<p>This thesis contributes to the development of novel methods and techniques for computer-aided electromagnetics (EM)-based modeling and design of microwave circuits exploiting two previously unrelated technologies: space mapping (SM) and artificial neural networks (ANNs). The conventional approach to EM-based modeling of microwave circuits is reviewed, as well as other state-of-the-art neuromodeling techniques. The fundamental space mapping concept is also reviewed. Developing neuromodels based on space mapping technology is addressed. Several SM-based neuromodeling techniques are described. Contrast with other neuromodeling approaches is realized. An algorithmic procedure to design, called Neural Space Mapping (NSM) optimization, is described. NSM enhances an SM-based neuromodel at each iteration. Other techniques for optimization of microwave circuits using artificial neural networks are reviewed. Efficient EM-based statistical analysis and yield optimization of microwave components using SM-based neuromodels is described. Other yield-driven EM optimization strategies are briefly reviewed. An innovative strategy to avoid extra EM simulations when asymmetric variations in the physical parameters are assumed is described. Neural Inverse Space Mapping (NISM) optimization for EM-based microwave design is described. A neural network approximates the inverse mapping at each iteration. The NISM step simply consists of evaluating this neural network at the optimal empirical solution. NISM step is proved to be a quasi-Newton step when the amount of nonlinearity in the inverse neuromapping is small. NISM optimization is compared with other SM-based optimization algorithms. The theoretical developments are implemented using available software on several advanced and industrially relevant microwave circuits. Suggestions for further research are provided.</p> / Doctor of Philosophy (PhD)

<p>As the number of elderly persons as well as their fraction of the total population<br />continues to rise, especially in the developed countries, providing an appropriate living<br />environment for them using smart home technology is rapidly gaining attention. Two<br />important tasks of a smart home technology are monitoring the daily activities and<br />the vital signs of the elderly to improve their quality of life and to monitor existing<br />or the onset of health abnormalities. In this thesis, we focus on the monitoring of<br />taking medicine by the elderly person using vision sensors (low-cost cameras). This<br />task is important since it helps both the person and the doctor in the treatment<br />of illnesses of elderly persons. The allocated resources of communication bandwidth<br />between the sensor nodes and the computational power, used for this task, affect<br />the implementation cost. Therefore, it is desired to develop an effective scheme<br />which efficiently allocates bandwidth and computational resources to achieve a high<br />reliability (detection performance) at low cost.<br />In this thesis, we have proposed two different approaches to solve this detection<br />and monitoring problem. As the input data are video frames, captured by cameras<br />from the same scene, the frames have inter-view redundancy. Taking advantage of<br />this inter-view redundancy, we proposed a video coding classification scheme based<br />on separate encoding and joint decoding, and have obtained significant compression<br />improvement compared to existing techniques. In the second approach, we studied<br />different parts of the detection and monitoring system to find an efficient design<br />for distribution of different event detection parts between the nodes and the central<br />processing unit so that the allocated resources are reduced. In this scheme, the<br />useful information of the frames are extracted in the form of their main features<br />such that decision making based on these features is the same as decision making<br />based on the raw frames. As a result, we could propose a new scheme which requires<br />significantly less bandwidth and computational resources while achieving the same<br />detection performance.</p> / Master of Applied Science (MASc)

<p>In recent years there has been a growing interest in computer-based surgical planning,<br />virtual-reality enabled training of medical procedures, and computer gaming all involving non-rigid deformable objects. High-fidelity simulations of haptic interaction with deformable objects is computationally demanding. The Finite Element Method (FEM) is known to produce relatively accurate solution for continuum mechanics-based models of soft-object deformation. Linear elastic FE models require solving a large sparse system of equations. The solution accuracy can be improved by increasing the resolution of the finite element mesh resulting in a larger number of equations and hence greater computational complexity. Depending on the mechanical characteristics of the soft-object, to maintain stability and high fidelity in haptic interaction, the update rate should be in the range of 1001000Hz. This, for example, means that for a moderately-sized three-dimensional mesh of 6000 nodes, a set of 18000 linear equations must be solved within 1-10ms.</p> <p>In this thesis, hardware-based parallel computing is proposed for finite-element<br />(FE) analysis of soft-object deformation models. In particular, a distributed implementation of the (CG) algorithms on .\' Field Programmable Gate Array (FPGA)<br />devices connected in a ring configuration is developed. This Parallel architecture<br />can be utilized to solve the large system of equations arising from FE models at high update rates required for stable haptic interaction. Massive parallelization of the computations is achieved by customizing the hardware architecture to the problem at hand and employing a large number of adaptive fixed-point computing units in parallel. The proposed hardware architecture satisfies three important criteria: (i) it meets the haptic rendering timing constraint by enabling an update rate of 400Hz; (ii) it attempts to simulate as many nodes as possible, given the available resources on the FPGA devices employed in this work and (iii) it is scalable both within an FPGA and also across multiple FPGA devices.<br />This research builds upon our group's earlier work in [1]. In that paper a novel highly parallelized single-FPGA architecture was proposed for solving system of equations arising from FEM using Conjugate gradient method. In this thesis, a multiple-FPGA architecture based on that design has been proposed. The contributions in the new multiple-FPGA design can be summarized as follows.</p> / Master of Applied Science (MASc)

