You are here
Search results
(21  26 of 26)
Pages
 Title
 Signal Processing Based Distortion Mitigation in Interferometric Radar Angular Velocity Estimation
 Creator
 Klinefelter, Eric
 Date
 2021
 Collection
 Electronic Theses & Dissertations
 Description

Interferometric angular velocity estimation is a relatively recent radar technique which uses a pair of widely spaced antenna elements and a correlation receiver to directly measure the angular velocity of a target. Traditional radar systems measure range, radial velocity (Doppler), and angle, while angular velocity is typically derived as the timerate change of the angle measurements. The noise associated with the derived angular velocity estimate is statistically correlated with the angle...
Show moreInterferometric angular velocity estimation is a relatively recent radar technique which uses a pair of widely spaced antenna elements and a correlation receiver to directly measure the angular velocity of a target. Traditional radar systems measure range, radial velocity (Doppler), and angle, while angular velocity is typically derived as the timerate change of the angle measurements. The noise associated with the derived angular velocity estimate is statistically correlated with the angle measurements, and thus provides no additional information to traditional state space trackers. Interferometric angular velocity estimation, on the other hand, provides an independent measurement, thus forming a basis in R2 for both position and velocity.While promising results have been presented for single target interferometric angular velocity estimation, there is a known issue which arises when multiple targets are present. The ideal interferometric response with multiple targets would contain only the mixing product between like targets across the antenna responses, yet instead, the mixing product between all targets is generated, resulting in unwanted `crossterms' or intermodulation distortion. To date, various hardware based methods have been presented, which are effective, though they tend to require an increased number of antenna elements, a larger physical system baseline, or signals with wide bandwidths. Presented here are novel methods for signal processing based interferometric angular velocity estimation distortion mitigation, which can be performed with only a single antenna pair and traditional continuouswave or frequencymodulated continuous wave signals.In this work, two classes of distortion mitigation methods are described: modelbased and response decomposition. Modelbased methods use a learned or analytic model with traditional nonlinear optimization techniques to arrive at angular velocity estimates based on the complete interferometric signal response. Response decomposition methods, on the other hand, aim to decompose the individual antenna responses into separate responses pertaining to each target, then associate like targets between antenna responses. By performing the correlation in this manner, the crossterms, which typically corrupt the interferometric response, are mitigated. It was found that due to the quadratic scaling of distortion terms, modelbased methods become exceedingly difficult as the number of targets grows large. Thus, the method of response decomposition is selected and results on measured radar signals are presented. For this, a custom singleboard millimeterwave interferometric radar was developed, and angular velocity measurements were performed in an enclosed environment consisting of two robotic targets. A set of experiments was designed to highlight easy, medium, and difficult cases for the response decomposition algorithm, and results are presented herein.
Show less
 Title
 Smartphonebased sensing systems for dataintensive applications
 Creator
 Moazzami, MohammadMahdi
 Date
 2017
 Collection
 Electronic Theses & Dissertations
 Description

"Supported by advanced sensing capabilities, increasing computational resources and the advances in Artificial Intelligence, smartphones have become our virtual companions in our daily life. An average modern smartphone is capable of handling a wide range of tasks including navigation, advanced image processing, speech processing, cross app data processing and etc. The key facet that is common in all of these applications is the data intensive computation. In this dissertation we have taken...
