 gases by using the steady state sensor response and an artificial neural network (ANN) structure is proposed for the concentration estimation of the same gases inside the sensor response time by using the transient sensor response. The Quartz Crystal Microbalance (QCM) type sensors were used as gas sensors. A computer controlled measurement and automation system with IEEE 488 card was used to control of the gas concentration values and to collect the sensor responses. Acceptable performance was obtained for the concentration estimation with fuzzy inference. And the appropriateness of the artificial neural network for the gas concentration determination inside the sensor response time is observed.

Abstract. Although conventional compression tools achieve good compression rates, they ignore the security issue. This study presents the design and implementation of a lossless compression algorithm with embedded security to fill that gap. The scheme introduced is a system with encoding and decoding and is oriented for text type of data. It is implemented on sample text files from stan-dard English Calgary Corpus. Two ideas, one hiding the encryption key by using a PRNG and the other employing multiple iterations to dissipate language statistics, are suggested to strengthen the security of the system. Both ideas are implemented and promising results are obtained.

The handling of the sparse matrix vector product(SMVP) is a common kernel in many scientific applications. This kernel is an irregular problem, which has led to the development of several compressed storage formats such as CRS, CCS, and JDS among others. We propose an alternative storage format, the Transpose Jagged Diagonal Storage(TJDS), which is inspired from the Jagged Diagonal Storage format and makes no assumptions about the sparsity pattern of the matrix. We present a selection of sparse matrices and compare the storage requirements needed using JDS and TJDS formats, and we show that the TDJS format needs less storage space than the JDS format because the permutation array is not required. Another advantage of the proposed format is that although TJDS also suffers the drawback of indirect addressing, it does not need the permutation step after the computation of the SMVP.

The paper reports the development of a software platform, named PES (Parallelized Evolution System), that parallelizes the fitness evaluations of evolutionary methods over multiple computers connected via a network. The platform creates an infrastructure that allows the dispatching of fitness evaluations onto a group of computers, running both Windows and Linux operating systems, parallelizing the evolutionary process. PES is based on the PVM (Parallel Virtual Machine) library and consists of two components; (1) a server component, named PES-Server, that executes the basic evolutionary method, the management of the communication with the client computers, and (2) a client component, named PES-Client, that executes programs to evaluate a single individual and return the fitness back to the server. Performance of PES is tested for the problem of evolving behviors for a swarm of mobile robots simulated as physics-based models, and the speed-up characteristics are analyzed.

This paper presents the design and evaluation of a cache coherence adapter for the cache-coherent non-uniform memory access multiprocessor system in which symmetric multiprocessor (SMP) nodes are interconnected via Xcent-Net. The SMP node is a 4-way symmetric multiprocessor subsystem based on the Intel Xeon processors, and Xcent-Net is a dual, adaptive-routed, virtual cut-through multistage interconnection network composed of hierarchical crossbar routers. The cache coherence adapter contains a directory and a remote access cache to support the directory-based cache coherence protocol customized for the SMP nodes on Xcent-Net. For any cache design, cache size and cache line size are crucial design parameters in terms of performance and implementation, and thus they are extensively evaluated. According to the simulation results, it is shown that a 128-Mbyte remote access cache with 64-byte lines is the best choice, where the average data access latency is 9.4 ms and the effective bandwidth is 6.8 Mbytes/s per node.

Modern Distributed Shared Memory (DSM) systems offer high speed application processing by allowing to use relaxed consistency models, such as processor consistency. Unfortunately, most of the existing coherence protocols implementing relaxed consistency in multiprocessors or loosely couples clusters use write-update strategy which incurs large communication overhead, and therefore is impractical for most distributed applications. This paper presents a new home-based coherence protocol for DSM systems with processor consistency. The protocol uses local invalidation paradigm introducing little overhead. No additional invalidation messages are required; all coherence information is piggybacked to update messages.

We show a two-phase approach to address the minimization of different communication-cost metrics in fine-grain partitioning of sparse matrices for parallel processing. In the first phase, we obtain a partitioning with existing tools on the matrix to determine computational loads of the processor. In the second phase, we try to minimize communication-cost metrics. For this purpose, we develop communication-hypergraph and partitioning models. We experimentally evaluate the contributions on a PC cluster.

