Information Retrieval By Exploiting the Unique Geometry of Sparse Arrays: A New Statistical and Algorithmic Framework
Abstract
The goal of the proposed research is to develop new analytical and algorithmic tools to enhance the capabilities of modern sensor arrays for extracting information from spatio-temporal signatures of impinging electromagnetic waves. Sensor arrays are widely used in a large number of imaging, sensing, and tracking applications, and their performance depends on a number ofphysical factors such as array geometry, aperture, statistical properties of the source scene, noise power, and so forth. A deeper understanding of how these factors interplay towards characterizing fundamental performance limits of modern sensing systems, will allow the development of more powerful, robust and computationally efficient algorithms. Recent advances in sampler design for sensor arrays have produced new insights into how nonuniform array geometry can play a critical role in overcoming long-standing bottlenecks in sourcelocalization. However, the superior performance of these arrays is typically achieved at the cost of using larger number of temporal snapshots. Since practical sensing systems often operate in a limited-sample regime, it is crucial to understand finite sample behavior of these arrays, anddevelop robust algorithms that can overcome possible performance degradations. This project will integrate novel ideas from optimization, random matrix theory and signal representation to analyze performance limits of non-uniform sensor arrays in presence of limited data, and when the number of sources to be localized can potentially exceed the number of sensors. In the latter scenario, it becomes essential to use non-uniform geometry. Our results will pave the way for design of computationally efficient robust algorithms with improved performance guarantees. Our research outcomes will positively impact many applications of sensing and imaging, ranging from radar, sonar to medical imaging. Moreover, the mathematical and computational frameworks go beyond specific applications and can lead to development of new optimization tools and analysis techniques.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 24, 2019
- Source ID
- N000141912227
Entities
People
- Piya Pal
Organizations
- Office of Naval Research
- United States Navy
- University of California, San Diego