Computer Engineering

King for their help. v I would like to thank my sister, Mukadder (Dağtekin) Bingöl, for selflessly coming here to help me during my most difficult years. I thank my parents, Ahmet and Hatice, for their unconditional love and support. They have been working like no other for as long as remember to see their children educated and taken care for. I thank my sister Meral Dağtekin and Fatih Volkan Dağtekin for their support and love. I must also thank my extended family. Well, if I forgot any names please forgive me. vi Table of contents LIST OF TABLES .

Presented herein is a fully integrated low-noise CMOS multi-channel amplifier for neural recording applications. The circuit employs the chopper modulation technique to reduce the effect of flicker noise and DC offset. A reduced area design implementation is achieved by trading off the increased noise margin performance of the chopper modulator for minimal amplifier area and analog multiplexing of the recording sites. A fully differential topology is used for the signal path to improve noise immunity. The analog amplifier exhibits 56 dB of gain with a 115 kHz bandwidth and a common mode rejection ratio (CMRR) of 80 dB. Simulation results show a total input referred noise less than 16 nV/√Hz. The system power consumption is approximately 750 µ µ µ µWatts. The fully integrated system was designed in ABN 1.6-um single poly n-well CMOS process.

A video processing and display system for performing high speed geometrical image transformations has been designed. It involves looking up the video image by using a pointer memory. The system supports any video format which does not exceed the clock rate that the system supports. It also is capable of changing the brightness and colormap of the image through hardware.

In this study, hybrid methods are proposed for feature selection and classification of gene expression datasets. In the proposed genetic algorithm/support vector machine (GA-SVM) and genetic algorithm/k nearest neighbor (GA-KNN) hybrid methods, genetic algorithm is improved using Pearson's correlation coefficient, Relief-F, or mutual information. Crossover and selection operations of the genetic algorithm are specialized. Eight different gene expression datasets are used for classification process. The classification performances of the proposed methods are compared with the traditional GA-KNN and GA-SVM wrapper methods and other studies in the literature. Classification results demonstrate that higher accuracy rates are obtained with the proposed methods compared to the other methods for all datasets.

This work is licensed under Creative Commons Attribution-NonCommercial 4.0 International License ÖZ Bilgisayar teknolojisinin hızlı gelişmesi sonucunda çok büyük miktarlarda verilerle çalışmak olağan bir durum haline gelmiştir. Bu verilerin işlenmesi, verilerden anlamlı bilgiler çıkarılması ve kararlar alınması için makine öğrenmesine ihtiyaç duyulmaktadır. Görüntü içerisinden nesnelerin algılanmasına yönelik son zamanlarda yapılan çalışmalar özellikle Anlamsal Tabanlı Görüntü Erişimi alanına doğru yönelmektedir. Bu alanda yapılan çalışmalar ile Anlamsal Boşluk olarak adlandırılan ve görüntülerden makineler tarafından çıkarılan düşük düzeydeki renk, şekil, doku (color, shape, texture) özellikleri ile insanlar tarafından resimlerden algılanan ve yüksek düzey olarak ifade edilen kavramlar arasındaki uyuşmazlıkların giderilmesine yoğunlaşmaktadır. Bu amaçla, belirli bir bilgi alanına (domain) ait kavramların özelliklerini ve aralarındaki ilişkileri göstermek için iyi tanımlanmış ontolojiler oluşturulmakta ve arama işlemi bu yönde ilerlemektedir. Ontolojiler kullanılarak bilgiler bilgisayarların işleyebileceği biçime dönüştürülmekte ve bilgiler arasında anlamlı ilişkiler oluşturulabilmektedir. Bu çalışmada Anlamsal Tabanlı Görüntü Erişimi (Semantic Based Image Retrieval-SBIR) üzerine bir derleme yapılmıştır. SBIR ile amaç İçerik Tabanlı Görüntü Erişimi (Content Based Image Retrieval-CBIR) ile yapılan arama işlemlerinde karşılaşılan ve Anlamsal Boşluk (Semantic Gap) olarak ifade edilen darboğazın aşılmasıdır. Yapılan çalışmalarda Ontoloji (Ontology) kavramının kullanılmasıyla problemin çözümünde önemli bir gelişme yaşandığı gözlemlenmiştir.

Wireless Ad hoc network is a network, which does not have any geographical restriction and fixed substructure. This kind of networks are divided into cellular and special networks. Cellular networks are comprised of a fixed spinal column and a radio-base station. When a node exits from the domain of base station and reconnects to another station, it continues to function. In this method, the base station is fixed but the connection of Special Network does not have a predetermined structure. Administration and configuration of these kinds of networks are not dependent on any special user. Such networks back up calculation at any time and in any place and their structures can change automatically. In special networks since all of the nodes are mobile and connect dynamically they are commonly named Ad hoc. There is increasing trend for applications in such fields as mobile vehicular communication, special field tasks and distributed mobile automation systems. In this kind of network, all of the nodes act as routers; they find paths and keep other nodes' information. In this paper, we discussed special networks, improved their routing protocol called DSR (unicast), and introduced new protocol named GBDSR (multicast). GBDSR allows a node located in the network to send data packets for other nodes. In GBDSR, multi distribution of data packets are turned into general distribution of data packets. This leads to have a new protocol with less network traffic, less access time, less packet loss and less energy usage. Of course, these advantages need more use of bandwidths.