<p>The first part of this thesis is concerned with efficient adaptive image interpolation techniques for real-time applications. A new image interpolation algorithm is developed that combines optimal data fusion and context modeling of images. Specifically, two estimates of missing pixels obtained by cubic interpolation in perpendicular directions<br />are optimally fused under minimum mean square (MMSE) criterion. The fused result is further improved by a context-based error feedback mechanism to compensate for the error of cubic interpolation. The proposed image interpolation algorithm preserves edge structures well and achieves superior visual quality. This is accomplished at low computational complexity, making the new algorithm suitable for hardware implementation.</p> <p>The main part of this thesis is devoted to a more sophisticated image interpolation<br />algorithm based on hidden Markov modeling (HMM). Most of existing interpolation<br />algorithms rely on point by point decisions to estimate the missing pixels. In contrast,<br />the HMM approach of image interpolation estimates a block of missing pixels via maximum a posterior (MAP) sequence estimation. The hidden Markov model can<br />incorporate the statistics of high resolution images into the interpolation process and<br />the MAP estimation technique can exploit high-order statistical dependency between<br />pixels. The proposed HMM-based image interpolation algorithm is implemented and its performance is evaluated and compared with existing methods. The comparison<br />study shows that the HMM-based image interpolation algorithm can reproduce<br />cleaner and sharper image details than its predecessors, while suppressing common<br />interpolation artifacts such as ringing, jaggies, and blurring.<br /><br /></p> / Master of Applied Science (MASc)

<p>Home health care continues to be an important and challenging issue in most countries, especially for the elderly. To improve home health care, it would be useful to have wireless biomedical systems that can monitor various vital signs ofthe elderly and to provide such information to a health care professional. In the orient, one powerful tool in diagnosing and predicting health issues is the Traditional Chinese Pulse Diagnosis (TCPD) technique. The TCPD technique is through the examination ofthe artery pulse pattern on three points along the radial artery and it usually requires a doctor to conduct pulse palpation with their fingers. Therefore, the diagnosis relies significantly on the experience of the doctor. However, for the more widespread use of TCPD, one concept is to bring the doctor's "fingers" and "encode" the doctor's experience to an individual in a home care setting. Using existing technologies in biomedical sensor, data acquisition, communication and microelectronics, it is possible to construct a "smart" TCPD system.</p> <p>In this thesis, a microcontroller based pulse monitoring system for TCPD's application in home care is proposed. The system consists of three main units for data acquisition, data processing and wireless transmission. The pulse data acquisition is with a liquid-filled digital pressure sensor module with the employment of applanation tonometry, a technique used in recording the peripheral artery waveform. Each sensor module is read by a corresponding microcontroller via its serial peripheral interface, and the measurement is then sent wirelessly to a personal computer (PC) via a 204GHz transceiver. The system was used to successfully record and transmit radial pulse pressure and body surface temperature measurements to a host PC. Pulse waveforms are then reproduced from the pulse pressure measurements to conduct offline analysis. The analysis is targeted to integrating TCPD diagnosis with quantitative pulse representation and measurement history, to use expert knowledge in classification and recognition, and therefore to provide supporting information for disease diagnosis and forecasting.</p> <p>The work performed in this thesis presents the proof-of-concept research and system implementation in the design of an individual health monitoring system using the TCPD method. Basic pattern matching and parameter extraction/comparison are performed and verified. Finally, the research work presented here provides a solid foundation for future work in this field ofTCPD and its application.</p> / Master of Applied Science (MASc)

