Browsing by Subject "Tracking"

Now showing 1 - 14 of 14

Acoustic motion tracking and its application
(2017-05) Zheng, Huihuang; Qiu, Lili, Ph. D.
Video games, Virtual Reality (VR), Augmented Reality (AR), and Smart appliances (e.g., smart TVs) all call for a new way for users to interact and control them. This thesis explores high precision acoustic motion tracking system which aims to replace traditional control devices such as mouse and let the user play games, interact with VR/AR headsets, and control smart appliances. We develop a lightweight system which can achieve mm-level tracking accuracy using inaudible sounds. At the heart of our system lies a distributed Frequency Modulated Continuous Waveform (FMCW) which is able to accurately estimate the absolute distance between a receiver and a transmitter that are separate and unsynchronized. We further develop an optimization framework to combine FMCW estimation with Doppler shifts and Inertial Measurement Unit (IMU) measurements to enhance the accuracy, and efficient algorithm to solve the optimization problem. We develop several interesting applications on top of our motion tracking technology, including audio ruler, drawing in the air, and playing motion-controlled games
Essays on economics of education
(2017-05) He, Ziwei; Linden, Leigh L., 1975-; Murphy, Richard J; Antonovics, Kate; Geruso, Michael L
This dissertation investigates the relationship between various input for education (such as class structure and academic environment) and academic outcomes (including majors and test scores). It also looks at the impact of education on individuals’ non-academic outcomes. This dissertation consists of three essays. The first chapter examines the impact of assigning students into different tracks on students’ academic performance and subject specialization in China. I make use of regression discontinuity design and find that track assignment significantly affects choice of majors and test scores in high school. For students around the tracking threshold, being assigned to a high track reduces the probability of choosing the science major by 7 percent for boys and 21 percent for girls. The second chapter examines the impact of international peers on domestic students STEM degree in U.S.. I use historical enrollment patterns as an instrumental variable to predict current enrollment of international students and find that the composition and ability of international peers significantly affect the likelihood of graduating with a STEM degree for female and minority domestic students. The third chapter explores the casual relationship from education to religious beliefs in China. I exploit the change in compulsory school law in China in 1986 and find that one additional year of schooling reduce the probability of being religious by 8 percent.
iTrak : a social mobile diary and web blogging utility for travelers
(2013-05) Dao, Tung Thanh, active 2013; Aziz, Adnan
iTrak is a combined mobile and web application that takes advantage of the GPS to allow travelers to share their experience while travelling. The application gathers GPS data and broadcasts it via a web interface or social networks such as Facebook to update user’s status during a trip. iTrak is also equipped with other features such as writing notes or recording video journals to offer a rich experience and provide an interactive diary, along with a real-time tracking ability, for travelers.
Learning to write in (networked) public: children and the delivery of writing online
(2014-12) Roach, Audra Katherine; Bomer, Randy; Hoffman, Jim; Maloch, Beth; Schallert, Diane; Hodgson, Justin
This investigation explored how three children (together with parents) developed networked publics that were diverse, well-connected, and powerful in the world. It was framed in response to calls in the field to better understand the new literacies young writers develop online and outside of school, and to increase literacy educators’ attention to the role of public audiences in writing and how writing is circulated. Performative case study methodology, ethnographic methods, and digital methods were employed to track and describe the online networks of three children (ages 11-13). These focal children were actively involved with their parents in social media, and had developed widespread networks with shared interests in children’s books and book reviews (Case 1), baseball (Case 2), and helping the homeless (Case 3). The children’s online networks were conceptualized as networked publics, drawing on Warner’s (2002) notion of publics as ongoing discursive relations among strangers, and on Actor-Network Theory’s notion of networks as assemblages of diverse interests that mobilize toward a common goal (Callon, 1986) and that develop stability in relation to ongoing circulations of texts (Latour, 1986; Spinuzzi, 2008). Research questions were framed broadly around the rhetorical canon of delivery [now digital delivery (Porter, 2009)], and were concerned with how writers distributed texts online, how those texts circulated, how the networked publics become more stable and powerful, and what instabilities children and parents had to negotiate in order to accomplish all of this. Data sources included interviews with 15 children and 28 adults, and fieldnotes observations of approximately 1,700 screen-captured webpages and other online artifacts. Findings showed that the young writers and their parents initiated and sustained networked publics through distribution practices that were oriented toward building trust; their texts displayed: interest, appreciation, reliability, service, credibility, and responsiveness. Both grassroots and commercial entities circulated texts in these networks, as they were sources of the ongoing renewal these different groups all needed in order to thrive. Sources of instability included conflicts over standards of writing quality, matters of profit, and the constancy of the demand to generate new interest and writing online. Children and their parents responded to these instabilities by welcoming and negotiating heterogeneous perspectives and partnerships. Implications of the study call for further research and teaching about the art of networked public discourse and digital delivery.