There are several alternative mechanisms for disseminating information among a group of participants in a distributed environment. An efficient model is to use epidemic algorithms that involve pair-wise propagation of information. These algorithms are based on the theory of epidemics which studies the spreading of infectious diseases through a population. Epidemic protocols are simple, scale well and robust again common failures, and provide eventual consistency as well. They have been mainly utilized in a large set of applications for resolving inconsistencies in distributed database updates, failure detection, reliable multicasting, network news distribution, scalable system management, and resource discovery. A popular distribution model based on the theory of epidem-ics is the anti-entropy. In this study, we focus on pull-based anti-entropy model used for multicast reliability as a case study, demonstrate its scalability and robustness, and give our comparative simulation results discussing the performance of the approach on a range of typical scenarios.

The static staging heuristics proposed so far for data staging adhere to a method by which only one data item is processed at each communication step. Concurrent scheduling, on the other hand, allows a communication step to include more than one transfer simultaneously. This is possible because some requests are achievable through separate paths, which will yield a better performance. In this study, we propose the Extended Partial Path (EPP) heuristic as a way to work in conjunction with the concurrent scheduling. Our simulation results show that the EPP performs considerably better than a static staging heuristic in the literature.

This paper presents nonlinear filtering using dynamic neural networks (DNNs). In addition, the general usage of linear filtering structure is given. DNN which has a quasi-linear structure has been effectively used as a filter with fast training algorithm such as Levenberg-Marquardt method. The test results are shown that the performance of DNN as linear and nonlinear filters is satisfactory.

This paper introduces a new approach based on artificial neural networks (ANNs) to identify a number of linear dynamic systems with single recurrent neural model. The structure of single neural model is capable of dealing with systems up to a given maximum number. Single recurrent neural model is trained by the backpropagation with momentum. Total nine systems from first to third orders have been used to validate the approach presented in this work. The results have shown that the recurrent single neural model is very successful in identifying a number of systems. The new approach presented in this work provides simplicity, accuracy and compactness.

This paper investigates the use of artificial neural networks (ANN) in risk estimation of asset returns. Istanbul Stock Exchange (ISE) index (XU100) is studied with a mixture of experts ANN architecture using daily data over a 12-year period. Results are compared to feed-forward neural networks, multilayer perceptron (MLP) and radial basis function (RBF) networks and recurrent neural networks(RNN). They are also compared to widely accepted Exponential Generalized Autoregressive Conditional Heteroskedasticity EGARCH) volatility model. These results suggest that mixture of experts (MoE) have the strength to capture the volatility in index return series and prepares a valuable basis for financial decision making.

Eukaryotic cells are subdivided into functionally separate membrane enclosed compartments. Each compartment and its vicinity contain functionally linked proteins related to the activity of that cell compartment. In an eukaryotic cell, each protein is targeted to its specific cell localization where it exerts full function. Large scale genome analysis provides high number of putative genes to be characterized. Therefore, prediction of the subcellular localization of a newly identified protein is invaluable for determination of its function. For the prediction of subcellular localization, we have developed a method, Prediction to Subcellular Localization (P2SL), using the amino acid composition and order. The composition represents the global features in the full or partial sequences while the other represents the local features, e.g. motifs of different window sizes. This method can be used to predict the signal peptides(SP), the mitochondrial targeting peptides(mTP), the nuclear and the cytosolic sequences. For classification purposes, other than the well known traditional techniques, self-organizing maps and probabilistic approaches like k-nearest neighbours are used. The experiments are made with different parameters. P2SL is the first method which uses human genes only and has a full success rate on redundancy-reduced data sets. It will be available on the web-server http://cgh.ceng.metu.edu.tr/P2SL

In this paper, a fuzzy variance analysis model with interaction between explanatory variables using Tanaka's model is investigated. A linear programming model is formulated for measuring the value of response factors, based on Tanaka's model. The proposed model measures both values of response factors, as well as their interactions.

The proposed method is the preliminary step for a human-machine interaction system, in which a robot arm mimics the movements of a human arm, visualized through a camera set-up. In order to achieve this goal, the posture of a model joint, which simulates a human arm, is determined by finding the bending and yaw angles from captured images. The image analysis steps consist of preprocessing of noise via median filtering, thresholding and connected component analysis. The relation between the relative positions of these markers can be used to determine the unknown bending and yaw angles of the model joint. This information is further passed to a PUMA 760 robot arm to finalize the goal. The preliminary simulation results are promising to present that the proposed system can be utilized in a real environment in which a human (arm) can be mimicked by a machine with visual sensor.

