Addis Ababa University Libraries Electronic Thesis and Dissertations: AAU-ETD!

Design and Performance Analysis of Energy Efficient Technique for Wireless Multimedia Sensor Networks using Machine Learning Algorithms

Thu, 13 Jun 2013 10:27:12 GMT

Title: Design and Performance Analysis of Energy Efficient Technique for Wireless Multimedia Sensor Networks using Machine Learning Algorithms

Authors: Kibrewerk, Akalu

Abstract: Wireless multimedia sensor networks (WMSNs) are developed from wireless sensor networks (WSNs) for acquiring and transmitting multimedia data such as images, audio and video streams and scalar data. Energy is the most critical factor in sensor networks. Its power requirement is satisfied by low capacity and low power battery. One of the reasons is the requirement of unattended operation in remote or even potentially hostile locations, sensor networks are extremely energy-limited. This constraint demands that techniques must not only be efficient but energy conscious as well, which requires new approach to addressing the common but substantial issues. Reduction of communicated multimedia volume is an important step to reduce energy consumption in WMSNs because of the relatively huge amount of data collected by the nodes compared with scalar sensors. One of the algorithms in machine learning which can reduce the dimensionality is unsupervised learning Artificial Neural Networks which typically perform dimensionality reduction through pattern clustering. In this thesis, an attempt has been made to reduce the amount of transmitted information in WMSNs using vector quantization technique using Self Organizing Map (SOM) algorithm in order to increase the lifetime of the network. In the Proposed Design, SOM is used to generate a codebook using single image and batch image training methods. An energy model has been designed to calculate the lifetime of the nodes taking into consideration the computational energy cost and communication energy cost. Using this energy model, the codebook size has been optimized to a size of 50 codewords through which the network lifetime has shown an increase of 158.03 percentages compared to the existing design. This amount of increase in the lifetime of the WMSNs is on a graceful-tradeoff with the image quality since the main purpose of sensor networks is the occurrence or non occurrence of things of interest rather than on excellence of image quality considering a surveillance which is the main application for the deployment of the sensors.

DEVELOPMENT OF A COMPUTER PROGRAM FOR STRUCTURAL DESIGN OF CONCRETE SLAB CULVERT DECK

Thu, 13 Jun 2013 09:19:25 GMT

Title: DEVELOPMENT OF A COMPUTER PROGRAM FOR STRUCTURAL DESIGN OF CONCRETE SLAB CULVERT DECK

Authors: Andinet, Zeleke

Abstract: Concrete slab culvert is an important structure used to convey trucks and pedestrian along a road corridor or in one of a range of other situations. This structure is highly constructed in highway road projects in Ethiopia. In this study, a FORTRAN program is developed for the structural design of reinforced concrete slab culvert deck according to the provisions given in AASHTO LRFD Bridge 2005 Edition. The developed program is expected to assist the structural designers and users to design the superstructure part of a reinforced concrete slab culvert deck efficiently with great accuracy. Both at grade and at fill slab deck thicknesses are computed according to the specification specified in AASHTO LRFD Bridge 2005 Edition. The reinforcement bars are also designed based on the requirements specified in the code. The program is developed in four steps. The first step is to define and analyze the problem; the second step is to develop an optimal solution and designing the program, the third step is coding the program and the final step is testing and documenting the program. FORTRAN 95 is a programming language used in the fields of scientific, numerical, and engineering fields. In this thesis, this language has been used to develop the program for the structural design of reinforced concrete slab culvert deck. The input data for at grade and at fill slab culverts are saved on a note pad in the external file folder which constitute the material properties, geometric features and proposed diameter of reinforcement bars of the slab culvert and its deck in the folder which contains FORTRAN 95 program. The output data is written on the note pad in the external folder based on the format assigned for each output in the folder which contains the design results of slab deck thickness and area, spacing and length of main, distribution and temperature reinforcement bars. Besides Edge beam design parallel to the traffic is executed and shown in the output result by the developed program.

MODELING OF SILICON NANOWIRE FIELD EFFECT TRANSISTORS

Thu, 13 Jun 2013 09:03:25 GMT

Title: MODELING OF SILICON NANOWIRE FIELD EFFECT TRANSISTORS

Authors: Kidist, Moges

Abstract: Modeling and simulation of Metal Oxide Semiconductor Field Effect Transistors (MOSFETs) is very essential in order to understand the device physics, electrostatics and other important phenomena occurring in this device. Therefore, in this thesis the modeling and simulation of Silicon Nano Wire field Effect Transistors (SiNW FETs) is done. The modeling is done assuming both ballistic transport and transport in the presence of scattering. The modeling of SiNW FETs assuming ballistic transport is an extension of the Natori’s theory of ballistic MOSFETs. The second part of the modeling, which is developed on the assumption of scattering transport, is based on McKelvey’s flux method. When the scattering effects are assumed to be absent, the scattering model reduces to the ballistic model. Therefore, the main novelty introduced in this thesis is the extension of the previous models and the incorporation of these two models together. After the derivation of the model, its benchmarking is also done. This is accomplished by comparing the simulation results of the developed model, which is implemented using MATLAB programming, with that of the experimental and numerical simulation results. Various important parameters are extracted and used for comparison, the main ones being the On-state current (Ion), the Off-state current (Ioff), the Subthreshold Slope (SS) and drain induced barrier lowering (DIBL). The comparison shows that there is a good agreement between the simulation results of the developed model and the experimental and numerical simulation results, which indicates the validity of the model. Finally, the effect of scaling of the physical parameters on the device performance is investigated. The main parameters chosen for this investigation are the diameter of the Nano Wire (NW) and the gate oxide thickness. When a simulation is done by varying these parameters, Ion and Ioff currents are found to be affected greatly.

Performance Evaluation of Adaptive Arrays for MIMO Smart Antenna Systems

Wed, 05 Jun 2013 09:11:50 GMT

Title: Performance Evaluation of Adaptive Arrays for MIMO Smart Antenna Systems

Authors: Fantaye, Mekonnen

Abstract: The demand for wireless systems has been growing rapidly over the recent years due to improved reliability, high data rates, seamless connectivity and low deployment costs. MIMO systems are the most efficient leading innovation of wireless systems for maximum capacity and improved quality and coverage. This theory has been around for a long while but the complexity involved and the signal processing required has been a major drawback to its widespread use. However, recent improvements in Digital Signal Processing (DSP) technology has made it possible to now construct such transmission systems. In this thesis we study different adaptive blind and nonblind algorithms for MIMO systems such as LMS, CMA, SMI, and combined algorithms, LMS-CMA, and SMI-CMA. Moreover, we compare these adaptive array algorithms with other known class of MIMO linear receiver (channel estimation) techniques like Zeroforcing (ZF) and minimum mean square error (MMSE) methods. In addition to this, we have discussed Capacity of MIMO systems and different MIMO transmission techniques such as spatial diversity (SD), Spatial multiplexing(SM). The results of performance evaluation for Adaptive array MIMO receivers revealed that LMS has better BER performance than SMI, SMI-CMA, and ZF and the same performance with MMSE with no need of CSI. LMS algorithm has slow convergence but low complexity compared to MMSE algorithm that has fast convergence with very high complexity. Moreover, the number of training signals can minimized by 62.5% at the cost of 2-4dB SNR using nonblind algorithm( LMS) combined with blind algorithm( CMA).