Listwise frameworks for ranking and rank aggregation
(2017-12) Pareek, Harsh Harivansh; Ghosh, Joydeep; Ravikumar, Pradeep; Mooney, Raymond; Price, Eric; Tewari, Ambuj
The goal in Learning to Rank (LETOR) is to learn to order a novel set of items, given training data comprising sets of items and their orderings. There are three important problems in LETOR: capturing individual item relevance, using item interaction information to diversify rankings and aggregating ranked lists. The first has been the subject of considerable recent work; this dissertation contributes to the latter two. LETOR is a key component in Information Retrieval (IR) systems. IR approaches first select relevant items using naïve but scalable strategies, then apply sophisticated LETOR algorithms to re-rank them before user presentation. Most existing LETOR approaches assign scores to each item independently of the other retrieved items and simply rank them according to these scores. These methods are described as listwise in the literature but we finesse these as pointwise ranking functions being used with listwise losses. However sophisticated such an approach may be, it can never hope to diversify rankings if it does not take interactions between items into account at test time. Our contributions are as follows: we introduce listwise ranking functions (LRFs) and a corresponding representation theorem. Then, we discuss a listwise boosting procedure for combining LRFs, and analyze boostability, weak-learnability in this context. Motivated by empirical concerns, we introduce and address the problem of learning a "best" surrogate function from among those consistent with an underlying metric, i.e. one most suited to the data distribution. We study fundamental limitations of unsupervised rank aggregation procedures by connecting them to impossibility results in social choice theory, namely Arrow's theorem, and consider relaxed variants of social choice axioms. Towards the aim of modeling ranking and diversity in a flexible manner, we study how to handle ranking factors in Graphical Models and test these ideas empirically in an oceanographic float-tracking problem setting which requires ranking as a key component. The ideas presented in this dissertation are supported by experiments primarily on MSLETOR datasets; we conclude with thoughts on the limitations of these datasets for learning LRFs and propose directions for the future.
Long-term tracking of neuronal clusters with ultraflexible oversampling electrode arrays
(2020-02-05) Zhu, Hanlin; Xie, Chong, Ph. D.
Biological being has displayed abilities to adapt to the ever-changing environment through neural and behavioral modifications and stably maintain such adaptations. Understanding the neural substrates that underlie such capacity is a critical question in neuroscience. Answering this question requires reliable, long-term, and large-scale tracking of the dynamical brain circuitry at the resolution of its basic unit, single neurons. Intracortical electrical recording remains the only option that simultaneously offers temporally resolved, depth independent, long-term tracking of local neuronal clusters. This thesis highlights data processing strategy and the performance of 4 different electrode designs whose properties vary substantially along two axes 1. The spatial resolution of the sensor, subsampling versus oversampling 2. Rigidity of the implanted device, rigid versus flexible. In the second part of this thesis, our work on long-term tracking of neuronal clusters with ultraflexible oversampling electrode array is summarized
Pathway through math : educator perspectives on middle school math acceleration
(2021-05-06) Heaton, Amy Joann; Riegle-Crumb, Catherine
This thesis reports a study of middle-school mathematics teachers’ attitudes about teaching Geometry in middle school, along with the difference between the factors they think should be used in placing students in the advanced (Geometry) track and what factors are actually considered. Mathematics is a subject which sees significant racialized tracking due to the sequential nature of its course progression coupled with inequitable data measurements and placement methods. While the Common Core State Standards (CCSS) present a standard course progression that does not include Algebra 1 at the middle school level, many school districts continue to include it, and in some cases, Geometry, as options for higher-performing students. In this study, three middle school teachers from two school districts that offered different middle school mathematics course progressions were surveyed, and the responses were then analyzed and coded. Though these teachers had idealized notions of placement tests being the best measure for a student’s mathematical readiness, additional considerations such as equity concerns and parental disagreement contributed to the actual placement of students into advanced course pathways. This thesis discusses implications for equity in middle school math.