<p>Ubiquitous wireless sensor networks (WSNs) are expected to play an important role in the future society for various applications. As a result, carefully managing the network resources to improve the network performance becomes a hot research topic. In this thesis, we study the perfonnance ofWSNs with a cluster tree topology, where all the cluster heads (CHs) fonn a tree topology. The sensor nodes transmit data to their directly associated CHs, which forward the traffic to the sink through other CHs in the cluster tree.<br />We first study the associations between sensor nodes and the cluster heads (CHs). In a WSN where there is a strong overlapping coverage area between the CHs, associating the sensor nodes to different CHs may result in different network performance. As the sensor node associations affect the traffic load within each cluster and that between the clusters, timeline of the CHs should be allocated accordingly. We formulate three optimization problems by jointly considering the sensor node associations and CH timeline allocations. The objectives are maximizing the throughput per sensor node, balancing the energy consumption among the CHs, and maximizing the network level throughput, respectively. Conesponding to each of the objectives, a heuristic association scheme is designed and the timeline allocations of the CHs are calculated. Numerical results based on computer simulation demonstrate that the proposed schemes achieve close-to-optimum performance.</p> <p>In the second part of the thesis we study the end-to-end transmission delay for traffic at different levels of a WSN with the cluster tree topology. The end-to-end delay includes both local transmission delay between the sensor nodes and their directly associated CHs and inter-CH transmission delay between the forwarding CHs along the path to the sink. Given the timeline allocations of each CH for local and inter-cluster traffic transmissions, we find the distribution of the local traffic transmission delay and that of the inter-CH transmission delay. Based on these results, we then derive the distribution of the end-to end transmission delay and the packet drop rate due to excessive delay. The results provide important guidelines for allocating the CH time resources in order to achieve certain delay or packet drop rate performance. By appropriately allocating the CH time resources, it is possible that traffic traversing more hops to the sink experiences better delay performance than that traversing a fewer number of hops.</p> / Master of Applied Science (MASc)

<p>Free-Space Optical (FSO) communication systems have recently attracted considerable attention in last-mile applications. High bandwidth, unlicensed spectrum, ease of installation, and high security have made them a good candidate for high data rate transmissions. However, distance-dependent atmospheric turbulence and channel loss degrade the optical link reliability and confine FSO systems to short-haul applications. This thesis addresses innovative all-optical relaying techniques to mitigate the degrading effects of atmospheric turbulence-induced fading by relaying data from the source to the destination using intermediate terminals. The proposed techniques, optical amplify-and-forward (OAF) relaying and optical regenerate-and-forward (ORF) relaying, are deployed in multihop FSO systems to extend the maximum accessible communicating distance of high data rate wireless optical systems.<br /> In all-optical relaying techniques, photo detection is performed once at the receiver and intermediate terminals process optical field envelopes instead of optical intensities. This major difference requires a new definition of channel model for propagation of optical waves through the atmosphere. By using the developed channel model, bit error rate (BER) performance of multihop OAF FSO systems is analyzed through Monte-Carlo simulations. The simulation results indicate that OAF relaying technique mitigates the channel impairments and enhances the BER performance. By employing more relays, longer distances become accessible, however distance improvement decreases due to accumulating background noise at relays. In order to remove background noise effects, another optical relaying technique is developed. The ORF relaying technique eliminates the received background noise at each relay and significantly outperforms OAF systems. For example at high bit rate BR= 10 Gbps, using two equally-spaced OAF relays during a 3 km turbulence-free link increases the total communicating distance by about 1.11 km. Replacing OAF relays by ORF relays extends the total communicating distance to 4.48 km which is 1.66 km longer than the similar OAF FSO system. By deploying more ORF relays, even longer distances are achievable.</p> / Master of Applied Science (MASc)