An important facet of flexible querying and information retrieval is the ranking of the possible alternatives in the result. Adequate ranking information simplifies decision making and helps the user in finding faster the requested information. This paper deals with the construction of ranking methods, which are based on the use of extended possibilistic truth values. Extended possibilistic truth values are a flexible means to model query satisfaction, which additionally allow to deal with cases where some of the imposed query criteria are not applicable. Three alternative ranking approaches are presented and compared with each other based on their applicability in flexible database querying.

Mining useful information and helpful knowledge from data transactions is evolving as an important research area. Current online techniques for mining association rules identify the relationship among transactions using binary values. However, transactions with quantitative values are commonly encountered in real-life applications. In this paper, we address this problem by introducing a fuzzy adjacency lattice, and then integrate the lattice structure with linguistic weights in a way to reflect the importance of items. Experiments conducted using synthetic data show the effectiveness of the proposed method for online generation of weighted fuzzy association rules.

Protein-protein interactions are key to understanding biological processes and disease mechanisms in organisms. There is a vast amount of data on proteins waiting to be explored. In this paper, we describe application of data mining techniques, namely association rule mining and ID3 classification, to the problem of predicting protein-protein interactions. We have combined available interaction data and protein domain decomposition data to infer new interactions. Preliminary results show that our approach helps us find plausible rules to understand biological processes.

This paper describes a rule discovery system that has been developed as part of an ongoing research project. The system allows discovery of multi-relational rules using data from relational databases. The basic assumption of the system is that objects to be analyzed are stored in a set of tables. Multi-relational rules discovered would either be used in predicting an unknown object attribute value, or they can be used to see the hidden relationship between the objects’ attribute values. The rule discovery system, developed, was designed to use data available from any possible ‘connected’ schema where tables concerned are connected by foreign keys. In order to have a reasonable performance, the ‘hypotheses search’ algorithm was implemented to allow construction of new hypotheses by refining previously constructed hypotheses, thereby avoiding the work of re-computing.

The sudden expansion of the web and the use of the Internet has caused some research fields to regain (or even increase) its old popularity. Of them, text categorization aims at developing a classification system for assigning a number of predefined topic codes to the documents based on the knowledge accumulated in the training process. We propose a framework based on an automatic inductive classifier, called ILA, for text categorization, though this attempt is not a novel approach to the information retrieval community. Our motivation are two folds. One is that there is still much to do for efficient and effective classifiers. The second is of ILA's (Inductive Learning Algorithm) well-known ability in capturing by canonical rules the distinctive features of text categories. Our results with respect to the Reuters 21578 corpora indicate (1) the reduction of features by information gain measurement down to 20 is essentially as good as the case where one would have more features; (2) recall/precision breakeven points of our algorithm without tuning over top 10 categories are comparable to other text categorization, namely similarity based matching, naive Bayes, Bayes nets, decision trees, and linear support vector machines.

A data stream is a massive unbounded sequence of data elements continuously generated at a rapid rate. Due to this reason, most algorithms for data streams sacrifice the correctness of their results for fast processing time. This paper proposes a clustering method over a data stream based on statistical ¥ì-partition. The multi-dimensional space of a data domain is divided into a set of mutually exclusive equal-size initial cells. A cell maintains the distribution statistics of data elements in its range. Based on the distribution statistics of a cell, a dense cell is dynamically split into two mutually exclusive smaller cells called intermediate cells. Eventually, the dense sub-range of an initial cell is recursively partitioned until it becomes the smallest cell called a unit cell. A cluster of a data stream is a group of adjacent dense unit cells. As the size of a unit cell is set to be smaller, the resulting set of clusters is more accurately identified. Through a series of experiments, the performance of the proposed algorithm is comparatively analyzed.

The technological advances in communication networks and computers are largely contributed to the developments of multimedia applications such as the videoconference. These applications require an efficient management of the Quality of Service (QoS) and consist of a great number of participants. In order to support such applications, we propose a hierarchical communication architecture and a multicast protocol allowing scalability. This architecture allows to guarantee the QoS. In fact, it allows to reduce efficiently the connection number and the communication delays. We examined and compared the performance of the proposed protocol and some existing protocols with simulations in NS2 tool.