Show more"Supported by advanced sensing capabilities, increasing computational resources and the advances in Artificial Intelligence, smartphones have become our virtual companions in our daily life. An average modern smartphone is capable of handling a wide range of tasks including navigation, advanced image processing, speech processing, cross app data processing and etc. The key facet that is common in all of these applications is the data intensive computation. In this dissertation we have taken steps towards the realization of the vision that makes the smartphone truly a platform for data intensive computations by proposing frameworks, applications and algorithmic solutions. We followed a datadriven approach to the system design. To this end, several challenges must be addressed before smartphones can be used as a system platform for dataintensive applications. The major challenge addressed in this dissertation include high power consumption, high computation cost in advance machine learning algorithms, lack of realtime functionalities, lack of embedded programming support, heterogeneity in the apps, communication interfaces and lack of customized data processing libraries. The contribution of this dissertation can be summarized as follows. We present the design, implementation and evaluation of the ORBIT framework, which represents the first system that combines the design requirements of a machine learning system and sensing system together at the same time. We ported for the first time offtheshelf machine learning algorithms for realtime sensor data processing to smartphone devices. We highlighted how machine learning on smartphones comes with severe costs that need to be mitigated in order to make smartphones capable of realtime dataintensive processing. From application perspective we present SPOT. SPOT aims to address some of the challenges discovered in mobilebased smarthome systems. These challenges prevent us from achieving the promises of smarthomes due to heterogeneity in different aspects of smart devices and the underlining systems. We face the following major heterogeneities in building smarthomes:: (i) Diverse appliance control apps (ii) Communication interface, (iii) Programming abstraction. SPOT makes the heterogeneous characteristics of smart appliances transparent, and by that it minimizes the burden of home automation application developers and the efforts of users who would otherwise have to deal with appliancespecific apps and control interfaces. From algorithmic perspective we introduce two systems in the smartphonebased deep learning area: DeepCrowdLabel and DeepPartition. Deep neural models are both computationally and memory intensive, making them difficult to deploy on mobile applications with limited hardware resources. On the other hand, they are the most advanced machine learning algorithms suitable for realtime sensing applications used in the wild. DeepPartition is an optimizationbased partitioning metaalgorithm featuring a tiered architecture for smartphone and the backend cloud. DeepPartition provides a profilebased model partitioning allowing it to intelligently execute the Deep Learning algorithms among the tiers to minimize the smartphone power consumption by minimizing the deep models feedforward latency. DeepCrowdLabel is prototyped for semantically labeling user's location. It is a crowdassisted algorithm that uses crowdsourcing in both training and inference time. It builds deep convolutional neural models using crowdsensed images to detect the context (label) of indoor locations. It features domain adaptation and model extension via transfer learning to efficiently build deep models for image labeling. The work presented in this dissertation covers three major facets of datadriven and computeintensive smartphonebased systems: platforms, applications and algorithms; and helps to spurs new areas of research and opens up new directions in mobile computing research."Pages iiiii.
Show less
 Title
 Stochastic modeling of routing protocols for cognitive radio networks
 Creator
 Soltani, Soroor
 Date
 2013
 Collection
 Electronic Theses & Dissertations
 Description

Cognitive radios are expected torevolutionize wireless networking because of their ability tosense, manage and share the mobile available spectrum.Efficient utilization of the available spectrum could be significantly improved by incorporating different cognitive radio based networks. Challenges are involved in utilizing the cognitive radios in a network, most of which rise from the dynamic nature of available spectrum that is not present in traditional wireless networks. The set of available...