Precision pinch isometric force, force variability, accuracy, and task time among the fourth through eighth decades of life
(2011-05) Herring-Marler, Trenah Lannette; Abraham, Lawrence D.; Spirduso, Waneen Wyrick; Eakin, Richard T; Griffin, Lisa; Hunter, Diana
This dissertation encompassed three studies involving precision pinch strength and 5% submaximal fine-motor control. One hundred participants (30-79 years old) were divided into 10-year categories, with 10 males and 10 females in each decade. A Manual Force Quantification System containing a platform and force-transducer apparatus, along with a computer and visual monitor, was used. Each subject performed four tasks -- maximal voluntary isometric contraction (MVIC), force-matching, tracing, and tracking -- by applying force on the transducers with the thumb and index finger while attempting to produce a desired force level or task displayed on the computer monitor. The first study measured MVIC, accuracy (rRMSE, Root Mean Square Relative Error), and force variability (Coefficient of Variation, CV) during a 5% MVIC force-matching task. The second study measured accuracy (rRMSE), task time, and group variability during a 5% MVIC tracing task. The third study measured accuracy and group variability during a 5% MVIC tracking task. Tracing and tracking were each divided into six Segments (S1-S6), three of which (S1-S3) required the increasing application of force from 50g up to 5% MVIC and the remaining three (S4-S6) requiring a release of force from MVIC down to 1% MVIC. The force-matching and force-tracking task times were scaled to each participant's MVIC, while the tracing task was performed at the participant's self-selected speed. The participants were encouraged to be accurate but also to trace the target line as quickly as possible. Declines in precision pinch strength and force control began to occur in the 70s for easier force-control tasks and in their 60s for more advanced force-tracking tasks. Men were stronger than women at all age levels. Participants in their 30s were the fastest; those in their 40s, 50s, and 60s slowed down to be accurate; and those in their 70s moved faster but were the least accurate. Three segmental factors affected error and time: low force level, releasing as opposed to applying force, and location along the target line with respect to reversal or ending points. Finally, variables for females were more heterogeneous at earlier decades than for men, and the older the age group was, the greater the variable heterogeneity was.
Predicting crowd trajectories using deep Graph Convolution Neural Networks
(2019-12) Mohamed, Abduallah Adel Omar; Claudel, Christian
Better machine understanding of pedestrian behaviors enables faster progress in modeling interactions between agents such as autonomous vehicles and humans. Pedestrian trajectories are not only influenced by the pedestrian itself but also by interaction with surrounding objects. Previous methods modeled these interactions by using a variety of aggregation methods that integrate different learned pedestrians states. We propose the Social Spatio-Temporal Graph Convolutional Neural Network (Social-STGCNN), which substitutes the need of aggregation methods by modeling the interactions as a graph. Our results show an improvement over the state of art by 20% on the Final Displacement Error (FDE) and an improvement on the Average Displacement Error (ADE) with 8.5 times less parameters and up to 48 times faster inference speed than previously reported methods. In addition, our model is data efficient, and exceeds previous state of the art on the ADE metric with only 20% of the training data. We propose a kernel function to embed the social interactions between pedestrians within the adjacency matrix. Through qualitative analysis, we show that our model inherited social behaviors that can be expected between pedestrians trajectories. Code will be released for the reproduction of the results.
