CU Theses
Pages
-
- Classification of Functional and Non-Functional Arm Movement Using Deep Learning and Wearable Camera Video Data
- Stroke is the loss of blood flow to part of the brain. It is one of the most common causes of death, global prevalence of stroke in 2019 was 101.5 million people. When a stroke happens, our brain cells do not receive enough oxygen that leads to the death of brain cells in some areas of the brain. Every year, it is estimated more than 800,000 people in the United States have strokes and approximately two-thirds of them survive but require rehabilitation. Commonly, movement control, cognitive ability, vision, speech, muscle control is affected. Upper Extremity (UE) rehabilitation is widely used to regain mobility and functionality of arm movements. The upper extremity begins at the shoulder joint, consists of three sections, the upper arm, forearm, and hand. The goal of UE rehabilitation is to improve the use of the upper extremity for post-stroke patients in their daily life. Although UE rehabilitation is commonly used in many clinicals and facilities, the results of the treatment in daily life are hard to evaluate. There are many different approaches to evaluate their effectiveness, however, most of them are based on wearable wrist-worn accelerometers when performing specific motor tasks associated with functional use, such as measuring grasp, grip, pinch, and gross movement. Bochniewicz, et al. successfully used machine learning to classify sensor data with an average of 94.80% in controls and 88.38% in stroke subjects in intra-subject test trials, and 91.53% for controls and 70.18% in stroke subjects in inter-subject test trials. However, these outcomes relate to the ability to perform various tasks in the clinic, not everyday life. The actual use of impaired limbs in limbs in everyday life is assessed by everyday life is currently assessed by a survey. It is realized that improving ability to perform a task does not mean increasing the use of impaired limbs when performing that task spontaneously in daily life. This project will investigate the use of a body worn camera to keep track of arm and hand movements in daily spontaneous activities. In this work, videos were collected when the participants performing a predefined task “Folding clothes” while wearing a GoPro camera mounted on the center of their chest. Videos collected from the camera are pre-processed by a command line toolbox called “FFMPEG” and analyzed by a deep learning algorithm to classify functional and non-functional movement of the upper limb. The algorithm is a combination of Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTMs) along with Time Distributed Layers. The results show that the deep learning architecture also learned the background features besides learning arms/hands features. When tested with data that has similar background to the training dataset, the model shows the ability to identify functional and non-functional movement, even with healthy participants who were not included in the training set. , Biomedical engineering, Biomedical Engineering, Degree Awarded: M.S.--Biomedical Engineering. The Catholic University of America
-
- Classifying Cognitive-Aesthetic Experience Through Machine Learning and EEG Data
- Humans have been designing and building sacred and secular structures for thousands of years. Designers in every culture and age have aspired to appeal to the emotional and spiritual senses of those who visit them. Judging by the high esteem in which we hold many cathedrals, palaces, towers, arenas, and homes, and by how well-known many of their designers remain after hundreds of years, they have clearly been effective. Yet, until now, we have had no way to examine what is happening at a physiological level when someone enters a building. With the recent advent of wearable technology such as mobile EEG devices and biometric wrist monitors that record the electrical activity of the brain and measure physiological data, we have new tools to inspect the internal states of building visitors and to determine whether a measurable, physiological response accompanies their experiences. This thesis applies machine learning to the data acquired from these devices to detect a difference in the effect that sacred and secular architecture has on believers. Data was collected from 32 subjects who were familiar with two buildings, the Basilica of the National Shrine of the Immaculate Conception (sacred structure) and Union Station (secular structure), and who reported being committed to the Catholic faith. Two machine learning algorithms, the random forest and the support vector machine, were used to construct both an intrasubject and an intersubject classifier that predicts whether a given second of EEG data was recorded in one of two buildings. The best intrasubject model is just under 80% accurate on average, and up to 90% accurate on some subjects, indicating a typical and measurable response in the subjects. However, individual variability in EEG measurements make an intersubject model difficult with a sample this small. Additional size or additional features may be required to develop a classifier that can operate successfully across a wide range of subjects., Computer science, Architecture, Electrical Engineering and Computer Science, Degree Awarded: M.S.C.S. Electrical Engineering and Computer Science. The Catholic University of America
-
- Clerical Education From the Time of Charlemagne to the Rise of the Universities
- by Thomas Vincent Cassidy., Typescript., Thesis--(S.T.L)--Catholic University of America, 1924., Bibliography: leaves 78-81.
-
- Coming of the Norsemen
- by P. W. Theibeau., Typescript., Thesis (M.A.--Philosophy)--Catholic University of America, 1918., Bibliography: leaves [iii-iv].
-
- Commentary on Gerson's Treatise:-"Leading the Little Ones to Christ."
- by Anthony Bernard Kruegler., Typescript., Thesis--(S.T.L)--Catholic University of America, 1920., Bibliography: leaves 79-81.