Defense Notices

Information used to assist biomedical and clinical research has largely comprised of data available in published sources like scientific papers and journals, or in clinical sources like patient health records, lab reports and discharge summaries. Information from such sources, though extensive and organized, is often not readily available due to its proprietary and/or privacy-sensitive nature. Collecting such information through clinical studies is expensive and the information is often limited to the diversity of the people who are involved in the study. With the growth of online social networks, more and more people openly share their health experiences with other similar patients through online healthcare forums. The data from these forum messages can act as an alternate source that provides for unrestricted, high volume, highly diverse and up-to-date information needed for assisting and guiding biomedical and pharmaceutical research. However, this data is often unstructured, noisy and scattered, making it unsuitable for use in its current form. This dissertation presents an Information Retrieval and Knowledge Discovery Framework that is capable of collecting data from online healthcare forums, extracting useful information and storing it in a structured form that facilitates knowledge discovery. A Healthcare Forum Mining Ontology developed as a part of this work is used to organize and capture the semantic relationships between patient related data like age, gender, ethnicity and habits, along with health related data like drugs, side-effects, diseases and symptoms which are extracted from the forum messages. The utility of this framework is demonstrated with the help of two applications: an Adverse Drug Reaction discovery tool that is able to assist pharmacovigilance by extracting adverse effects of drugs from forum messages and an ontology-based visualization tool that can be used for exploring and analyzing associations between patient and health related data extracted from forum messages. 

The project presents the development of modified differential evolution algorithm based on harmony search algorithm for linear aperiodic array synthesis. The modified algorithm has the combine capability from the classical DE as well as harmony search algorithm. This differential evolution algorithm method optimizes a problem by iteratively trying to improve a solution with regards to given measure of quality. The objective is to optimize the linear aperiodic arrays with a minimum peak side lobe level (PSSL). The algorithm follows the steps of initializing the model parameters and generate corresponding base vectors followed by selection of two spacing vectors from the base vectors. Perform mutation and crossover in order to generate a new spacing vector. By calculation of PSSL along with execution of selection operation in DE, we update the vector base. Finally we adjust the parameters to meet the criteria, otherwise the iteration starts all over from the selection of two spacing vectors randomly. Numerical results shows that the HSDEA gives us a better PSSL performance. Comparison of PSSL using HSDEA and other differential evolution algorithm are performed which proves that the algorithm in study produces better PSSL performance with less number of evaluations.

Event Studies research focuses on the statistical impact that an event has on a traded company. In Finance, a financial press-release announcing company earnings is an example of an event. Unlike earnings announcements, media events may arise unexpectedly. By using the framework of an Event Study, this proposal will explore unexpected events in modern media -- particularly Twitter. Measuring statistical impact is not the central goal. Instead, listed here are the selected implementation objectives. Utilizing natural language processing, identify events on Twitter that influence stock prices of firms. Create text and time-series models, by applying machine learning techniques, to classify events. Develop quantitative trading strategies by associating prediction outputs as trading signals. The implementation objectives combine Event Studies and Machine Learning to produce an actionable system that guides trading decisions.

Nowadays, companies, governments, and civilians are moving towards using remote sensing drones for tasks that are too expensive, too risky, or too mundane for humans to do in order to retrieve visual intelligence. With this new age of drones being used for work, it is crucial to understand what goes into designing and constructing sensor systems, and how to build a vision system that preserves image integrity so that it can be successful in supplying data from aerial reconnaissance missions. This work focuses on the development of two such sensor systems, one containing a single camera and the other containing a rigid pair of cameras for implementation in unmanned aerial vehicles (UAVs) for the purpose of geographic information system (GIS) and surveillance applications. Calibration results for the cameras used in each system are given, and 
an analysis of camera capture frequency and synchronization is presented to 
understand how various automated camera trigger methods affect the integrity of image data during UAV flights. 

This thesis presents the design and implementation of high power, wide bandwidth transmit/receive (T/R) switches and modules for use in multi-channel ice-penetrating imaging radars. The switches were designed to address the lack of standard off-the shelf (COTS) devices that meet our technical requirements. 
The design of these switches was accomplished using electronic design automation (EDA) tools and implemented with quadrature hybrids and actively biased PIN diodes. Three different circuits were developed for three different frequency bands: 160-230 MHz (VHF band), 150-600 MHz (VHF/UHF), and 10-45 MHz (HF band). The circuits are capable of transmitting at least 1000 W of peak power and exhibit an insertion loss lower than 1.3 dB for 160-230 MHz, 1.6 dB for 150-600 MHz, and 1.95 dB for 10-45 MHz ranges. A fourth, miniaturized prototype for the 150-600 MHz range was implemented for use in future multi-channel systems. The circuits developed exhibit turn-on times better than 1.3 µs for the VHF/UHF circuits; and 2.1 µs for the HF circuits. The turn-off times were better than 200 ns for the first two bands and 1.36 µs for the HF band. Both the VHF and VHF/UHF have been demonstrated in field operations with two different radar systems. 