Recently, real-time packet schedulers based on Earliest Deadline First(EDF) policy have been extensively studied to support end-to-end bounded delay for real-time traffic. However, the packet scheduler could not guarantee the QoS requirements of real-time traffic since it receives a number of non-real-time traffic for the purpose of management and control. In this paper, we study a packet scheduling scheme servicing non-real-time traffic using the available time of link (i.e. slack time). Proposed scheme assigns the deadline to non-real-time packets on-line and services them under EDF policy. Since proposed scheme services the non-real-time traffic when the link bandwidth is not used, it can guarantee the schedulability of real-time flows.

We present a new optimal scheduling algorithm called Periodic Write-Read-Compute (PWRC) scheduling for stream based parallel video processing. Although PWRC scheduling exploits the properties of the video data, it is applicable to any type of periodic data over which a data independent application is to run. The PWRC algorithm is designed considering a bust based parallel architecture allowing point-to-point communication between the host and the workers. A client-server system software is designed for the parallel system and SPMD type of programming is used. The PWRC requires a high level atomic write-read command for data transmission. Such an indivisible command couple can be created in various ways depending on the system capabilities. We analysed the PWRC scheduling algorithm by the help of a few cost model, which comprise system and application characteristics. The analysis of the cost model provides information either to form a parallel video processing system or to predict the overall performance of an existing system. A data transfer bandwidth constraint is also defined, in order to ensure that the system running the PWRC scheduling algorithm meets real-time requirements of video processing.

This paper introduces a new error-resilient mechanism based on video redundancy coding (VRC) of H.263 Version 2, formerly known as H.263+. VRC is a mechanism to achieve temporal error resilience in error-prone environments. However, VRC is not suitable for the time-varying error-prone channel. The proposed network-aware VRC mechanism adaptively changes the prediction structure with multiple threads according to the channel status. Also, scene-adaptiveness is incorporated with the network-aware VRC to reduce the motion-jerkiness occurred by an abrupt scene change. Simulation results show that the proposed adaptive VRC mechanism does not only utilize network resources efficiently, but also reduce the fluctuation of the video quality.

Neural-network-based optical network restoration is illustrated over an example in which multiple classes of traffic are considered. Over the preplanned primary and backup capacity, optimal routing and wavelength assignment is carried out. In case of a network failure, protection routes and optimum flow values on these protection routes are extracted from a previously trained feed-forward neural network which is distributed over the optical data communications network.

Delegation in computer systems plays an important role in relieving security officer's management efforts, especially in a large-scale, highly decentralized environment. By distributing management authorities to a number of delegatees, scalable and manageable security management functionality can be achieved. Recently, a number of researches are proposed to incorporate delegation concept into Role-Based Access Control(RBAC) model, which is becoming a promising model for enterprise environment with various organization structures. In this paper, we propose a new role-based delegation model using sub-role hierarchies supporting restricted inheritance functionality, in which security administrator can easily control permission inheritance behavior using sub-roles. Also, we describe how role-based user-to-user, role-to-role delegations are accomplished in the proposed model and analyze our delegation model against various delegation characteristics.

Collection, storage and analysis of multiple hosts’ audit trails in a distributed manner are known as a major requirement, as well as a major challenge for enterprise-scale computing environments. To ease these tasks, and to provide a central management facility, a software-suit, named as “Log-Hunter” has been developed. Log-Hunter is a secure distributed log server system which involves log collection and consolidation in a large-scale environment having multiple hosts that keeps at least one audit trail. This architecture also eases the inspection and monitoring of the audit trails generated on multiple hosts. By consolidating all the audit trails on a centralized server, it significantly reduces the manpower requirement, and also provides secure log storage for inspection of log entries as it becomes necessary. This paper presents the functional specifications, architecture and some preliminary performance results of the Log-Hunter.

This research features a new method for skin color segmentation using a 2D plane in the RGB color space. The RGB color values of the input color image do not need to be converted into HSI or YIQ color coordinates that have popularly been used for color segmentation. We have observed an important fact that skin colors in the RGB color space are approximately distributed in a linear fashion. Based on this fact, we have applied PCA (Principal Component Analysis) techniques to RGB values of skin colors from a set of training images. We detect skin regions by the lookup of skin color histogram computed based on a 2D color plane of which two axes correspond to two directions with smallest spread of skin colors. The proposed 2D color plane for color histogram lookup has an advantage over HS or IQ color planes. By using this plane, the problem of color constancy is much relieved. A learned color histogram contains most skin colors detected in the input images and at the same time, the distribution of skin colors in the plane is invariant compared to those in the HS or IQ planes. We have evaluated the performance of the proposed method by comparing with the performance of color histogram lookup methods based on HS or IQ color plane. The experimental results show that the performance of our method is robust to illumination changes.

