tactical CONtext EXtraction (CONEX)
Abstract
Enriching a primary data stream with contextual information (i.e., the circumstances or facts such as who, what, and where that surround a particular event) can be accomplished by fusing data from multiple sensors. For this task, modern systems rely heavily on man-made reference signals, such as Global Positioning Systems (GPS), and preprogrammed algorithms with limited adaptability. Object recognition using Deep Learning and related approaches has been demonstrated, but these methods require significant offline training. The tactical CONtext EXtraction (CONEX) program will develop compact sensors and adaptive processors for extracting contextual information from resource-constrained environments. CONEX sensors will collect information from the landscape and natural sources, such as the relative position of stars, to supplement inertial measurement systems and other sensor feeds in GPS-denied areas. CONEX processors will contain embedded real-time learning algorithms that operate over multiple timescales. These adaptive methods efficiently capture complex spatial and temporal structure in noisy, ambiguous data streams that are beyond the analysis capabilities of state-of-the-art signal/image processing systems.
Document Details
- Document Type
- Accomplishment
- Publication Date
- Oct 01, 2017
- Source ID
- 1a87968e8c0d849305a6d8d9dc299dd1