<p>In this thesis an efficient approach to nonlinear non-Gaussian state estimation based on spline filtering is presented. The estimation of the conditional probability density of the unknown state can be ideally achieved through Bayes rule. However, the associated computational requirements make it impossible to implement this online filter in practice. In the general particle filtering problem, estimation accuracy increases with the number of particles at the expense of increased computational load. In this thesis, B-Spline interpolation is used to represent the density of the state pdf through a low order continuous polynomial. The motivation is to reduce the computational load and to improve accuracy. The motion of spline control points and corresponding coefficients is achieved through implementation of the Fokker-Planck equation, which describes the propagation of state probability density function between measurement instants. The solution of the Fokker Planck equation is achieved by calculating the state transition probability matrix. The state transition matrix is calculated using Dirac Feynman approximation. This filter is applicable for a general state estimation problem as no assumptions are made about the underlying probability density. Finally, simulation results are presented to demonstrate the effectiveness of the proposed algorithm.</p> / Master of Applied Science (MASc)

<p>Optical flow is a well known technique for the measurement of motion in images. Although it has many applications, calculating the optical flow remains computationally expensive and challenging to use in time-critical tasks. This thesis describes an accelerated approach to optical flow computation using foveation and parallel processing on a Graphics Processing Unit (GPU). Foveation reduces the amount of image data to process by mimicking the variable resolution structure of the human visual system. The resulting image data is processed in parallel on a 240 processor GPU to achieve high frame rates on high resolution images. The newly introduced Compute Unified Device Architecture (CUDA) framework is utilized to create an efficient mapping of optical flow and foveation algorithms to the GPU. <br /> The performance and error of the algorithm is characterized using synthetic and real data. The non-foveated optical flow algorithm is found to perform up to 100× faster than a CPU implementation. Foveated optical flow is found to give an additional performance gain of up to 27× over non-foveated optical flow with a corresponding increase in angular error. The results are shown to match or outperform FPGA and non-CUDA GPU implementations. Finally, the application of the described system to real-time control of a robot arm is demonstrated.</p> / Master of Applied Science (MASc)

<p>Sequentially plasma activated bonding (SPAB) of silicon wafers has been investigated to facilitate chemical free, room temperature and spontaneous bonding required for integration of nanostructure on the wafer scale. The SPAB consists of surface activation using reactive ion etching (RIE) plasma followed by microwave (MW) radicals. The drop shape analysis and atomic force microscopy (AFM) results show that O2 RIE plasma is the most efficient in removing surface contaminations while keeping smooth surfaces. On the other hand, MW N2 radicals offer highly reactive, smooth and hydrophilic surfaces. These highly reactive, smooth and hydrophilic surfaces allow strong and spontaneous bonding of silicon/silicon at room temperature. Electrical characteristics show that the current transportation across the nano-bonded interface is dependent on plasma parameters. The infrared images show that plasma induced voids' nucleation at the bonded interface is dominated by O2 RIE power over O2 RIE activation time. The bonding strength achieved at room temperature in SPAB is about 30 times higher than that in hydrophilic bonding.<br /> In order to explore the reliability of SPAB at high temperature, the bonded wafers are annealed from 200 to 900°C. The thermal induced voids' nucleation occurred preferentially at the plasma induced defect sites. The nucleation of void density is quantitatively determined and explained using high resolution transmission electron microscopy (HRTEM) observations. The electron energy loss spectroscopy results reveal the existence of silicon dioxide at the bonded interface. The reduction in bonding strength after annealing at high temperature is correlated to the increase in void density. The plasma induced defect sites such as nanopores and craters are identified using an AFM. The porous surface allows easy removal of interfacial water and spontaneous covalent bonding at room temperature. The HRTEM results confirm nanometer scale bonding which is needed for the integration of nanostructures. Based on the results, a bonding mechanism of SPAB is presented. <br /> In order to expand the applicability of SPAB for diverse materials, a novel hybrid plasma bonding (HPB) process is developed to achieve void-free and strong silicon/glass and germanium/glass bonding at low temperature. The HPB combines sequential plasma activation with anodic bonding process. Void-free interface with high bonding strength is observed both for silicon/glass and germanium/glass at 200oe. The bonding strength of the silicon/glass and germanium/glass in the HPB at 2000e is 30 MPa and 9.1 MPa, respectively. The improved characteristic behavior of the interface in the HPB is attributed to higher hydrophilicity and smooth surfaces of silicon, glass and germanium after sequential plasma activation and high electrostatic force associated with anodic bonding. Based on the results, a bonding mechanism of HPB is discussed. <br /> The chemical free strong bonding of silicon/silicon in SP AB at room temperature and void-free strong bonding of silicon/glass and germanium/glass in HPB at low temperature can be applied in spontaneous integration of nanostructures on the wafer scale.</p> / Master of Applied Science (MASc)