Show moreCognitive radios are expected torevolutionize wireless networking because of their ability tosense, manage and share the mobile available spectrum.Efficient utilization of the available spectrum could be significantly improved by incorporating different cognitive radio based networks. Challenges are involved in utilizing the cognitive radios in a network, most of which rise from the dynamic nature of available spectrum that is not present in traditional wireless networks. The set of available spectrum blocks(channels) changes randomly with the arrival and departure of the users licensed to a specific spectrum band. These users are known as primary users. If a band is used by aprimary user, the cognitive radio alters its transmission power level ormodulation scheme to change its transmission range and switches to another channel.In traditional wireless networks, a link is stable if it is less prone to interference. In cognitive radio networks, however, a link that is interference free might break due to the arrival of its primary user. Therefore, links' stability forms a stochastic process with OFF and ON states; ON, if the primary user is absent. Evidently, traditional network protocols fail in this environment. New sets of protocols are needed in each layer to cope with the stochastic dynamics of cognitive radio networks.In this dissertation we present a comprehensive stochastic framework and a decision theory based model for the problem of routing packets from a source to a destination in a cognitive radio network. We begin by introducing two probability distributions called ArgMax and ArgMin for probabilistic channel selection mechanisms, routing, and MAC protocols. The ArgMax probability distribution locates the most stable link from a set of available links. Conversely, ArgMin identifies the least stable link. ArgMax and ArgMin together provide valuable information on the diversity of the stability of available links in a spectrum band. Next, considering the stochastic arrival of primary users, we model the transition of packets from one hop to the other by a SemiMarkov process and develop a Primary Spread Aware Routing Protocol (PSARP) that learns the dynamics of the environment and adapts its routing decision accordingly. Further, we use a decision theory framework. A utility function is designed to capture the effect of spectrum measurement, fluctuation of bandwidth availability and path quality. A node cognitively decides its best candidate among its neighbors by utilizing a decision tree. Each branch of the tree is quantified by the utility function and a posterior probability distribution, constructed using ArgMax probability distribution, which predicts the suitability of available neighbors. In DTCR (Decision Tree Cognitive Routing), nodes learn their operational environment and adapt their decision making accordingly. We extend the Decision tree modeling to translate video routing in a dynamic cognitive radio network into a decision theory problem. Then terminal analysis backward induction is used to produce our routing scheme that improves the peak signaltonoise ratio of the received video.We show through this dissertation that by acknowledging the stochastic property of the cognitive radio networks' environment and constructing strategies using the statistical and mathematical tools that deal with such uncertainties, the utilization of these networks will greatly improve.
Show less
 Title
 TENSOR LEARNING WITH STRUCTURE, GEOMETRY AND MULTIMODALITY
 Creator
 Sofuoglu, Seyyid Emre
 Date
 2022
 Collection
 Electronic Theses & Dissertations
 Description

With the advances in sensing and data acquisition technology, it is now possible to collect datafrom different modalities and sources simultaneously. Most of these data are multidimensional in nature and can be represented by multiway arrays known as tensors. For instance, a color image is a thirdorder tensor defined by two indices for spatial variables and one index for color mode. Some other examples include color video, medical imaging such as EEG and fMRI, spatiotemporal data...
Show moreWith the advances in sensing and data acquisition technology, it is now possible to collect datafrom different modalities and sources simultaneously. Most of these data are multidimensional in nature and can be represented by multiway arrays known as tensors. For instance, a color image is a thirdorder tensor defined by two indices for spatial variables and one index for color mode. Some other examples include color video, medical imaging such as EEG and fMRI, spatiotemporal data encountered in urban traffic monitoring, etc.In the past two decades, tensors have become ubiquitous in signal processing, statistics andcomputer science. Traditional unsupervised and supervised learning methods developed for one dimensional signals do not translate well to higher order data structures as they get computationally prohibitive with increasing dimensionalities. Vectorizing high dimensional inputs creates problems in nearly all machine learning tasks due to exponentially increasing dimensionality, distortion of data structure and the difficulty of obtaining sufficiently