Probabilistic multi-object tracking for autonomous vehicles
(2019-10-10) Motro, Michael; Ghosh, Joydeep; Bhat, Chandra; Heath, Robert; Shakkottai, Sanjay; Thomaz, Edison
Interactive robots such as self-driving cars require accurate hardware and methods to locate relevant objects such as other traffic participants. They also must predict other participants' actions or understand their role in the environment. Given imperfect information about present objects at each time, a multi-object tracker maintains an estimate of all present relevant objects and infers motion or other information that can be deduced from viewing an object over time. Trackers are often built around a probabilistic model that includes known characteristics of object motion and sensor behavior. This thesis discusses several details for designing a probabilistic multi-object tracker for vehicular environments, as well as ways to utilize probabilistic tracked estimates for autonomous vehicle applications. The increasingly complex environments perceived by robots have demanded new paradigms of perception. In particular, camera and laser-based perception of urban settings is solved using learned algorithms that directly transform raw data into object estimates. We present a probabilistic model of modern object detectors that can be integrated with standard trackers. The primary effects that are modelled are line-of-sight limitations to sensor detection, and correlation in algorithmic detection errors over time. Each of these modifications are shown to improve performance on a public benchmark for vehicle tracking, without fundamental modifications to the tracking algorithm. Accurate tracking can require intensive computation on its own. We examine the implementation of multiple hypothesis tracking, a high-performance probabilistic tracker, and improve the computational efficiency of its data association algorithm in several ways. The modified algorithm is tested on vehicular tracking data as well as simulated large-scale and multisensor problems. The improved speed of the algorithm allows for more hypotheses to be propagated at a given speed, which in turn improves tracking performance. In addition to improving the current estimate of the environment, tracking enables prediction of the future environment by determining object motion and history. The uncertainty of these estimates can be quantified by a probabilistic tracker and should be considered when making predictions or deciding actions. However, probabilistic estimates are difficult to translate into interpretable and actionable concepts, such detection of impending collisions between objects. We disambiguate the error rate in collision detection into inevitable errors from uncertain object estimation and further errors incurred by fast approximate calculation of the probability of collision from these estimates. Various methods for collision detection from uncertain data are compared and tested on vehicle simulations. Automated overtaking assistants are studied as a specific application of collision detection. These assistants alert drivers in advance that entering the opposite lane to pass a slower vehicle will be unsafe. We characterize the expected design of these systems, including sensor or communication accuracy and limitations as well as driver variability and uncertainty in future motion. Overtaking assistant simulations demonstrate that the assistant can fulfill its purpose at expected levels of tracking and prediction uncertainty, provided that the chosen sensor or communicating device has a sufficient operating distance
Single station Doppler tracking for satellite orbit prediction and propagation
(2015-05) Dykstra, Matthew C.; Fowler, Wallace T.; Lightsey, E. Glenn
Presently, there are two main methods of launching a cube satellite into Earth orbit. The first method is to purchase a secondary payload slot on a major launch vehicle. For the second method, the satellite must first be transported via a major launch vehicle to the International Space Station. From there, the satellite is loaded into one of two deployment mechanisms, and deployed at a specified time. In each case, the satellite's initial orbit is not accurately known. For ground operators this poses a problem of position uncertainty. In order to solve this problem, a satellite tracking algorithm was developed to use an initial two-line element set for coarse orbit prediction, followed by Doppler measurements for continuous processing and updating. The system was tested using simulated data. The analysis showed that this low-cost, scalable system will satisfy the tracking requirements of many cube satellite missions, including current missions at the University of Texas.
Texas's House Bill 5 as modern tracking structure : social stratification reified?