In recent years, a number of mixtures of local PCA models have been proposed. Most of these models require the user to set the number of submodels (local models) in the mixture and submodels dimensionalities (i.e. number of PC's). To make the model free of parameters, we propose a greedy EM algorithm to find a suboptimal number of submodels. For a global retained variance ratio, the proposed algorithm estimates for each submodel the dimensionality that retains the given variability ratio. We test the proposed method on texture segmentation.

2D Gel Electrophoresis (2DGE) image is the most widely used method for the isolation of the objective protein by comparative analysis of the protein spot pattern in the gel plane. The process of protein analysis is the method, which compares the protein pattern that is spread in the gel plane with the contrast group and finds interesting protein spot by image analysis. Previous 2DGE image analysis is composed of gaussian fitting, and segments protein spots by watersheds, a morphological segmentation. Watersheds have a benefit that is fast in global threshold, but induces under-segmentation and over-segmentation of spot area when gray level is continuous. The drawback was somewhat solved by marker point institution, but needs the split and merge process. This paper introduces a novel segmentation of protein spots by watersheds using hierarchical threshold, which can resolve the problem of marker-driven watersheds.

This paper presents a study for recognizing isolated Turkish handwritten uppercase letters. In the study, first of all, a Turkish Handprint Character Database has been created from the students in Istanbul Technical University (ITU). There are about 20000 uppercase and 7000 digit samples in this database. Several feature extraction and classification techniques are realized and combined to find the best recognition system for Turkish characters. Features, obtained from Karhunen-Loéve Transform, Zernike Moments, Angular Radial Transform and Geometric Features, are classified with Artificial Neural Networks, K-Nearest Neighbor, Nearest Mean, Bayes, Parzen and Size Dependent Negative Log-Likelihood methods. Geometric moments, which are suitable for Turkish characters, are formed. KLT features are fused with other features since KLT gives the best recognition rate but has no information about the shape of the character where other methods have. The fused features of KLT and ART classified by SDNLL gives the best result for Turkish characters in the experiments.

A hyper-star graph HS(m,k) has been introduced as a class of lower cost interconnection networks. Hyper-star graph has more merit than hypercube when degree X diameter is used as a cost measure. In other words, they have smaller degree and diameter than hypercubes. In this paper, we consider some of the important properties of hyper-star graphs such as symmetry, w-diameter, and fault diameter. We show that HS(2n,n) is node-symmetric. We also show that the w-diameter of HS(2n,n) is bounded by the shortest path length plus 4, and fault diameter of HS(2n,n) is bounded by its diameter plus 2. In addition, we introduce an efficient broadcasting scheme in hyper-star graphs based on a spanning tree with minimum height.

When a node in a multicast tree fails, the tree is divided into several fragments. To achieve a fault-tolerant communication, failure recovery schemes are necessary to restore the tree. We present a simple recovery scheme for overlay multicast trees that involves only failure-neighboring nodes into the restoration and keeps the original structure of the rest of the tree. The scheme is based on virtual bypass rings providing alternative paths to eliminate the faulty node and reroute the traffic. Our scheme is scalable, independent of message source and traffic direction in the tree, restores the multicast tree in real time without a significant delay penalty and our experiments show that it is efficient even un-der a heavy traffic in the tree.

This paper proposes an information system that classifies web pages according a taxonomy, which is mainly used from seven search engines/ directories. The proposed classifier is a four-layer Generalised Regression Neural Network (GRNN) that aims to perform the information segmentation according to web page features. Many types of web pages were used in order to evaluate the robustness of the method, since no restrictions were imposed except for the language of the content, which is English. The system can be used as an assistant and consultative tool in order to help the work of human editors.