large training sample size.In this thesis, we develop tensorbased approaches to various machine learning tasks. Existingtensor based unsupervised and supervised learning algorithms extend many wellknown algorithms, e.g. 2D component analysis, support vector machines and linear discriminant analysis, with better performance and lower computational and memory costs. Most of these methods rely on Tucker decomposition which has exponential storage complexity requirements; CANDECOMPPARAFAC (CP) based methods which might not have a solution; or Tensor Train (TT) based solutions which suffer from exponentially increasing ranks. Many tensor based methods have quadratic (w.r.t the size of data), or higher computational complexity, and similarly, high memory complexity. Moreover, existing tensor based methods are not always designed with the particular structure of the data in mind. Many of the existing methods use purely algebraic measures as their objective which might not capture the local relations within data. Thus, there is a necessity to develop new models with better computational and memory efficiency, with the particular structure of the data and problem in mind. Finally, as tensors represent the data with more faithfulness to the original structure compared to the vectorization, they also allow coupling of heterogeneous data sources where the underlying physical relationship is known. Still, most of the current work on coupled tensor decompositions does not explore supervised problems.In order to address the issues around computational and storage complexity of tensor basedmachine learning, in Chapter 2, we propose a new tensor train decomposition structure, which is a hybrid between Tucker and Tensor Train decompositions. The proposed structure is used to imple ment Tensor Train based supervised and unsupervised learning frameworks: linear discriminant analysis (LDA) and graph regularized subspace learning. The algorithm is designed to solve ex tremal eigenvalueeigenvector pair computation problems, which can be generalized to many other methods. The supervised framework, Tensor Train Discriminant Analysis (TTDA), is evaluated in a classification task with varying storage complexities with respect to classification accuracy and training time on four different datasets. The unsupervised approach, Graph Regularized TT, is evaluated on a clustering task with respect to clustering quality and training time on various storage complexities. Both frameworks are compared to discriminant analysis algorithms with similar objectives based on Tucker and TT decompositions.In Chapter 3, we present an unsupervised anomaly detection algorithm for spatiotemporaltensor data. The algorithm models the anomaly detection problem as a lowrank plus sparse tensor decomposition problem, where the normal activity is assumed to be lowrank and the anomalies are assumed to be sparse and temporally continuous. We present an extension of this algorithm, where we utilize a graph regularization term in our objective function to preserve the underlying geometry of the original data. Finally, we propose a computationally efficient implementation of this framework by approximating the nuclear norm using graph total variation minimization. The proposed approach is evaluated for both simulated data with varying levels of anomaly strength, length and number of missing entries in the observed tensor as well as urban traffic data. In Chapter 4, we propose a geometric tensor learning framework using product graph structures for tensor completion problem. Instead of purely algebraic measures such as rank, we use graph smoothness constraints that utilize geometric or topological relations within data. We prove the equivalence of a Cartesian graph structure to TTbased graph structure under some conditions. We show empirically, that introducing such relaxations due to the conditions do not deteriorate the recovery performance. We also outline a fully geometric learning method on product graphs for data completion.In Chapter 5, we introduce a supervised learning method for heterogeneous data sources suchas simultaneous EEG and fMRI. The proposed twostage method first extracts features taking the coupling across modalities into account and then introduces kernelized support tensor machines for classification. We illustrate the advantages of the proposed method on simulated and real classification tasks with small number of training data with high dimensionality.
Show less
 Title
 Theory, synthesis and implementation of currentmode CMOS piecewiselinear circuits using margin propagation
 Creator
 Gu, Ming
 Date
 2012
 Collection
 Electronic Theses & Dissertations
 Description

Achieving high energyefficiency is a key requirement for many emerging smart sensors and portable computing systems. While digital signal processing (DSP) has been the defacto technique for implementing ultralow power systems, analog signal processing (ASP) provides an attractive and alternate approach that can not only achieve high energy efficiency but also high computational density. Conventional ASP techniques are based on a topdown design approach, where proven mathematical...