(2018-05) Arrington, Katherine Leigh; Reyes, Pedro, 1954-; Treisman, Uri.; Holme, Jennifer J; Green, Terrance L
In 2013, Texas policymakers passed House Bill 5 (HB 5), which changed high school graduation requirements to a multitiered set of plans called the Foundation High School Program (FHSP). This hierarchical set of graduation plans groups students based on a chosen career endorsement and offers different content instruction based on their choices, mirroring tracking structures that categorize students into groups and then provide those groups with dissimilar instructional experiences. This project investigated whether HB 5 is achieving the hope of the bill’s authors—to increase student engagement through allowing students to choose programs tailored to their career aspirations—or if the policy functionally operates as tracking. This study used a quantitative analysis of the data available through the Texas Education Agency (TEA) to look for descriptive patterns in the offerings and outcomes for students using the predictor variables of the type of or urbanicity of the district and the racial and socioeconomic composition of each district. Generalized linear models and generalized linear multilevel models indicate the extent to which relationships between both the HB 5 graduation plan offerings in each district and outcomes for students enrolling and graduating under the HB 5 plans and the district’s characteristics. This study found significant differences in the endorsements offered by districts based on urbanicity of the district, specifically differences between rural districts and the rest of the state. The study found differences in who was enrolled in FHSP while enrollment was considered optional, with significant differences by year and for those students enrolled in rural districts as well as specifically for students in districts with higher proportion of African American/Black and Hispanic/Latino students. There are significant differences in graduates under FHSP who earned the distinguished level of achievement based on these predictors and specific differences in the odds of students in suburban districts with higher proportions of African American/Black students graduating under FHSP and earning the distinguished level of achievement. Implications indicate that FHSP operates as a means to uphold the system of student tracking
The mountain : a journey from self to summit
(2018-05-03) Haas, Jonathan Edward; Lynn, Kirk; Ortel, Sven
As a reflective theatrical practitioner, I must engage with my design process and perpetually investigate the methods I employ in my craft as a multi-disciplinary designer. Concurrently, recent trends in technology have challenged traditional models of theater and numerous artists are inventing new methods of storytelling. Just as the smart phone has put the user in the center of their digital sphere, installation art is putting the participant in control of their story and their experience. This thesis was developed to investigate my creative process and how I might integrate my own personal experiences in alpine mountaineering into an installation that sought to empower participants. Further, the installation explored interactive design techniques and technologies including positional motion tracking. This thesis documents the development and implementation of an immersive art installation, The Mountain, which incorporated concepts of video games and theater within a mythological narrative framework directly informed by Joseph Campbell’s Hero’s Journey. This mixed methods study analyzes the experiences of a collaborative design team and audience participants as a way to assess the impact of the project. Findings include a deepened understanding of my process and aspirations as a generative artist alongside an increased experience with integrated technical systems. This document concludes with a conceptual outline for future installation work.
Towards unified object recognition in the wild
(2022-05-02) Zhou, Xingyi, 1994-; Krähenbühl, Philipp; Mooney, Raymond J.; Zhu, Yuke; Ramanan, Deva
Large-scale well-curated datasets are the fuel of computer vision. However, most datasets only focus on one single domain with a specific task and a fixed label set. Computer vision models trained on a single dataset only apply to a subset of applications in the real world. The goal of my research is to remove the artificial barriers of datasets and make object recognition generalize in the wild. There should be one single computer vision model, not a zoo of dataset-specific models. The model should be trained on a diverse set of datasets and should be able to recognize objects from different data sources in all domains. Towards this goal, my thesis focuses on three aspects: a point-based object representation that unified multiple vision tasks, a unified framework that detects and tracks objects through time, and a unified vocabulary between detection and classification annotations. First, we propose to represent an object using the simplest-possible representation --- a point. All object properties, like its size, pose, depth, and velocities, are attributes of the points and are inferred from the point features. We developed a point-based object detector, CenterNet, using standard keypoint detection techniques. We extend this point-based detector to many vision tasks by just adding task-specific regression outputs. The point-based representation achieves the state-of-the-art-level performance and runs fast with a unified framework. Second, we show the point-based representation also simplifies linking objects through time. We extend our point-based detector into a local tracker by regression the inter-frame motion of each object. The resulting point-based tracker is efficient, accurate, robust, and unified under different domains, tasks, and framerates. Going further, we develop a tracker that associates and classifies objects from the whole video clip. The global association uses a transformer that looks at all objects in a long temporal window, and directly produces trajectories. Finally, we study how to extend the vocabulary of our recognition system. We explored two directions: 1) merging multiple object detection datasets in different vocabularies and domains with an automatic label-space unification algorithm; 2) introducing additional classification annotations with a much large vocabulary, i.e., twenty-thousand classes. The resulting unified detector has a broad vocabulary, is more robust to changes in the visual domain, and generalizes readily to new unseen environments and taxonomies.