Wireless Sensor Network consist of large number of sensing devices, which are equipped with limited radio communication capabilities and limited computing which is used for many purposes in the present and the future. Clustering and routing in wireless sensor networks for the transfer of information from sensor nodes to base station are especially important. Energy is one the most important items to determine the network lifetime due to low power energy nodes included in the network. For improving this major parameter we have to design best network architecture and consider all impact components. In this article, firstly we discuss all important concepts in WSN architecture and the impact factors that effect in performance directly or indirectly, and then we focus to power management to find best design. Finally we propose a power usage model with considering these discussed impact factors.

Cognitive impact factor is one of the main measures for media-human interaction. In this study, we explain design and implementation of a semi-autonomous humanoid character display system that provides high cognitive impact on people. It combines marketing, entertainment, announcement, guidance and service sales on one singular device. As different from conventional automated teller machine (ATM) approach, interactive humanoid character (IHC) first detects the people around and targets starting the communication with them. A finite state machine model is used to determine type of possible requirement of the person in communication by navigating in pre-defined behavioral states due to interaction. This way of auto-controlled telepresence saves marketing, sales and service man-hours while it maximizes the media reach factor.

In optical networks, a connection is used to transmit data between source and destination nodes via lightpaths. The optical signal transmitted along a lightpath requires cross-connect switches (OXCs) and telecommunication carriers to switch high-speed optical signals in a fiber optic network. As a signal propagates from the source to the destination, the signal quality is continuously degraded by optical network components and impairments. Wavelength routing networks have two main problems: network design (fiber topology) and traffic requirements (traffic matrix). The network design problem is divided into lightpath topology design (LTD) and routing and wavelength assignment (RWA) problems. The RWA problem is more important for increasing the effectiveness of optical networks. The lightpath routing sub-problem requires determining the physical links for each lightpath that consist of optical channels. The wavelength assignment (WA) sub-problem requires determining the wavelength that each lightpath uses, i.e., assigning a wavelength to each lightpath in the logical topology such that wavelength restrictions are obeyed for each physical link. In this article, we discuss the important concepts of optical networks and the factors that affect in RWAs directly or indirectly.

This article presents a comparative study between Optical Burst Switching (OBS) and Optical Packet Switching (OPS) with variable-size packets on mono-service and multiservice network with Quality of Service (QoS) support. In terms of performance criteria, loss rate in byte and access delay are considered. A 2-state Markov Modulated Poisson Process (MMPP) traffic generator is used. A 14-nodes mesh network topology illustrating NSFNET and a ring topology are designed as the testbed. NS2 Network Simulation tool is used for our tests. According to the simulation results, we have seen that in both mono-service and multi-service networks, OPS is better than OBS under low traffic loads. However, OBS performs better than OPS in mono-service networks and low priority service classes under high traffic loads. In this study, we have shown that choosing one of these two switching technologies (OPS or OBS) in the future depends on the traffic load and the topology of network, together with the considered service differentiation scheme.

In parallel with the enormous spread of Internet usage, the tendency of users for high capacity applications (HDTV, Online Games, P2P File Transfer) are rising. This increase in needs of data communication, compel the service providers to act to end users better than before. At this point, Passive Optical Networks (PONs) distinguish in access network solutions due to its capability of carrying high amount of data over longer distances. Although they are still in development, existing electronic based data communication techniques (ADSL, double twisted copper telephony systems like VDSL, cable modem coaxial cable CATV systems) can operate over short distances and have large amount of energy consumption. In this paper, basic schemes of dynamic bandwidth allocation algorithms on EPON have been presented and a novel online-offline based method called hcDBA (Half Cycling Dynamic Bandwidth Allocation) has been introduced. In addition, hcDBA algorithm has been improved with early prediction mechanism (p-hcDBA). By the simulation studies, it has been investigated that hcDBA algorithm performs better than existing mechanism in terms of byte loss ratio and access delays criteria. p-hcDBA also improved the performance in terms of access delay. In this study, we stated that when the bandwidth increases the prediction approached has more impact of the overall performance.

In this study, Earth’s magnetic field signals have been investigated to determine mobile user’s location. In theory, Earth’s magnetic field does not change during the day at a certain point. But, the various noise effects that are exposed during the measurement causes deviations in the measured signal. In this study; Kalman Filter, LOESS, Savitzky-Golay filters are adapted with two different approaches to purge Earth’s magnetic field values from noise. KNearest Neighbour and Random Forest models have been trained with filtered signals and the locations of the mobile user are determined. Relevant systems have been tested by using RFKONDB which is existed in literature. The purpose of this study is to measure how these filters should be adapted to an Earth’s magnetic field based indoor localization systems. Digital sensors, which are integrated mobile devices, can use different measurement techniques. In a heterogeneous environment, noise reduction filters can show a different effect....