Show moreAchieving high energyefficiency is a key requirement for many emerging smart sensors and portable computing systems. While digital signal processing (DSP) has been the defacto technique for implementing ultralow power systems, analog signal processing (ASP) provides an attractive and alternate approach that can not only achieve high energy efficiency but also high computational density. Conventional ASP techniques are based on a topdown design approach, where proven mathematical principles and related algorithms are mapped and emulated using computational primitives inherent in the device physics. An example being the translinear principle, which is the stateoftheart ASP technique, that uses the exponential currenttovoltage characteristics for designing ultralowpower analog processors. However, elegant formulations could result from a bottomup approach where device and bias independent computational primitives (e.g. current and charge conservation principles) are used for designing "approximate" analog signal processors. The hypothesis of this proposal is that many signal processing algorithms exhibit an inherent calibration ability due to which their performance remains unaffected by the use of "approximate" analog computing techniques. In this research, we investigate the theory, synthesis and implementation of high performance analog processors using a novel piecewiselinear (PWL) approximation algorithm called margin propagation (MP). MP principle utilizes only basic conservation laws of physical quantities (current, charge, mass, energy) for computing and therefore is scalable across devices (silicon, MEMS, microfluidics). However, there are additional advantages of MPbased processors when implemented using CMOS currentmode circuits, which includes: 1) the operation of the MP processor requires only addition, subtraction and threshold operations and hence is independent of transistor biasing (weak, moderate and strong inversion) and robust to variations in environmental conditions (e.g. temperature); and 2) improved dynamic range and faster convergence as compared to the translinear implementations. We verify our hypothesis using two analog signal processing applications: (a) design of highperformance analog lowdensity parity check (LDPC) decoders for applications in sensor networks; and (b) design of ultralowpower analog support vector machines (SVM) for smart sensors. Our results demonstrate that an algorithmic framework for designing margin propagation (MP) based LDPC decoders can be used to tradeoff its BER performance with its energy efficiency, making the design attractive for applications with adaptive energyBER constraints. We have verified this tradeoff using an analog currentmode implementation of an MPbased (32,8) LDPC decoder. Measured results from prototypes fabricated in a 0.5 μm CMOS process show that the BER performance of the MPbased decoder outperforms a benchmark stateoftheart minsum decoder at SNR levels greater than 3.5 dB and can achieve energy efficiencies greater than 100pJ/bit at a throughput of 12.8 Mbps. In the second part of this study, MP principle is used for designing an energyscalable support vector machine (SVM) whose power and speed requirements can be configured dynamically without any degradation in performance. We have verified the energyscaling property using a currentmode implementation of an SVM operating with 8 dimensional feature vectors and 18 support vectors. The prototype fabricated in a 0.5μm CMOS process has integrated an array of floating gate transistors that serve as storage for up to 2052 SVM parameters. The SVM prototype also integrates novel circuits that have been designed for interfacing with an external digital processor. This includes a novel currentinput currentoutput logarithmic amplifier circuit that can achieve a dynamic range of 120dB while consuming nanowatts of power. Another novel circuit includes a varactor based temperature compensated floatinggate memory that demonstrates a superior programming range than other temperature compensated floatinggate memories.
Show less
 Title
 Unconstrained 3D face reconstruction from photo collections
 Creator
 Roth, Joseph (Software engineer)
 Date
 2016
 Collection
 Electronic Theses & Dissertations
 Description

This thesis presents a novel approach for 3D face reconstruction from unconstrained photo collections. An unconstrained photo collection is a set of face images captured under an unknown and diverse variation of poses, expressions, and illuminations. The output of the proposed algorithm is a true 3D face surface model represented as a watertight triangulated surface with albedo data colloquially referred to as texture information. Reconstructing a 3D understanding of a face based on 2D input...
Show moreThis thesis presents a novel approach for 3D face reconstruction from unconstrained photo collections. An unconstrained photo collection is a set of face images captured under an unknown and diverse variation of poses, expressions, and illuminations. The output of the proposed algorithm is a true 3D face surface model represented as a watertight triangulated surface with albedo data colloquially referred to as texture information. Reconstructing a 3D understanding of a face based on 2D input is a longstanding computer vision problem. Traditional photometric stereobased reconstruction techniques work on aligned 2D images and produce a 2.5D depth map reconstruction. We extend face reconstruction to work with a true 3D model, allowing us to enjoy the benefits of using images from all poses, up to and including profiles. To use a 3D model, we propose a novel normal fieldbased Laplace editing technique which allows us to deform a triangulated mesh to match the observed surface normals. Unlike prior work that require large photo collections, we formulate an approach to adapt to photo collections with few images of potentially poor quality. We achieve this through incorporating prior knowledge about face shape by fitting a 3D Morphable Model to form a personalized template before using a novel analysisbysynthesis photometric stereo formulation to complete the fine face details. A structural similaritybased quality measure allows evaluation in the absence of ground truth 3D scans. Superior largescale experimental results are reported on Internet, synthetic, and personal photo collections.
Show less