Effects of network parameters on the performance of mobile ad hoc networks (MANETs) have been widely investigated. However, there are certain issues related to the hardware implementation and internal parameters of the nodes present in the network that effect performance dramatically. In this work a source routed wireless network is implemented in VHDL. Basic modules in the design are mobile ad hoc nodes and a switch that emulates the wireless channel and controls the connectivity between nodes. By varying hardware parameters of the nodes under different network conditions performance of the overall system is measured and simulation results are presented.

In this paper, we present performance results of our previously developed rate control algorithm under various controlled packet loss rates. Since IP has no QoS support for real-time delivery of multimedia data, QoS control service has to be provided by the applications. Unlike traditional rate control approaches that take into account only the loss statistics, our algorithm integrated receiver buffer status into the feedback messages from the clients. Under various packet loss rates, we compared our approach to traditional two watermark approach and observed that our algorithm is more robust to varying network conditions.

The new emerging video coding standard H.264 employs variable block size motion compensation using multiple references with quarter-pel motion vector accuracy. This scheme is a key feature to accomplish higher coding gain, however, also a decisive factor that increases overall computational complexity. To overcome this, we propose a novel fast mode decision scheme suited for variable block sizes by classifying coding modes based on rate-distortion cost. The experimental results show that the proposed method provides significant reduction in computational complexity without any noticeable coding loss and additional operations.

The image quality enhancement based on digital signal processing (DSP) has been addressed as one of the crucial issues in digital display technology and market. Since real-time digital imaging should be controlled in limited speed and bits, colored motion pictures with digital display generally generate many problems such as limited operating speed of algorithm and gamut range of expressed color compared to analog cathode ray tube (CRT) display. Especially, the contrast enhancement with limited values sometimes may result in overpassing contrast limit. The changes of hue and saturation cause a damage of color nature of original image. Moreover, the abruptly changed gain during contrast enhancement causes the flickering problem as time passes. This paper introduces the image contrast enhancement algorithm named Advanced Contrast Enhancement (ACE) with adaptive and recursive method which operates at 135MHz, SXGA, keeps the color tone of original image and prevents gain from an abrupt change. The chip with ACE algorithm will be useful not only for general imaging but also for special treatment such as medical imaging to detect problems in human body pictures such as X-ray picture, ultrasound, and CT in real-time.

Variable-length codes (VLCs) improve coding performance using statistical characteristics of source symbols; however, VLCs have disastrous effects from bit errors in noisy transmission environments. In order to overcome problems with VLCs, reversible variable-length codes (RVLCs) have been introduced as one of the error resilience tools due to their error recovering capability for corrupted video streams. Still, existing RVLCs are complicated in the design and have some rooms for improvement in coding efficiency. In this paper, we propose a new design method for a symmetrical RVLC from the optimal Huffman code table. The proposed algorithm has a simpler construction process and also demonstrates an improved performance in terms of the average codeword length than other symmetrical RVLC algorithms.

Input-queued switch is useful for high bandwidth switches and routers because of lower complexity and fewer circuits than output-queued switch.However, it suffers HOL-blocking, which limits the throughput to 58%. To overcome HOL-blocking problem, many input-queued switches are controlled by sophisticated scheduling algorithms at centralized schedulers which restrict the design of the switch architecture. In this paper, we propose a simple scheduler called Pipelined and Prioritized Round Robin(PPRR) which is intrinsically distributed by each input-port. An iterative prioritized round robin scheduling algorithm in a pipelined fashion is provided. The proposed algorithm has less complexity and yet comparable performance with respect to other algorithms such as iSLIP and RPA. The effectiveness of PPRR is demonstrated with simulations under uniform and bursty traffic conditions.

In this work we examine the performance of arbitration algorithms for core high-speed serial switches (HSSS). Taking Infiniband as an example, experimental results show that, for a homogeneous network load, the average queuing delay for the switch output ports under the weighted round-robin (WRR) algorithm defined for InfiniBand is similar to the average queuing delay resulting from the largest delay-first (LDF) or first-in-first-out (FIFO) algorithms. For a 4-port switch, when the average load is high (close to 95% of the network capacity) and unbalanced with respect to the WRR weights, the difference in average delay between WRR and LDF or FIFO is large. However, for a 20-port switch and the same high unbalanced average network load, this difference is less than half of that of the 4-port switch. Further, the average queuing delay difference between WRR and LDF or FIFO diminishes quickly as the average load decreases. Hence, even for a fairly high average load, LDF or FIFO can be suggested for arbitration in core HSSS.