This study is about a prediction approach for our previous dynamic bandwidth allocation algorithm for Ethernet Passive Optical Networks (EPON). Our previous work (hcDBA) is based on half cycle timing for bandwidth allocation. That can be handled as a middle way between online and offline bandwidth allocation schemes. In PONs, prediction is used for bandwidth allocation to grant loaded nodes with early responses. However, due to the versatile nature of data traffic, prediction algorithms have a handicap to provide a better solution in classical approaches. In this study, a novel prediction approach integrated with hcDBA algorithm described. Performance comparison of hcDBA with & without prediction and IPACT algorithm is given. According to the simulation results, prediction on hcDBA seems to give some performance improvements in terms of access-delay.

This article presents a comparative study between Optical Burst Switching (OBS) and Optical Packet Switching (OPS) with variable-size packets on mono-service and multiservice network with Quality of Service (QoS) support. In terms of performance criteria, loss rate in byte and access delay are considered. A 2-state Markov Modulated Poisson Process (MMPP) traffic generator is used. A 14-nodes mesh network topology illustrating NSFNET and a ring topology are designed as the testbed. NS2 Network Simulation tool is used for our tests. According to the simulation results, we have seen that in both mono-service and multi-service networks, OPS is better than OBS under low traffic loads. However, OBS performs better than OPS in mono-service networks and low priority service classes under high traffic loads. In this study, we have shown that choosing one of these two switching technologies (OPS or OBS) in the future depends on the traffic load and the topology of network, together with the considered service differentiation scheme.

A performance comparison according to different number of wavelengths and topologies on OBS has been studied in this paper. Preemptive Channel Scheduling Algorithm (PCSA) has been used as reservation mechanism in OBS. In terms of performance criteria, loss rate in bytes, access delay and end-to-end delay are considered. A 2-state MMPP (Markov Modulated Poisson Process) traffic generator is used. Four different types of Mesh and Ring topologies are used. NS2 Network Simulation tool is used for our tests. In OBS algorithms, bursts are created using a hybrid model that takes into account both timeout and maximum length threshold mechanisms. In nodes, in order to satisfy QoS requirements, priority based queuing and Regulative Wavelength Grouping (RWG) are used. In priority based queuing, packets (bursts) are sent according to their priority order. In this study, the effects of generated traffic according to the topologies and the effects of increased number of wavelengths are shown by access delays. According to the simulation studies, the success of byte drop rate increases while the number of wavelengths increases. The results that obtained on mesh topologies are better than the results of ring topologies according to our simulation results.

In this study, Earth's magnetic field signals have been investigated to determine mobile user's location. In theory, Earth's magnetic field does not change during the day at a certain point. But, the various noise effects that are exposed during the measurement causes deviations in the measured signal. In this study; Kalman Filter, LOESS, Savitzky-Golay filters are adapted with two different approaches to purge Earth's magnetic field values from noise. K-Nearest Neighbour and Random Forest models have been trained with filtered signals and the locations of the mobile user are determined. Relevant systems have been tested by using RFKONDB which is existed in literature. The purpose of this study is to measure how these filters should be adapted to an Earth's magnetic field based indoor localization systems. Digital sensors, which are integrated mobile devices, can use different measurement techniques. In a heterogeneous environment, noise reduction filters can show a different effect. Two different test scenarios and two different noise reduction models, with the 3 noise reduction techniques, are developed to find the best case.

According to the upstream TDM in the system of Ethernet passive optical network (EPON), this paper proposes a novel dynamic bandwidth allocation algorithm which supports the mechanism with correction-based the multiple services estimation. To improve the real-time performance of the bandwidth allocation, this algorithm forecasts the traffic of high priority services, and then preallocate bandwidth for various priority services is corrected according to Gaussian distribution characteristics, which will make traffic prediction closer to the real traffic. The simulation results show that proposed algorithm is better than the existing DBA algorithm. Not only can it meet the delay requirement of high priority services, but also control the delay abnormity of low priority services. In addition, with rectification scheme, it obviously improves the bandwidth utilization.

Dans ce travail de these, un apercu general sur les systemes PON est presente et sont etudies les mecanismes et classification PON existants. Apres, nous introduisons notre premiere contribution qui est un algorithme d'allocation dynamique de bande passante pour EPON. Cet algorithme propose est designe comme «hcDBA». Par la suite, une amelioration de l'algorithme de hcDBA avec mecanisme de prediction precoce est presentee. Notre simulation montre bien que notre algorithme hcDBA est performant compare aux mecanismes existants en termes de taux de perte de paquets et de delai d’acces. Dans notre seconde contribution, nous sommes interesses au probleme de consommation d’energie qui est un sujet d’actualite dans les reseaux de telecommunication. Les etudes montrent aujourd’hui que les reseaux d'acces constituent une partie remarquable de la consommation totale d'energie dans les reseaux de telecommunication. Les organisations ITU-T et IEEE ont publie la recommandation po...