Understanding network behavior that undergoes challenges is essential to constructing a resilient and survivable network. Due to the mobility and wireless channel properties, it is more difficult to model and analyze mobile ad hoc networks under various challenges. We provide a comprehensive model to assess the vulnerability of mobile ad hoc networks in face of malicious attacks. We analyze comprehensive graph-theoretical properties and network performance of the dynamic networks under attacks against the critical nodes using both synthetic and real-world mobility traces. Motivated by Minimum Spanning Tree and small-world networks, we propose a network enhancement algorithm by adding long-range links. We compare the performance of different enhancement strategies by evaluating a list of robustness measures. Our study provides insights into the design and construction of resilient and survivable mobile ad hoc networks.

A 10 bit pipeline ADC has been designed using three 4bit SAR stages in pipeline in IBM 180nm CMOS IC Technology using Cadence Spectre simulator. The ADC runs at 20Msamples/sec speed thereby handing signals up to 10MHz bandwidth. The 20Msamples/sec, 10bit ADC is a state of the art design in this class of ADCs at 180nm Technology node. SAR ADCs run at Nyquist rate and they consume lower power (~50fJ/conversion) compared to other popular ADCs – Delta Sigma ADCs(~90fJ/conversion) and Flash ADCs. Secondly SAR ADCs do not employ op-amp or any other block that can’t be easily scaled with technology and hence it is easily portable saving designer’s effort. It therefore becomes an ideal candidate for battery run mobile devices that require intermediate resolution (9-12 bits) and intermediate speed (10-50MS/s). Each SAR stage has a sampler, comparator, SAR logic, Capacitive DAC and synchronizer blocks. The pipeline ADC is built using three SAR stages in pipeline and Residue Amplifiers in between two successive SAR stages. This project goes through the design cycle of the complete ADC- Schematic design, Schematic simulations, Layout and Parasitic extracted simulations.

The least-squares mismatched filter (LS MMF) is a pulse compression method used to suppress range sidelobes. Though initially derived for codes, this work provides a description of the adjustments needed such that the LS MMF can be applied to FM waveforms, a topic that had not previously been published (to the best of our knowledge). Additionally, the effects of range straddling and Doppler on the LS MMF are examined. The effects of straddling on mismatch loss is well known, what is less appreciated is the effect straddling has on the range sidelobes. This work outlines methods that alleviate some of the degradation in sidelobe levels due to straddling. Making the LS MMF more robust to Doppler is also investigated. Adaptive Pulse Compression (APC) is another pulse compression algorithm that has been adjusted to be applicable to FM waveforms. Although the derivation of these adjustments is not part of this work, the analysis via simulation and measured data are. The effects of straddling and Doppler on APC are also investigated, and improvements to APC are analyzed. Lastly, these pulse compression methods are applied to measured data, showing their viability for application in real FM-based systems. 

Optical orthogonal frequency division multiplexing (OFDM) offers high spectral efficiency, resilience to fiber distortion, and simple equalization that make it a suitable technology for next generation optical communication systems. The suitability of optical OFDM to convey data and services in the next generation of optical networks has been extensively investigated for both direct and coherent detection. The key point of OFDM is that all sub-carriers in frequency domain are orthogonal to each other in order to completely eliminate the inter-channel interface (ICI). Nyquist pulse modulation is relatively new technique in optical communication, but the format is very similar to OFDM. It can be derived by simply interchanging time and frequency domain for orthogonal sub-carriers. Therefore, Nuquist pulse modulation could be referred an orthogonal time division multiplexing (OTDM) technique. 
In this project, we investigate the design of a field programmable gate array (FPGA) based optical OFDM modulation and Nyquist pulse modulation transmitters implementing digital signal processing. The transmitters were utilized to generate QAM-OFDM signals and QAM-Nyquist signals. We study the impact of different IFFT algorithms for OFDM and different FIR filter orders for Nyquist on the system performance. In addition to that, we make some comparisons between these two modulation techniques in terms of resource requirements on FPGA, spectral efficiency and peak-to-power ratio. 

Super-Heterodyne receiver is still a predominant receiver architecture used today. In this receiver design one of the most important design trade-offs is the selection of IF frequency. The IF frequency should be low because at the higher (GHz) frequencies the signal processing circuits performs poorly and the cost goes higher. Selectivity of the receiver also affects the IF frequency as the bandwidth of the filter increases with the IF frequency so that the adjacent signals may not get enough attenuation. Another reason which makes selection of IF frequency more complicate is, it should be free from interference and we could achieve this by getting enough attenuation for Image and Murphy signals which creates mixer product terms exactly at the IF filter center frequency. 
In this project an application has been developed in the Mathematica environment which reduces the complexity in rejecting all the Murphy signals and in selecting the IF frequency. And selection of IF frequency using the application is discussed. An interface has been developed with the filter response with Image and all the Murphy signal bands positioned on it. Filter responses are shown for various filter types and orders, as well as Image and Murphy signal bands are shown for Low side, High side and Up conversion tuning solutions with the values of most problematic frequency signals in each band.