Non-stationary signal analysis with applications to blind-source processing and imaging
Abstract
Most real-world signals that arise in natural phenomena and from man-made systems are mainly nonlinear and non-stationary. Typical natural signals include meteorological, oceanographic, geological and physiological observations, while man-made signals, that are equally important, include spectral curves of hyperspectral images acquired from airborne and satellite imaging. Unfortunately there is still no rigorous and dependable methods for the divide-and-conquer strategy in separating such signals into their fundamental building blocks for better understanding and visualization, as well as for analysis and manipulation. The goal of our current project is to contribute to this important area by developing a rigorous mathematical theory together with effective methods and efficient algorithms, using data samples that may be continuous-time or discrete samples that may be irregular and contaminated with noise. Our most recent results, including contribution to non-numeric signals, have met various aspects of this challenge, but still require much more time and effort. Potential applications are too numerous to list exhaustively. The obvious ones include multi-directional seismic imaging, non-numeric physiological data for health care, cloud data processing, energy network, commodity trading, real-time analysis of millimeter-wave imaging for TSA, and video imageries for other homeland security applications.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 12, 2017
- Source ID
- W911NF1510385
Entities
People
- Hrushikesh Mhaskar
Organizations
- Army Contracting Command
- Claremont Graduate University
- United States Army